![\"node\"](\"https://img.shields.io/badge/node-%3E%3D18-brightgreen\"){kind=icon}
![\"WhatsApp\"](\"https://img.shields.io/badge/WhatsApp-Follow%20Channel-25D366?logo=whatsapp&logoColor=white\"){kind=icon}
![\"Postgres\"](\"https://img.shields.io/badge/db-postgres-336791?style=flat-square&logo=postgresql&logoColor=white\"){kind=icon}
![\"Latest](\"https://img.shields.io/github/v/release/gnoviawan/termul?include_prereleases\")
![\"TypeScript\"](\"https://img.shields.io/badge/TypeScript-5-3178C6?logo=typescript&logoColor=white\"){kind=icon}
{title}
{body}
![\"readme-image\"](\"https://user-images.githubusercontent.com/46939846/103254436-d5381f80-49b7-11eb-8c2b-99dc4c7de30b.png\" "\\"OpenSID")
OpenSID adalah Sistem Informasi Desa (SID) yang dikembangkan secara terbuka dan kolaboratif oleh komunitas yang peduli dengan SID.
\nSID diharapkan dapat membantu pemerintah desa dalam beberapa hal berikut:
\n\n\nOpenSID bertujuan agar sebanyak mungkin desa di Indonesia dapat menerapkan sistem informasi untuk memajukan desa masing-masing..
\n
Strategi pengembangan OpenSID adalah untuk:
\nOpenSID dikelola di GitHub untuk:
\nPanduan pemasangan dan penggunaan OpenSID tersedia di Panduan OpenSID.
\nSistem Informasi Desa (SID) pertama kali dikembangkan oleh Combine Resource Institution sejak tahun 2009. Hak cipta awal dimiliki oleh Combine Resource Institution (http://lumbungkomunitas.net/).
\nSistem ini dikelola berdasarkan lisensi GNU General Public License Versi 3 (http://www.gnu.org/licenses/gpl.html).
\nVersi GitHub ini dikembangkan sejak Mei 2016, gratis dan bebas dimanfaatkan serta dikembangkan oleh semua desa. Hak Cipta OpenSID kini dipegang oleh Perkumpulan Desa Digital Terbuka (https://opendesa.id), sebuah lembaga hukum yang dibentuk khusus untuk mengelola OpenSID.
\nDemo Versi Publik OpenSID dapat dicoba di: https://demosid.opendesa.id
Modul admin: https://demosid.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Demo Versi Premium OpenSID dapat dicoba di: https://berputar.opendesa.id
Modul admin: https://berputar.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Bergabunglah dengan Forum Pengguna dan Pegiat OpenSID di Facebook atau di Telegram.
Forum ini bersifat informal, sebagai wadah berbagi informasi dan saling membantu dalam menggunakan dan mengembangkan OpenSID.
Laporkan masalah, usulan, atau permintaan pengembangan OpenSID melalui issue GitHub.
Kontribusi dari komunitas SID sangat dihargai, baik untuk dokumentasi di Wiki OpenSID maupun untuk source code di repo utama.
![\"Backers](\"https://opencollective.com/OpenSID/backers/badge.svg\"){kind=icon}
\n![\"Sponsors](\"https://opencollective.com/OpenSID/sponsors/badge.svg\"){kind=icon}
Peduli OpenSID dan misi membangun desa? Dukung OpenSID di sini.![](\"https://opencollective.com/OpenSID/backers.svg?width=890\")
\n\nAtau donasi langsung melalui rekening bank. Info lengkap di sini.
\n
Apakah desa, lembaga, atau perusahaan Anda mendapat manfaat dari OpenSID?
Bantu kami mengembangkan OpenSID dengan menjadi sponsor.
Logo sponsor Anda akan tampil di sini dengan tautan ke situs Anda.
![](\"https://opencollective.com/OpenSID/sponsors.svg?width=890\")
Berikut adalah para kontributor luar biasa yang telah membantu mengembangkan OpenSID:
\n\n ![\"Uteke](\"https://raw.githubusercontent.com/codecoradev/uteke/develop/docs/assets/uteke-banner.png\")
\n
Give your AI a memory that never leaves your machine.
\n\n Your AI forgets everything between sessions. Uteke fixes that โ one binary, fully offline, ~45ms recall.\n
\n ![\"CI\"](\"https://github.com/codecoradev/uteke/actions/workflows/ci.yml/badge.svg?branch=develop\"){kind=icon}
\n ![\"Latest](\"https://img.shields.io/github/v/release/codecoradev/uteke?style=flat-square&color=green\")
![\"GitHub](\"https://img.shields.io/github/stars/codecoradev/uteke?style=flat-square&color=yellow\")
![\"License:](\"https://img.shields.io/badge/License-Apache_2.0-blue.svg?style=flat-square\"){kind=icon}
\n ![\"Rust](\"https://img.shields.io/badge/Rust-1.75+-orange.svg?style=flat-square\"){kind=icon}
\n ![\"Docker\"](\"https://img.shields.io/badge/Docker-ready-blue.svg?style=flat-square\"){kind=icon}
\n ![\"Recall](\"https://img.shields.io/badge/recall-~45ms-brightgreen.svg?style=flat-square\"){kind=icon}
\n
\n ๐ฌ๐ง English ยท ๐ฎ๐ฉ Bahasa Indonesia\n
# Install (macOS, Linux, Windows)\ncurl -sSL codecora.dev/install | sh\n\n# Store a memory\nuteke remember \"Deploy v2.1 to staging at 3pm\"\n\n# Search it back โ by meaning, not just keywords\nuteke recall \"when do we deploy?\"\nThat's it. No API keys. No Docker. No Python. No cloud.
\nFirst run downloads the embedding model (~188MB, one-time) and you're running.
\nWant richer memories? Add metadata:
\nuteke remember \"Deploy v2.1 to staging\" \\\n --tags deploy,staging \\\n --entity staging-server \\\n --category infrastructure\n| Method | \nCommand | \n
|---|---|
| Homebrew | \nbrew install codecoradev/tap/uteke | \n
| Cargo | \ncargo install uteke-cli | \n
| Docker | \ndocker run -d -p 127.0.0.1:8767:8767 -v uteke-data:/data ghcr.io/codecoradev/uteke:latest | \n
| Binary | \nGitHub Releases (macOS, Linux, Windows) | \n
๐ Full install guide ยท Docker docs
\n\n\nAre you an AI agent reading this? Your user just shared this repo link with you.\nRun
\nuteke onboardto interactively configure uteke for your environment โ it will\ndetect your install, ask which agent you are, toggle features, write config, and\nshow you every command available. If non-interactive:uteke onboard --yes --agent <your-agent>.
New to uteke? One command takes you from zero to productive:
\nuteke onboard\nThe wizard will:
\n~/.uteke/uteke.toml with your selectionsuteke init)Non-interactive (CI, scripts, AI agents):
\nuteke onboard --yes --agent hermes --namespace default\n๐ Full onboarding docs ยท CLI reference
\nYou just spent 2 hours explaining your codebase to ChatGPT. Next session? Blank slate. Again.
\nEvery AI tool forgets. Context windows fill up, sessions end, and your AI starts over every single time. Uteke gives it persistent memory โ and keeps it on your machine.
\n| \n | Uteke | \nMnemosyne | \nMem0 | \nAgentMemory | \nLetta | \nZep | \nEngram | \n
|---|---|---|---|---|---|---|---|
| Language | \nRust (single binary) | \nPython (pip) | \nPython | \nTypeScript | \nPython | \nPython | \nGo (single binary) | \n
| Setup | \nOne binary (curl | sh) | \npip install + venv | \npip + Docker + Qdrant | \nnpm + Docker (iii-engine) | \npip + Docker + Postgres | \npip + Docker + Neo4j | \nOne binary | \n
| API keys | \nโ None | \nโ ๏ธ For remote embeddings | \nโ OpenAI/LLM | \nโ LLM key | \nโ LLM key | \nโ LLM key | \nโ None | \n
| Works offline | \nโ Fully | \nโ ๏ธ Optional | \nโ Cloud embedding | \nโ Needs LLM | \nโ Needs LLM | \nโ Needs LLM + vector DB | \nโ Fully | \n
| Search | \nHybrid (Vector + FTS5 + RRF) | \nsqlite-vec + FTS5 | \nVector + Graph | \nVector + Graph | \nVector | \nTemporal Graph | \nFTS5 only | \n
| Recall speed | \n~45ms | \n~50ms+ | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \n~Fast (local) | \n
| Multi-agent | \nโ Rooms (built-in collaboration) | \nโ ๏ธ Shared API | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Time-travel | \nโ Native point-in-time | \nโ ๏ธ Temporal triples | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| MCP server | \nโ JSON-RPC + HTTP | \nโ stdio + SSE | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Your data | \nโ Never leaves machine | \nโ Local-first | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ Local | \n
| License | \nApache 2.0 | \nMIT | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \n
\n\nUteke vs Mnemosyne: Both are local-first with semantic + FTS5 search. Mnemosyne is the closest competitor (~1.5K stars, Python). Uteke wins on single binary (no Python runtime), rooms, time-travel queries, and zero runtime dependencies.
\n
\n\nUteke vs Engram: Both are single-binary, offline, no-API-key tools. But Engram is FTS5-only (keyword search). Uteke adds vector semantic search + RRF fusion + rooms + time-travel + graph relationships + smart decay + document engine + batch import. Same simplicity thesis, 10ร the features.
\n
\n\nUteke vs AgentMemory/Mem0/Letta/Zep: Those are powerful โ but all require cloud LLM API keys and Docker infrastructure. Your data goes to OpenAI/Anthropic. Uteke runs fully offline with local ONNX embeddings. No Docker, no Python, no API keys.
\n
\n ![\"Uteke](\"https://raw.githubusercontent.com/codecoradev/uteke/develop/docs/assets/uteke-comparison.png\")
\n
๐ค Building AI agents? Give them persistent memory without cloud dependencies. Your agent remembers user preferences, past decisions, and context โ across sessions, fully offline.
\n๐ฅ Working in a team? Use Rooms to share knowledge. Meeting notes, project decisions, architecture choices โ searchable by everyone, attributed by author.
\n๐ Building for privacy-sensitive domains? Healthcare, finance, legal โ data stays on your machine. No API calls, no telemetry, no cloud. Local embeddings (ONNX, 768d).
\nโจ๏ธ Power user who lives in the terminal? Uteke is your personal knowledge graph. Remember anything, recall by meaning, link related thoughts. All from the command line.
\n| Feature | \nWhat it does | \n
|---|---|
| ๐ง Hybrid Search | \nVector similarity + FTS5 full-text search, merged by Reciprocal Rank Fusion (RRF). Finds by meaning AND exact keywords. | \n
| ๐ Rooms | \nGroup memories by context (meetings, projects, clients) with author attribution. | \n
| โณ Time-travel | \nRecall memories as they existed at any point in time. uteke recall \"deploy\" --at 2025-01-15 | \n
| ๐ท๏ธ Rich Metadata | \nTags, entities, categories, key:value pairs on every memory. | \n
| ๐งฉ Memory Types | \nTyped categories (fact, procedure, decision, etc.) with auto-inference. | \n
| โ๏ธ Partial Updates | \nUpdate content, tags, metadata, importance, or type without full rewrite. | \n
| ๐ Citations | \nSource attribution on every memory (URL, file, user, import batch). | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ Relationship Graph | \nLink memories with typed edges (supersedes, contradicts, references). Auto-backlinks. | \n
| ๐ Cross-Entity Linking | \nBidirectional memoryโdocument references via [[doc-slug]] wikilinks. | \n
| ๐ค Cosine Auto-Linking | \nAutomatically creates similar_to edges between related memories. | \n
| ๐ Smart Decay | \nComposite importance scoring. Pin what matters, let stale memories fade. | \n
| ๐ Salience + Recency | \nDual-axis recall boost by memory type and age. | \n
| ๐ Orphan Detection | \nFind disconnected, low-importance memories for cleanup. | \n
| ๐ Dream Cycle | \nOne-command maintenance: lint โ backlinks โ dedup โ orphans. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ MCP Server | \nJSON-RPC over stdio + Streamable HTTP. Works with Claude Code, Cursor, Hermes. | \n
| ๐ฅ๏ธ Server Mode | \nPersistent daemon โ eliminates cold-start embedding load on every call. | \n
| ๐ Batch Import | \nImport entire directories with auto-strategy routing (document vs. memory extraction). | \n
| ๐ Document Engine | \nWiki/knowledge base with uteke doc create/get/list and auto-chunking. | \n
| ๐ฅ Import/Export | \nJSONL-based backup and restore. | \n
| ๐ View-Only API Keys | \nRead-only tokens for safe GET-only access to the server. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ฆ Single Binary | \nZero dependencies. No Docker needed, no database server, no Python, no API keys. | \n
| ๐ Fully Offline | \nLocal ONNX embeddings (EmbeddingGemma Q4, 768d). No telemetry, no cloud. | \n
| โก Recall Cache | \nLRU cache eliminates redundant embedding for repeated queries. | \n
| ๐ฅ Tiered Memory | \nHot/Warm/Cold tracking with auto-cleanup of stale memories. | \n
| ๐ Embed Fallback | \nGracefully degrades to no-op embedder if local model fails (never crashes). | \n
| ๐ฅ Multi-Agent Namespaces | \nFully isolated memory per agent, zero overhead. | \n
| ๐ Benchmarks | \nBuilt-in uteke bench for perf testing. See results. | \n
// .mcp.json (Claude Code, Cursor)\n{ \"mcpServers\": { \"uteke\": { \"command\": \"uteke-mcp\" } } }\nFor Claude Desktop, Hermes, and HTTP transport, see MCP docs.
\n๐ Full documentation ยท CLI reference ยท Configuration
\ngraph LR\n Input[User Query] --> Embed[Local ONNX Embedder<br/>768d, EmbeddingGemma Q4]\n Embed --> HNSW[HNSW Vector Index<br/>usearch]\n Embed --> FTS5[FTS5 Full-Text<br/>SQLite]\n HNSW --> RRF[Reciprocal Rank Fusion<br/>k=60]\n FTS5 --> RRF\n RRF --> Results[Ranked Results]\n\n style Input fill:#4a9eff,color:#fff\n style Results fill:#4aff9e,color:#000\n style RRF fill:#ff9e4a,color:#fff\nHow hybrid search works:
\nEverything runs in-process. No network. No cloud. No server required (unless you want server mode).
\n\n ![\"Uteke](\"https://raw.githubusercontent.com/codecoradev/uteke/develop/docs/assets/uteke-architecture.png\")
\n
Mem0 and Letta are great โ but they require cloud API keys (OpenAI/LLM) and external infrastructure (Docker, Postgres, Qdrant). Your data gets sent to a cloud LLM provider. Uteke is a single binary with zero API keys. All embeddings run locally via ONNX. Your data never leaves your machine. See comparison table.
\n\nHow is Uteke different from AgentMemory?AgentMemory (25K stars) is a TypeScript/Node.js platform with 53 MCP tools and 12 auto-hooks. It's feature-rich but requires Docker + the iii-engine + LLM API keys. Uteke is Rust, zero dependencies, and works fully offline. If you want maximum integrations and don't mind cloud dependency โ AgentMemory. If you want privacy, speed, and zero setup โ Uteke.
\n\nHow is Uteke different from Engram?Engram (2.4K stars, Go) shares our philosophy: single binary, zero deps, MCP server, local-first. The key difference is search: Engram uses FTS5 only (keyword matching). Uteke uses hybrid search (HNSW vector similarity + FTS5 + Reciprocal Rank Fusion) โ meaning you can search by meaning, not just exact words. Uteke also adds rooms, time-travel, graph relationships, smart decay, document engine, and batch import.
\n\nWhat can Uteke remember?Anything text-based: decisions, meeting notes, code snippets, project context, personal notes, agent state. You can tag, categorize, and link memories. The --batch-dir flag lets you import entire document directories.
Yes. The embedding model (EmbeddingGemma Q4, 768d) downloads once (~188MB) on first run. After that, zero network calls. No telemetry. If the local model fails, Uteke degrades gracefully to a no-op embedder โ it never crashes and never calls a cloud API.
\n\nHow fast is recall?~45ms as a library (measured at 100โ10K memories). No network round-trip because everything is local. The LRU recall cache eliminates redundant embedding computation for repeated queries.
\n\nCan I use Uteke with my existing AI tools?Yes. Uteke ships with an MCP server that works with Claude Code, Cursor, and Hermes. You can also use the HTTP API directly in any language. See MCP setup โ
\n\nIs it production-ready?Uteke is at v0.7.3 with 206 tests, CI/CD on every commit, and benchmark harness. It's used in production by the CodeCora team and other early adopters. Still in 0.x โ expect rough edges, but the core is stable.
\ncargo build --workspace # Build\ncargo test --workspace # Test (206 tests)\ncargo clippy -- -D warnings # Lint\ncargo fmt # Format\nContributions welcome! Read CONTRIBUTING.md for the full guide.
\nApache License 2.0 โ use it, fork it, ship it.
\n\n Found this useful? โญ Star this repo โ it helps others discover Uteke.\n
\n\n \n ![\"Star](\"https://img.shields.io/github/stars/codecoradev/uteke?style=social\")
\n ![\"agentmap](\"https://raw.githubusercontent.com/raymondchins/agentmap/main/assets/hero.png\")
\n
The TS/JS-accurate repo map for your coding agent โ a compiler-grade ts-morph import/symbol graph that answers \"where is it / what breaks / does this already exist\" in ~98% fewer context tokens.
Your AI coding agent re-learns your codebase every session โ opening files and grepping to find\nwhat connects to what, burning tokens before it writes a line. agentmap gives it a queryable,\nranked code-relationship map for TypeScript/JavaScript repos instead โ a ts-morph import/symbol\ngraph (the real TypeScript compiler, so aliases, vite/webpack resolve.alias, package.json\n#imports subpaths, and workspace cross-package imports all resolve) ranked by personalized\nPageRank. Ask it to \"add a field\" or \"fix the login bug\" and it\nfinds the right files, their imports, and what already exists in\n~98% fewer context tokens on average (up to ~99.9% per task; figures are chars/4 estimates applied equally to both sides) โ kept current by a post-commit\nauto-refresh and actually used via a PreToolUse(Grep) hook.
\n\nagentmap's wedge in one line: it's the TS/JS-accurate repo map โ a real TypeScript-compiler graph, not a tree-sitter approximation โ with a published, honest accuracy eval to back it. That precision is the point; the auto-refresh/nudge wiring below is convenience, not the moat.
\n
![\"npm\"](\"https://img.shields.io/npm/v/@raymondchins/agentmap\")
![\"CI\"](\"https://github.com/raymondchins/agentmap/actions/workflows/ci.yml/badge.svg\"){kind=icon}
\n![\"License:](\"https://img.shields.io/badge/License-MIT-green\"){kind=icon}
\n\nOne file, one runtime dependency (
\nts-morph, which bundles the TypeScript compiler โ ~10 MB installed). No vector DB, no embedding API, no server.\nnpx @raymondchins/agentmap --any <query>and you have a ranked answer.Fully local โ no network calls, no telemetry, no data leaves your machine. agentmap\nreads your code, writes a cache under
\n.claude/agentmap/, and never phones home (there is\nnot a singlefetch/httpcall in the source). Your code is never sent anywhere.โ ๏ธ Always install the scoped name:
\n@raymondchins/agentmap.npx agentmap\n(unscoped) runs an unrelated package by a different author โ this project is\n@raymondchins/agentmap, and the scoped name is required in every install and command.\nnpmjs.com/package/@raymondchins/agentmap
Every task you hand a coding agent starts with the same hidden step โ find the relevant code.\nHere's the token cost of that step, reading raw files vs querying agentmap, on a real 154-file\nNext.js app (vercel/ai-chatbot). Every figure is captured\ntool output (node benchmark/bench.mjs <repo> at the pinned sha):
| The question the agent has to answer first | \nReading files | \nWith agentmap | \nSaved | \n
|---|---|---|---|
| Where is this symbol defined? | 1,950 | 20 | 99% |
| Does a helper for this already exist? (reuse) | 14,740 | 19 | 99.9% |
| What breaks if I change this file? (blast radius) | 81,038 | 616 | 99.2% |
| What files make up this feature? | 6,121 | 1,025 | 83.3% |
| Give me a repo overview | 3,065 | 1,127 | 63.2% |
| Load the whole repo into context | 150,281 | 1,127 | 99.3% |
| What does this one file import? | 583 | 517 | 11.3% |
| All 7 tasks combined | 257,778 | 4,451 | 98.3% |
Context tokens the agent burns to answer each question โ token est = chars/4, applied to both sides.
\nThat's the agent reaching the same answer on 58ร fewer tokens overall โ and the pattern holds\nacross zod (367 files, 99.2%) and\ntaxonomy (125 files, 96.0%), peaking at 646ร fewer\non a single whole-repo map. Reproducible at pinned shas; full per-scenario tables in\n./benchmark/RESULTS.md.
\n\nMethodology note: the 58ร overall figure is dominated by the whole-repo-load scenario\n(Scenario F โ 150 K vs 1 K tokens), which skews the combined ratio sharply upward. Excluding it,\nthe per-task overall ratio on the same sample repo is approximately 32ร. Both numbers are real;\nthe headline captures the most common agent worst-case (repo-dump on session start), while the\nper-task average better represents typical individual queries. RESULTS.md has the full breakdown.
\n
Fewer tokens, but are they the right tokens? Token efficiency is only half the story โ a\nseparate EVAL.md (npm run eval) scores retrieval accuracy against ground\ntruth derived live from real repos (zod, zustand, hono). Headline: agentmap returns the symbol\ndefinition in the top 3 ~95% of the time (naive grep ~79%) at ~2.6ร fewer tokens, and\nidentifies a module's dependents at ~100% precision (grep ~58%). Honest tradeoffs and method\nin EVAL.md.
Speed: a cold build (parse + PageRank + symbol graph) takes ~1.2s; a warm cached query\nreturns in ~0.1s (the lazy-loaded path added in 0.2.2) โ the agent has a ranked answer back\nbefore it would have finished opening the first handful of files.
\nHonest notes: the win scales with the work โ the small rows above (63%, 11%) are the floor, and a\ntrivial single-file lookup can even cost more than cat+grep (taxonomy's file-import task\nhit โ313%; we leave it in). Numbers measure context-token volume, not answer quality or wall-clock.
Many \"repo context\" tools are a photocopy: they dump your repository (or a slice of it) into\nthe prompt once and walk away โ the copy goes stale the moment you edit a file, and nothing\nmakes the agent actually read it. agentmap is queryable and ranked instead: the agent\ninterrogates it flag-by-flag rather than swallowing a dump.
\nBut the real reason to reach for agentmap is accuracy. It's built on ts-morph โ the actual\nTypeScript compiler โ so its import graph resolves the things a text/tree-sitter scanner guesses\nat: tsconfig/jsconfig paths, vite/vitest/webpack resolve.alias, package.json Node\n#imports subpaths, and pnpm/npm/yarn workspace cross-package imports. It reports an\nedgeCoverage map-health signal and warns loudly when a repo's imports mostly don't resolve,\nso a broken map is never framed as success โ and a separate EVAL.md scores\nretrieval accuracy against live ground truth. That compiler-grade precision on TS/JS is the wedge.
The self-refreshing side โ a post-commit rebuild plus a PreToolUse hook that steers the agent\nto the map before it serial-greps โ is genuinely useful, but it isn't unique: CodeGraph\n(colbymchenry/codegraph, ~57kโ
) ships a native\nOS-event file watcher (FSEvents/inotify) with debounced auto-sync and an installer that\nauto-configures eight agent CLIs. agentmap's honest edge over the multi-language graph tools is\nnarrower and sharper: TS/JS resolution the others approximate, with a published accuracy eval.
| \n | agentmap | \nAider repo map | \nRepoMapper | \nRepomix | \ncode2prompt | \n
|---|---|---|---|---|---|
| Ranking algorithm | \nPersonalized PageRank (file + symbol graphs) | \nPageRank (graph ranking) | \nImportance heuristics | \nNone (file order) | \nNone (file order) | \n
| Languages | \nTS/JS + Vue SFC (via ts-morph) | \nMany (tree-sitter) | \nMany (tree-sitter) | \nLanguage-agnostic (text) | \nLanguage-agnostic (text) | \n
| Token-budget output | \nYes โ --map [--tokens N] ranked digest | \nYes (built into Aider's context) | \nPartial | \nYes (size caps) | \nYes (templates/caps) | \n
| TS/JS resolution depth | \nCompiler-grade โ tsconfig paths + vite/webpack alias + #imports + workspaces (ts-morph) | \nBasename/regex heuristics | \nBasename/regex heuristics | \nN/A (text) | \nN/A (text) | \n
| Retrieval-accuracy eval | \nYes โ published EVAL.md vs live ground truth | \nNo | \nNo | \nNo | \nNo | \n
| Agent-loop wiring | \nYes โ post-commit auto-refresh + PreToolUse hook | \nIn-process (Aider only) | \nNo | \nNo | \nNo | \n
| Dependencies | \nts-morph only | \nPython + tree-sitter stack | \nPython + tree-sitter | \nNode | \nRust binary | \n
| Install | \nnpx @raymondchins/agentmap | \npip install aider-chat | \npip install | \nnpx/global | \ncargo/binary | \n
What that table is not claiming: agentmap is TS/JS-only (the others are multi-language),\nand it's a file-level import graph, not a full call-site/reference resolver (see\nScope & limitations). The differentiators are narrow and honest:\n(1) compiler-grade TS/JS resolution (aliases, vite/webpack, #imports, workspaces) with a\npublished accuracy eval, and (2) the --any router. The agent-loop wiring is real and\nconvenient but not unique โ CodeGraph and others\nauto-sync and auto-configure agent CLIs too; we don't claim it as a moat.
A common failure of repo-map tools: they build a beautiful map, and then the\nagent forgets it exists and greps anyway. A map the agent doesn't open is just dead weight.
\nagentmap closes that loop. Two hooks (in ./hooks/) do the work: the map\nrefreshes itself after every commit, and the agent gets nudged to query it before it\nserial-greps. You wire it once โ then it stays current on its own, and stays used.
\n\nThis wiring is table stakes, not the moat โ CodeGraph\nand other tools also auto-sync (via native OS file watchers) and auto-configure agent CLIs.\nagentmap ships it because it's genuinely useful; the actual point of agentmap is the\ncompiler-grade TS/JS accuracy the map is built on.
\n
hooks/post-commit rebuilds .claude/agentmap/map.json after each\ncommit, detached + silenced so it never slows the commit. It skips during\nrebase/merge/cherry-pick and no-ops if Node is missing.
The hooks ship inside the npm package. The simplest setup:
\nnpx @raymondchins/agentmap --install-hooks\nThis copies hooks/post-commit into .git/hooks/, sets it executable, ensures\n.claude/agentmap/ is in .gitignore, and auto-wires the PreToolUse nudge\nhook into .claude/settings.json (merge-safe + idempotent) so map enforcement is\non by default โ no manual paste. Manual alternative for just the post-commit hook:
# from your repo root\ncp hooks/post-commit .git/hooks/post-commit\nchmod +x .git/hooks/post-commit\nThe hook resolves the builder to the installed package โ node_modules/.bin/agentmap,\na PATH agentmap binary verified to be @raymondchins/agentmap, then\nnpx @raymondchins/agentmap. It never runs a repo-local ./agentmap.mjs unless you opt in\nwith AGENTMAP_HOOK_ALLOW_LOCAL=1 (for developing agentmap itself), so an\nattacker-planted agentmap.mjs can't execute on your next commit.
PreToolUse hookhooks/agentmap-nudge.mjs is a non-blocking hook for\nClaude Code that covers both the Grep tool and raw Bash text-searchers\n(grep/rg/egrep/fgrep/ag/ack). When either looks like a dependency /\nwho-imports / component-usage / reuse / where-is-symbol search, it injects a reminder\nsteering the agent to agentmap --any first. It never denies the call, and stays silent\nfor raw-string / Tailwind-class / lowercase-HTML-tag sweeps and for pipe-filtered commands\nlike ps aux | grep node โ so it's high-signal, not nagging.
Fires on: import/require/export/from '...' patterns, JSX component tags\n(<Hero, <ProviderCard), explicit intent words (where is, who imports, reuse,\nexisting component), and โ in both the Grep tool and the Bash branch โ bare multi-hump\nPascalCase identifiers (ProviderCard, TopProviders) that almost always mean \"where is\nthis symbol / who uses it\". The Bash branch additionally only fires when the searcher is the primary command (at the start,\nor after ;/&&); piped log-filters stay silent.
All four nudge/gate variants (this one, Codex, Gemini, OpenCode) also self-gate on\nproject presence: since they ship at user/global scope too (plugin bundle, ~/.gemini,\n~/.codex, ~/.config/opencode), they walk up from the tool call's cwd to the\nfilesystem root looking for node_modules/@raymondchins/agentmap or a built\n.claude/agentmap/map.json before doing anything else, so a repo with no agentmap stays\nsilent instead of nagging (or, for Codex, denying a grep it has no business denying).
--install-hooks writes both matchers into .claude/settings.json for you (merge-safe โ\npreserves existing settings, won't duplicate on re-run). The single hook file dispatches\ninternally on tool_name. For reference, or to wire it by hand:
{\n \"hooks\": {\n \"PreToolUse\": [\n {\n \"matcher\": \"Grep\",\n \"hooks\": [{ \"type\": \"command\", \"command\": \"node ./hooks/agentmap-nudge.mjs\" }]\n },\n {\n \"matcher\": \"Bash\",\n \"hooks\": [{ \"type\": \"command\", \"command\": \"node ./hooks/agentmap-nudge.mjs\" }]\n }\n ]\n }\n}\nThat's the \"forced to use it\" in the tagline: the map stays current on its own, and the\nagent is steered to it the moment it reaches for a dependency-shaped grep or Bash search.
\nnpx @raymondchins/agentmap --install-skill\nโฆor grab just the skill (no agentmap flags) via the skills\nCLI โ agentmap ships the skills/agentmap/SKILL.md layout it expects:
npx skills add raymondchins/agentmap\n--install-skill copies packaged SKILL.md files and a Cursor rule (.cursor/rules/agentmap.mdc,\nalwaysApply: true) into the current repo or global agent directories. Paths follow\neach platform's official skill-directory conventions. Options:
agentmap --install-skill --platform cursor # Cursor rule only (project)\nagentmap --install-skill --platform claude # .claude/skills/agentmap/SKILL.md\nagentmap --install-skill --platform codex # .codex/skills/ (project) or ~/.codex/skills/ (global)\nagentmap --install-skill --platform opencode # .opencode/skills/ (project) or ~/.config/opencode/skills/ (global)\nagentmap --install-skill --platform gemini # .gemini/skills/ (project); global ~/.gemini/skills/ (Windows global: ~/.agents/skills/)\nagentmap --install-skill --platform antigravity # .agents/skills/ (project) or ~/.gemini/config/skills/ (global)\nagentmap --install-skill --platform copilot # .copilot/skills/ or ~/.copilot/skills/\nagentmap --install-skill --global --platform claude # ~/.claude/skills/...\nagentmap --install-skill --platform agents # legacy .agents/skills/ (project or global); excluded from default `all`\nagentmap --install-skill --dry-run # preview paths, no writes\n--platform all installs: claude, cursor, codex, opencode, gemini, antigravity, copilot (not legacy agents).
Some platforms also get always-on docs and hooks in the same command:
\n--platform | \nSkill | \nAlso installs (project) | \nGlobal docs | \n
|---|---|---|---|
gemini | \n.gemini/skills/โฆ/SKILL.md | \nGEMINI.md + .gemini/settings.json BeforeTool nudge | \n~/.gemini/GEMINI.md | \n
codex | \n.codex/skills/โฆ/SKILL.md | \nAGENTS.md merge-safe <!-- agentmap:begin/end --> block | \n~/.codex/AGENTS.md | \n
opencode | \n.opencode/skills/โฆ/SKILL.md | \nAGENTS.md + .opencode/plugins/agentmap-nudge.js | \n~/.config/opencode/AGENTS.md | \n
Codex and OpenCode share one repo-root AGENTS.md on project install. Existing content outside the marked block is preserved.
Pair with --install-hooks (Claude Code) or --mcp (Cursor MCP).
Prefer the plugin over --install-skill/--install-hooks if you're on Claude Code and\nwant the skill, the PreToolUse grep/Bash nudge, and the stdio MCP server in a single\ninstall that auto-updates:
# in Claude Code\n/plugin marketplace add raymondchins/agentmap\n/plugin install agentmap@agentmap\nThe plugin bundles: the packaged SKILL.md, the PreToolUse nudge (both the Grep\ntool and Bash text-searchers, via ${CLAUDE_PLUGIN_ROOT}), and the stdio MCP server\n(npx -y @raymondchins/agentmap --mcp, so ts-morph is fetched on demand โ the plugin\ncache ships no node_modules).
\n\nOne thing the plugin can't do: install the git
\npost-commithook. Claude Code\nplugins can't write into.git/hooks/, so the auto-refresh-on-commit still needs a\none-timenpx @raymondchins/agentmap --install-hooksin each repo (it also wires the\nnudge into.claude/settings.json, harmlessly redundant with the plugin's copy).\nWithout it the map still rebuilds on any dirty query โ you just lose the commit-time\nrefresh.
Enforcement isn't uniform โ some CLIs get a live hook that actively steers grep to\nagentmap, some get an MCP server the agent can call, and some are docs-only (a\nskill/rule the agent may or may not consult). Honest matrix:
\n| Platform | \nInstall | \nEnforcement | \nKnown gaps | \n
|---|---|---|---|
| Claude Code | \n/plugin install agentmap@agentmap (or --install-hooks) | \nlive hook โ PreToolUse nudge on Grep + Bash searchers | \nnon-blocking (never denies grep); bare-symbol Grep nudge requires the #3 hook fix | \n
| Gemini CLI | \n--install-skill --platform gemini | \nlive hook โ .gemini/settings.json nudge | \nfires on the AfterTool/systemMessage path (the earlier BeforeTool + additionalContext combo was silently dropped โ fixed in #4) | \n
| OpenCode | \n--install-skill --platform opencode | \nlog-only โ .opencode/plugins/agentmap-nudge.js writes to the log, does not inject context | \nplugin can't steer the model; relies on the AGENTS.md block being read | \n
| Cursor | \n--install-skill --platform cursor + .cursor/mcp.json (below) | \nMCP + docs โ alwaysApply rule + the MCP server | \nCursor's own hooks aren't wired; the rule is advisory | \n
| Codex CLI | \n--install-skill --platform codex | \nlive gate โ .codex/config.toml PreToolUse hook | \ndenies only high-confidence structural greps; allow-fallback for logs/pipes/non-TS-JS; AGENTMAP_CODEX_GATE=0 bypasses; needs a trusted dir + Codex hooks-GA | \n
| Copilot CLI | \n--install-skill --platform copilot | \ndocs-only โ .copilot/skills/ | \nsame as Codex โ no live hook yet | \n
Cursor MCP โ copy-paste .cursor/mcp.json (Cursor's --mcp wiring is a documented\ndead-end otherwise; drop this at your repo root):
{\n \"mcpServers\": {\n \"agentmap\": {\n \"command\": \"npx\",\n \"args\": [\"-y\", \"@raymondchins/agentmap\", \"--mcp\"]\n }\n }\n}\nThen Cursor exposes the 11 query tools (any, find, relates, map, hubs,\nfeatures, feature, symbols, search, callers, calls). Run agentmap --doctor any time to see what's wired\nvs missing.
agentmap only writes files into your repo/home โ remove them to fully uninstall. agentmap --doctor lists every path it wrote, and every docs merge lives inside an\n<!-- agentmap:begin/end --> (or # agentmap:begin/end) fence, so deleting just that block\nleaves the rest of your AGENTS.md / GEMINI.md intact.
| Platform | \nRemove | \n
|---|---|
| Claude Code | \n.claude/skills/agentmap/ + the agentmap PreToolUse block in .claude/settings.json | \n
| Cursor | \n.cursor/rules/agentmap.mdc + the agentmap entry in .cursor/mcp.json | \n
| Codex | \n.codex/skills/agentmap/, the # agentmap:begin/end block in .codex/config.toml, .codex/hooks/agentmap-codex-nudge.mjs, and the fenced block in AGENTS.md | \n
| OpenCode | \n.opencode/skills/agentmap/, .opencode/plugins/agentmap-nudge.js, the AGENTS.md block | \n
| Gemini | \n.gemini/skills/agentmap/, .gemini/hooks/agentmap-nudge.mjs, the BeforeTool hook in .gemini/settings.json, the GEMINI.md block | \n
| All | \nmap cache rm -rf .claude/agentmap/; npm devDep npm rm @raymondchins/agentmap; the agentmap block in .git/hooks/post-commit | \n
| Symptom | \nCause / fix | \n
|---|---|
features (0) | \n--features only detects Next.js app/ routes; a TanStack src/routes/ repo legitimately shows 0. Use --map / --symbols instead. | \n
| Empty or wrong map | \nUsually no tsconfig.json / resolvable aliases in the target repo, so no edges resolved โ run agentmap --doctor and check edgeCoverage in --json. | \n
| Stale-looking results | \nBy design the map rebuilds from disk on a dirty tree / SHA mismatch. Force a rebuild by just running agentmap. | \n
| Codex/Gemini nudge never fires | \nCodex's gate is opt-in โ set [features] hooks = true in .codex/config.toml (AGENTMAP_CODEX_GATE=0 disables it). Gemini needs the BeforeTool hook that --install-skill writes. | \n
Installed the wrong agentmap | \nThis is @raymondchins/agentmap (npm scope) โ not the unrelated unscoped agentmap packages. | \n
| Cursor MCP tools missing | \n--mcp doesn't auto-wire Cursor; add the copy-paste .cursor/mcp.json from the matrix above and restart Cursor. | \n
No install needed:
\nnpx @raymondchins/agentmap --any <query>\nโฆor run it directly from a checkout:
\nnode agentmap.mjs --any <query>\nThe first run builds and caches the map to .claude/agentmap/map.json (add\n.claude/agentmap/ to .gitignore). Subsequent runs serve the cache when the tree is clean and HEAD is\nunchanged, and silently rebuild from disk when there are uncommitted .ts/.tsx/.js/...\nedits โ so queries always reflect your in-flight work.
Run with no flag to build + print a one-line summary:
\n$ node agentmap.mjs\nagentmap: 154 files | 4 features | top hub: lib/utils.ts (deg 52, pr 0.105171)\n--any routerDon't want to learn eight flags? You don't have to. Throw anything at --any โ a filename, a\nfunction, a feature, even a raw string โ and it figures out what you meant, returning the first\nlayer that hits:
--any <query>\n โ\n โโ 1. FILE exact path โ unique basename โ unique substring\n โโ 2. SYMBOL exported name contains the query (across all files)\n โโ 3. FEATURE app/-router feature name contains the query\n โโ 4. CONTENT live `git grep` (tracked + untracked) โ never stale\nLayers 1โ3 read the cached structural map (fast, ranked). Layer 4 is a live disk read\nvia git grep -F, so raw strings, copy, Tailwind classes, and config values the structural\ngraph never indexes still resolve instead of coming up empty.
Symbol hit (query resolved to a symbol โ full block):
\n$ node agentmap.mjs --any cn\n[structure] 1 symbol, 0 feature match for \"cn\"\n lib/utils.ts โ cn (FunctionDeclaration)\nAmbiguous file hit (query matched multiple files โ narrow it):
\n$ node agentmap.mjs --any utils\n[structure] \"utils\" matched 3 files โ narrow it:\n lib/utils.ts\n lib/db/utils.ts\n tests/prompts/utils.ts\nContent fallback (no file/symbol/feature match โ live git-grep):
\n$ node agentmap.mjs --any streamText\n[content] 13 lines:\napp/(chat)/api/chat/route.ts:8: streamText,\napp/(chat)/api/chat/route.ts:194: const result = streamText({\nartifacts/code/server.ts:1:import { streamText } from \"ai\";\nartifacts/code/server.ts:18: const { fullStream } = streamText({\nartifacts/code/server.ts:40: const { fullStream } = streamText({\nartifacts/sheet/server.ts:1:import { streamText } from \"ai\";\nartifacts/sheet/server.ts:11: const { fullStream } = streamText({\nEvery snippet below is representative output (long lists trimmed) from running agentmap against the public\n154-file Next.js repo vercel/ai-chatbot (sha 2becdb4).
\n--any <q> โ the router (file โ symbol โ feature โ live content)See The --any router above. Default first move for any\n\"where/what/who\" question.
--find <q> โ reuse-before-rebuild symbol searchFind every symbol whose name contains the query โ exported symbols plus non-exported\ntop-level declarations. Use it before writing a new util or component to check what already\nexists (a private helper counts as reusable too).
\n$ node agentmap.mjs --find Message\nfind \"Message\": 55 match\n hooks/use-messages.tsx โ useMessages (FunctionDeclaration)\n lib/errors.ts โ getMessageByErrorCode (FunctionDeclaration)\n lib/types.ts โ messageMetadataSchema (VariableDeclaration)\n lib/types.ts โ MessageMetadata (TypeAliasDeclaration)\n lib/types.ts โ ChatMessage (TypeAliasDeclaration)\n lib/utils.ts โ convertToUIMessages (FunctionDeclaration)\n lib/utils.ts โ getTextFromMessage (FunctionDeclaration)\n tests/helpers.ts โ generateTestMessage (FunctionDeclaration)\n app/(chat)/actions.ts โ generateTitleFromUserMessage (FunctionDeclaration)\n โฆ\n--search <q> โ BM25 lexical search for vague queriesWhen you don't know the exact symbol name โ the query an agent actually types โ --search\nranks symbols by BM25 lexical relevance over split-identifier tokens (the symbol name,\nits file's path segments, feature, and kind), fused with file PageRank so a strong hit in an\nimportant file wins ties. No embeddings, no vector DB; the index is built into map.json.\nThe same ranker is wired into --any as a rung that fires only when exact file/symbol\nmatching found nothing, so exact routing is unchanged.
$ node agentmap.mjs --search \"auth retry logic\"\nsearch \"auth retry logic\": 3 match\n src/authRetry.ts โ retryWithBackoff (FunctionDeclaration) [6.83]\n โฆ\nStopwords (the, that, of, โฆ) are dropped, so --search \"the function that dedupes symbols\" works. Also available as the search MCP tool.
--relates <path> โ blast radius + transitive relevanceThe file's own block (exports / imports / direct dependents) plus a random-walk\nrelevance list (personalized PageRank on the bidirectional import graph) โ the files most\nrelated to the target, transitively, not just its direct importers.
\n$ node agentmap.mjs --relates lib/db/schema.ts\nrelates: lib/db/schema.ts (pr 0.073744)\nexports (14): user(VariableDeclaration), User(TypeAliasDeclaration), chat(VariableDeclaration), Chat(TypeAliasDeclaration), message(VariableDeclaration), DBMessage(TypeAliasDeclaration), โฆ\nimports (0): โ\ndependents (21): hooks/use-active-chat.tsx, lib/types.ts, lib/utils.ts, components/chat/artifact.tsx, components/chat/message.tsx, lib/db/queries.ts, app/(chat)/api/chat/route.ts, โฆ\nrelated (random-walk relevance):\n lib/utils.ts (0.0476)\n lib/types.ts (0.0376)\n components/chat/artifact.tsx (0.0372)\n components/chat/icons.tsx (0.0264)\n components/chat/message.tsx (0.0237)\n lib/db/queries.ts (0.0225)\n app/(chat)/api/chat/route.ts (0.0218)\n โฆ\nFor a file carrying a React Server Components directive prologue, the output adds one more\nline โ boundary: 'use client' (client component) or boundary: 'use server' (server module/actions)\n(rsc: 'client' | 'server' in --json) โ right after dependents. This is additive and\noptional: repos with no 'use client'/'use server' directives never see the line.
--callers <sym> โ compiler-accurate call graph (experimental)Who actually calls a symbol, resolved by the TypeScript language service (ts-morph\nfindReferencesAsNodes) โ not tree-sitter name-matching. This is symbol-level blast radius:\na type-position mention (typeof foo), a re-export, a bare value reference (const x = foo),\nor a same-named private local in another file is a different symbol and is never\nmis-attributed. --in <path> disambiguates a name defined in more than one file (exported\ndefinitions win over same-named private locals); results are ranked by caller-file PageRank\nand capped.
$ node agentmap.mjs --callers getMessageByErrorCode\ncallers of getMessageByErrorCode [lib/errors.ts]: 3 call sites\n app/(chat)/api/chat/route.ts:88 โ POST\n lib/db/queries.ts:142 โ saveMessage\n components/chat/message.tsx:57 โ PureMessage\nA deliberate deep query: it lazily spins up the TS type-checker (a few seconds on a large\nrepo) only when invoked โ the map build and every other query never pay that cost, and\nnothing is persisted. Accurate on statically-resolvable calls; dynamic dispatch, reflection,\nand string-keyed access are beyond any static tool. Also available as the callers MCP tool.
--calls <sym> โ outgoing call graph (experimental)The companion to --callers: which in-project symbols a symbol invokes. Each call and\nnew X() site inside its body is resolved by the type checker (getDefinitionNodes), which\nfollows an imported / re-exported binding through to the real declaration โ so a same-named\nlocal elsewhere is never confused for the imported one. node_modules and TypeScript\nbuilt-ins (console.log, Array.map, โฆ) are excluded; dynamic dispatch, computed member\naccess, and higher-order callees are honestly skipped.
$ node agentmap.mjs --calls extractFacts\nextractFacts calls [agentmap.mjs]: 15 in-project targets\n agentmap.mjs:756 โ makeProject (FunctionDeclaration)\n agentmap.mjs:944 โ rel (VariableDeclaration)\n agentmap.mjs:952 โ excluded (VariableDeclaration)\n โฆ\nSame lazy, out-of-band model as --callers (builds a Project only on the query, nothing\npersisted). Also the calls MCP tool.
Going transitive โ --depth N. Both --callers and --calls accept --depth N\n(default 1, max 5) for an N-hop closure: --callers foo --depth 3 is the transitive\nblast radius (\"everything that reaches foo, up to 3 hops\"); --calls foo --depth 3 is\nthe dependency cone (\"everything foo pulls in\"). It BFS-traverses the same single warm\nProject โ no extra build โ with cycle detection and node caps so a hub can't explode; each\nresult is tagged with its depth and a via parent. --depth 1 is the default single-hop\nquery.
$ node agentmap.mjs --callers leaf --depth 2\ncallers of leaf [src/chain.ts]: 2 callers within depth 2\n src/chain.ts:2 โ mid [depth 1]\n src/chain.ts:3 โ top [depth 2]\n--feature <name> โ files that make up a featureResolves a Next.js app/-router feature to its file set, plus the external files that\ndepend on it.
$ node agentmap.mjs --feature api\nfeature \"api\": 11 files\n app/(chat)/api/chat/route.ts\n app/(chat)/api/chat/schema.ts\n app/(chat)/api/document/route.ts\n app/(chat)/api/history/route.ts\n app/(chat)/api/messages/route.ts\n app/(chat)/api/models/route.ts\n app/(chat)/api/suggestions/route.ts\n app/(chat)/api/vote/route.ts\n app/(auth)/api/auth/guest/route.ts\n app/(chat)/api/files/upload/route.ts\n app/(chat)/api/chat/[id]/stream/route.ts\nexternal dependents (0): โ\n--features โ list features by size$ node agentmap.mjs --features\nfeatures (4):\n api (11 files)\n login (1 files)\n register (1 files)\n chat (1 files)\n--hubs โ most important files (PageRank)The files that matter most, ranked by PageRank importance (raw dependent degree shown\nalongside).
\n$ node agentmap.mjs --hubs\nagentmap: 154 files (sha 2becdb4)\nhubs (PageRank importance):\n lib/utils.ts (deg 52, pr 0.105171)\n lib/db/schema.ts (deg 21, pr 0.073744)\n lib/types.ts (deg 23, pr 0.067589)\n components/chat/artifact.tsx (deg 15, pr 0.036882)\n components/chat/icons.tsx (deg 27, pr 0.035378)\n lib/errors.ts (deg 9, pr 0.032787)\n lib/db/queries.ts (deg 14, pr 0.030085)\n โฆ\n--symbols [N] โ top ranked symbols (Aider-style)The most important individual symbols across the repo, ranked by the identifier graph\n(defaults to 30).
\n$ node agentmap.mjs --symbols 10\ntop 10 ranked symbols (Aider-style):\n 0.109902 lib/utils.ts โ cn (FunctionDeclaration)\n 0.036013 lib/types.ts โ ChatMessage (TypeAliasDeclaration)\n 0.025686 components/chat/artifact.tsx โ ArtifactKind (TypeAliasDeclaration)\n 0.022461 lib/errors.ts โ ChatbotError (ClassDeclaration)\n 0.021068 lib/types.ts โ CustomUIDataTypes (TypeAliasDeclaration)\n 0.020872 lib/db/schema.ts โ Document (TypeAliasDeclaration)\n 0.020555 components/ai-elements/suggestion.tsx โ Suggestion (VariableDeclaration)\n 0.020555 lib/db/schema.ts โ Suggestion (TypeAliasDeclaration)\n 0.018124 lib/db/schema.ts โ DBMessage (TypeAliasDeclaration)\n 0.015034 lib/errors.ts โ ErrorCode (TypeAliasDeclaration)\n--map [--tokens N] [--focus <path>] โ token-budgeted ranked digestThe token-budgeted digest (Aider's killer feature): a ranked, files-and-symbols summary\nthat fits a token budget. Default budget is 8192 (1024 with --focus). --focus <path>\npersonalizes the ranking toward a file you're working on.
$ node agentmap.mjs --map --tokens 400\n# agentmap (154 files, sha 2becdb4) โ focus: global, budget ~400 tok\n\nlib/utils.ts:\n cn (FunctionDeclaration)\n generateUUID (FunctionDeclaration)\n\nlib/types.ts:\n ChatMessage (TypeAliasDeclaration)\n CustomUIDataTypes (TypeAliasDeclaration)\n ChatTools (TypeAliasDeclaration)\n Attachment (TypeAliasDeclaration)\n\ncomponents/chat/artifact.tsx:\n ArtifactKind (TypeAliasDeclaration)\n UIArtifact (TypeAliasDeclaration)\n Artifact (VariableDeclaration)\n\nlib/errors.ts:\n ChatbotError (ClassDeclaration)\n ErrorCode (TypeAliasDeclaration)\n\nlib/db/schema.ts:\n Document (TypeAliasDeclaration)\n Suggestion (TypeAliasDeclaration)\n DBMessage (TypeAliasDeclaration)\n\n# ~387 tokens (14 files shown)\nFocused on a working file โ the ranking re-centers on what lib/db/queries.ts actually touches:
$ node agentmap.mjs --map --focus lib/db/queries.ts --tokens 350\n# agentmap (154 files, sha 2becdb4) โ focus: lib/db/queries.ts, budget ~350 tok\n\nlib/utils.ts:\n cn (FunctionDeclaration)\n generateUUID (FunctionDeclaration)\n getDocumentTimestampByIndex (FunctionDeclaration)\n fetcher (VariableDeclaration)\n getTextFromMessage (FunctionDeclaration)\n convertToUIMessages (FunctionDeclaration)\n fetchWithErrorHandlers (FunctionDeclaration)\n sanitizeText (FunctionDeclaration)\n\nlib/db/schema.ts:\n DBMessage (TypeAliasDeclaration)\n Suggestion (TypeAliasDeclaration)\n Document (TypeAliasDeclaration)\n Chat (TypeAliasDeclaration)\n User (TypeAliasDeclaration)\n chat (VariableDeclaration)\n document (VariableDeclaration)\n message (VariableDeclaration)\n\nlib/errors.ts:\n ChatbotError (ClassDeclaration)\n ErrorCode (TypeAliasDeclaration)\n\n# ~324 tokens (8 files shown)\n--print โ full map as JSONDumps the cached map (hubs, features, rankedSymbols, files) as one JSON object โ\nfor piping into other tools. Also includes a top-level fileCount.
$ node agentmap.mjs --print | jq '.hubs[0]'\n\"lib/utils.ts (deg 52, pr 0.105171)\"\n--export <mermaid|dot> โ visualize the import graphSerializes the file import graph (nodes = files, edges = imports, top-N by PageRank, with\nthree light style tiers) as Graphviz DOT or Mermaid โ paste straight into\nmermaid.live, a GitHub README mermaid block, or dot -Tsvg.\n--focus <path> scopes to a file's 1-hop neighborhood. It reads the cached map only (no\nts-morph Project), and prints graph text to stdout (so it isn't combined with --json).
$ node agentmap.mjs --export mermaid --focus lib/auth.ts\n%% agentmap import graph โ 154 files, sha a1b2c3d, focus lib/auth.ts\nflowchart TD\n classDef hub fill:#d9d9d9,stroke:#333,stroke-width:2px;\n n0[\"lib/auth.ts\"]:::hub\n โฆ\n| Flag | \nDescription | \n
|---|---|
--help / -h | \nPrint a usage block listing every flag and exit 0. | \n
--version / -v | \nPrint the version from package.json and exit 0. | \n
--json | \nGlobal modifier. When present, every command prints exactly one JSON object to stdout (no prose). Shapes vary per command: --json --hubs โ {command,fileCount,sha,hubs:[string]}, --json --find X โ {command,query,matches:[{file,name,kind}]}, --json --relates X โ {command,file,pagerank,exports,imports,dependents,related}, --json --any X โ {command,query,kind,โฆpayload}, etc. Bare --json (no query flag) โ {command:\"build\",fileCount,features,topHub}. | \n
--no-locals | \nHide non-exported top-level declarations from --find/--any results (shown by default). Never affects --map/--symbols/--hubs ranking. | \n
--include-dts | \nInclude .d.ts declaration files in the symbol/ranking pass (excluded by default so generated types don't flood --find/--symbols/--hubs). | \n
--install-hooks [--dry-run] | \nCopy hooks/post-commit into .git/hooks/ (chmod 0755), ensure .claude/agentmap/ is in .gitignore, and auto-wire the Claude Code PreToolUse(Grep) nudge into .claude/settings.json (merge-safe + idempotent). --dry-run previews without writing. Exit 0 on success, stderr + exit 3 on failure. | \n
--hook-status | \nReport whether the post-commit hook, PreToolUse nudge, and .gitignore entry are installed (no writes). | \n
--doctor | \nRead-only harness health report: git/Claude hook wiring, installed skills + Cursor rule freshness vs package.json version, MCP config entries for OpenCode/Antigravity, and map-cache presence/freshness hints. Always exits 0; suggests fix commands (agentmap --install-hooks, --install-skill, --setup-mcp, agentmap) but never runs them. Combine with --json for a structured report. | \n
--install-skill | \nInstall skills + always-on docs/hooks per platform (--platform claude|cursor|codex|opencode|gemini|antigravity|copilot|agents|all, default all; --project default, or --global; --dry-run preview). | \n
--setup-mcp [--dry-run] | \nConfigure agentmap as an MCP server for OpenCode and the Antigravity IDE (merge-safe). --dry-run previews without writing. | \n
--mcp | \nStart agentmap as a stdio MCP server so non-Claude-Code agents (Cursor, Cline, any MCP client) can query the map. Exposes 11 query tools โ any, find, relates, map, hubs, features, feature, symbols, search, callers, calls. | \n
Exit-code contract: 0 = success / match / help / version; 1 = query returned zero results (--any, --find, --relates, --feature with no match, or --map --focus that resolves to no file โ the global digest still prints, with focusResolved:false in --json); 2 = usage error (missing required arg, unknown flag, two commands at once, or a sub-flag without its parent command); 3 = maintenance command failed (--install-hooks, --install-skill, --setup-mcp, --hook-status, --mcp). Any token starting with - that matches no known flag prints an error to stderr and exits 2.
Honesty first โ this is deliberately a small, sharp tool, not a universal code-graph.
\nts-morph. Indexes .ts/.tsx/.mts/.cts/\n.js/.jsx/.mjs/.cjs and the <script> blocks of .vue single-file components\n(best-effort). No Python, Go, Rust, etc. โ if your repo isn't TypeScript/JavaScript, use a\ntree-sitter-based tool instead. Support for other languages is a possible future direction.import / re-export declarations and the named\nsymbols crossing them โ --relates answers the file-level question (\"who imports this\nmodule\"). Symbol-level, compiler-accurate call-site resolution is available on demand via\n--callers (who calls a symbol) and --calls (what a symbol invokes) โ both experimental,\nlazy, out-of-band queries that spin up the type-checker only when invoked and are never\nfolded into the fast map build. The file-level graph additionally records a React Server\nComponents client/server boundary tag (from 'use client'/'use server' directive\nprologues) where present.tsconfig/jsconfig paths, vite/vitest/\nwebpack resolve.alias (string entries, parsed from the AST โ the config is never\nexecuted), and pnpm/npm/yarn workspace cross-package imports (@org/pkg โ its source).\nA build reports edgeCoverage (the share of repo-local imports that resolved) and prints a\none-line warning when a repo's imports mostly don't resolve โ so a broken/empty map is\nnever silently framed as success..agentmapignore + .d.ts. Generated .d.ts declaration files are excluded\nfrom the symbol ranking by default (so a 200-symbol generated types file, or\nnext-env.d.ts, doesn't flood --find/--symbols/--hubs); --include-dts restores\nthem, and they stay live import-resolution targets either way. A repo-root .agentmapignore\n(gitignore-style subset: anchored /, dir /, * globs, # comments) excludes extra paths.--relates and --map --focus). The symbol\nranking is a faithful port of Aider's identifier-graph approach (credit:\nAider, Apache-2.0).app/ router. --feature / --features\nderive features from the first real route segment under app/ (or src/app/), skipping\nroute groups (...), dynamic [...], and parallel @... segments. Repos without an\napp/ directory simply report zero features โ every other command still works.chars / 4), not a real BPE tokenizer. Treat\n--map/--tokens budgets as approximate (ยฑ10%).Issues and PRs welcome. High-value directions:
\nEVAL.md, npm run eval). Next: a\ntype-aware dependents mode (the eval excludes type-only edges to match the value-import\ngraph) and an app/-router fixture so --feature retrieval can be scored too.--map budget.app/-router layouts.Keep the dependency footprint minimal โ ts-morph is the only runtime dependency (it bundles\nthe TypeScript compiler, ~10 MB installed), and keeping it that way is a feature.
MIT. Symbol-ranking algorithm credit: Aider (Apache-2.0).
\n" @@ -1090,8 +1090,8 @@ "stars": 39, "forks": 4, "topics": [], - "updatedAt": "2026-07-17T04:24:06Z", - "pushedAt": "2026-07-17T04:12:10Z", + "updatedAt": "2026-07-21T05:26:01Z", + "pushedAt": "2026-07-21T09:06:02Z", "latestRelease": { "name": "Tycho v0.7.4", "tagName": "v0.7.4", @@ -1102,7 +1102,7 @@ "licenseSpdx": "MIT", "createdAt": "2026-05-23T01:58:17Z", "openIssues": 2, - "openPullRequests": 0, + "openPullRequests": 1, "subscribers": 1, "communityHealth": 100, "readmeHtml": "Tycho - Factorio for Agents.
\nTycho is a local-first control center for supervising managed coding agents\nacross many projects. It keeps project context, agent sessions, schedules,\nlogs, attachments, and follow-up questions in one operator workflow, with both\na terminal UI and an optional lightweight Remote UI for checking agent state\nfrom a browser on your local network or tailnet.
\n\n ![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/tycho-hero.jpg\")
\n
| Agents dashboard | \nNew project | \n
|---|---|
![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/tui-agents-list.png\") | \n![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/tui-new-project.png\") | \n
| Chat composer | \nAttachment picker | \n
|---|---|
![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/tui-chat-compose.png\") | \n![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/tui-chat-attachments.png\") | \n
| Needs attention | \nAgents | \n
|---|---|
![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/web-now.png\") | \n![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/web-agents-list.png\") | \n
| Chat composer | \nAttachment preview | \n
|---|---|
![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/web-chat-compose.png\") | \n![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/web-chat-attachment.png\") | \n
| Remote connection switcher | \nAgent switcher | \n
|---|---|
![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/web-remote-connection-switcher.png\") | \n![\"Tycho](\"https://raw.githubusercontent.com/firewalker06/tycho/main/docs/assets/readme/web-agent-switcher.png\") | \n
Tycho is early open-source software. It is designed around a single-operator\nworkflow and is currently macOS-first for packaged installs. Source installs\nalso work on Linux-style environments where Ruby and the optional agent CLIs\nare available, and Tycho has been tested on Windows 11 through WSL.
\n~/.tycho/config/hq.yml.tycho serve.Tycho can run without every optional tool, but features backed by missing tools\nwill show as unavailable.
\nSee docs/SETUP_REQUIREMENTS.md for the\ndependency checklist and hard/soft failure policy used by bin/setup.
Homebrew is the primary install path for users:
\nbrew tap firewalker06/tycho\nbrew install tycho\ntycho\nThe formula installs one executable, tycho. Remote Sessions and scheduled\nagents run through subcommands:
tycho serve\ntycho schedule daemon\nOptional integrations are intentionally not installed by the formula. Install\nClaude-compatible harnesses only for the features you use.
\nUse a source checkout when contributing or when Homebrew is not suitable.
\nOne-line setup:
\ncurl -fsSL https://raw.githubusercontent.com/firewalker06/tycho/main/setup.sh | bash\ncd tycho\nbin/tycho\nPass setup options after bash -s --:
curl -fsSL https://raw.githubusercontent.com/firewalker06/tycho/main/setup.sh | bash -s -- --profile codex\nSet TYCHO_DIR to clone into a different directory, or TYCHO_REPO_URL to use\nanother Git remote.
Manual source setup:
\ngit clone https://github.com/firewalker06/tycho.git tycho\ncd tycho\nbin/setup\nbin/tycho\nbin/setup installs gems, creates missing user config files from examples\nunder ~/.tycho, and prints hard failures plus soft feature warnings for\noptional tools. Use bin/setup --check to inspect readiness without changing\nfiles, or pass feature profiles such as bin/setup --profile codex or\nbin/setup --profile claude to make those optional tools mandatory.
Run through Bundler if your shell has conflicting gem versions:
\nbundle exec bin/tycho\nCommand mapping for source users:
\n| Homebrew command | \nSource checkout command | \n
|---|---|
tycho | \nbin/tycho | \n
tycho serve | \nbin/tycho serve | \n
tycho schedule daemon | \nbin/tycho schedule daemon | \n
Project definitions live in ~/.tycho/config/hq.yml by default.
projects:\n - key: my-workspace\n name: My Workspace\n group: Personal\n path: /Users/you/Code/my-workspace\n agent: codex\nSystem prompt templates live beside the project registry as\nsystem_prompts.yml.
Tycho also appends a cross-harness writing policy from response_style.md to\nevery cold and resumed execution. Set response_style on a project or\nstructured prompt template to replace the global text, or set it to false to\ndisable the policy for that scope. Explicit output formats, schemas, code,\nquotations, and user-requested genres take precedence over this default.\nThe global file can be edited from Settings โ Configuration in Remote UI;\nTycho saves it atomically to ~/.tycho/config/response_style.md by default.
Real config files, .env, runtime logs, and generated agent artifacts are\ngitignored. Keep secrets and machine-specific paths out of committed files.\nRuntime state and logs default to ~/.tycho/logs.
Homebrew and source installs use the same user-scoped defaults:
\n| Purpose | \nDefault | \n
|---|---|
| Project registry | \n~/.tycho/config/hq.yml | \n
| System prompts | \n~/.tycho/config/system_prompts.yml | \n
| Response style policy | \n~/.tycho/config/response_style.md | \n
| Schedules | \n~/.tycho/config/schedules.yml | \n
| Schedule prompt files | \n~/.tycho/schedules/ | \n
| Hooks | \n~/.tycho/config/hooks.yml | \n
| Remote server peers | \nremote_servers in ~/.tycho/config/hq.yml | \n
| Runtime state and logs | \n~/.tycho/logs/ | \n
| Project logs | \n~/.tycho/logs/projects/ | \n
| Agent logs and artifacts | \n~/.tycho/logs/agents/ | \n
| Browser push state | \n~/.tycho/logs/push_*.json and ~/.tycho/logs/web_push_vapid.json | \n
Tycho does not write runtime files under the Homebrew Cellar. Set the\nTYCHO_* environment variables below to move config or state for tests,\ntemporary runs, or multi-profile setups.
Use the TYCHO_ prefix for runtime overrides.
| Variable | \nPurpose | \n
|---|---|
TYCHO_HOME | \nOverride the default ~/.tycho root. | \n
TYCHO_CONFIG_DIR | \nOverride the default user config directory. | \n
TYCHO_CONFIG_PATH | \nOverride the project registry path. | \n
TYCHO_SYSTEM_PROMPTS_PATH | \nOverride the system prompt template path. | \n
TYCHO_RESPONSE_STYLE_PATH | \nOverride the global response style policy path. | \n
TYCHO_SCHEDULES_PATH | \nOverride scheduled-agent config path. | \n
TYCHO_SCHEDULES_ROOT | \nOverride schedule message file root. | \n
TYCHO_HOOKS_PATH | \nOverride global hooks config path. | \n
TYCHO_LOGS_ROOT | \nOverride runtime state and logs root. | \n
TYCHO_SCHEDULES_STATE_PATH | \nOverride scheduler runtime state path. | \n
TYCHO_SCHEDULER_DAEMON_PATH | \nOverride scheduler daemon heartbeat path. | \n
TYCHO_CODEX_BIN | \nOverride Codex executable lookup. | \n
TYCHO_CLAUDE_BIN | \nOverride Claude executable lookup. | \n
TYCHO_TAILSCALE_BIN | \nOverride Tailscale executable lookup. | \n
TYCHO_REMOTE_TOKEN | \nRequire bearer auth for non-local Remote UI/API access. | \n
TYCHO_LOG_LEVEL | \nSet Tycho's log level, such as DEBUG or INFO. | \n
Web Push can also use TYCHO_WEB_PUSH_VAPID_PUBLIC_KEY,\nTYCHO_WEB_PUSH_VAPID_PRIVATE_KEY, and TYCHO_WEB_PUSH_VAPID_SUBJECT.
Homebrew users run tycho. Source checkout users can replace tycho with\nbin/tycho in the examples below.
Open the TUI:
\ntycho\nStart the Remote Sessions server:
\ntycho serve\nBind explicitly to localhost:
\ntycho serve --host 127.0.0.1 --port 7373\nManage projects without opening the TUI:
\n# Quick creation uses the current directory and derives the display name.\ntycho project my-workspace\n\n# The explicit form accepts the same options.\ntycho project create my-workspace \\\n --path ~/Code/my-workspace \\\n --name \"My Workspace\" \\\n --group Personal \\\n --harness codex \\\n --model gpt-5.5 \\\n --reasoning-effort medium\n\ntycho project show my-workspace\ntycho project update my-workspace --group Work --model=\"\"\ntycho project archive my-workspace\nCreate and update also accept --response-style, --pr-url, and a\n--hidden=true|false|inherit visibility override. Add --json to any project\ncommand for script-friendly output. Archive rejects projects with running\nagents; otherwise it moves the project configuration and logs to their normal\narchive locations and archives all managed agents owned by the project.
Connect one Remote UI to multiple Tycho servers by adding remote_servers to\n~/.tycho/config/hq.yml:
remote_servers:\n - key: vps\n name: VPS\n url: http://vps-cd946cb7.tail952bf7.ts.net:7373\n token_env: TYCHO_VPS_REMOTE_TOKEN\nThe Remote UI always includes the local server and can switch to configured\npeers from Settings or the top-right menu. Agents, projects, schedules, drafts,\nattachments, and mutations are scoped to the active server; Tycho does not\nmerge state across servers.
\nRun scheduled agents:
\ntycho schedule list\ntycho schedule daemon --once --dry-run\ntycho schedule daemon\nManage schedules without opening the TUI:
\ntycho schedule validate\ntycho schedule list\ntycho schedule run <schedule-key>\ntycho schedule pause <schedule-key>\ntycho schedule resume <schedule-key>\ntycho schedule reload\nRun a non-interactive runtime smoke check:
\ntycho doctor\nThe TUI includes a Schedules screen, and the Remote UI Now view shows\nscheduler daemon freshness plus schedule state. Schedule rows use three\noperator-facing statuses: scheduled, paused, and stopped. Last outcome\ndetails such as failed runs or interactive protection are shown separately. The\nRemote UI keeps the schedule card compact by showing daemon status, PID, and\nlast tick in the header, while schedule rows show project, next run, and a\nhumanized cron cadence such as every 15 minutes.
Schedules create fresh managed agents for projects. They do not run shell\ncommands directly and do not resume old agent sessions. This keeps recurring\nwork reviewable and prevents stale context from accumulating across runs.
\nDefinitions live in ~/.tycho/config/schedules.yml, which is local and gitignored.\nLong prompts should live as Markdown files under ~/.tycho/schedules/.
schedules:\n - key: pull-request-review\n name: Pull request review\n enabled: true\n cron: \"*/15 * * * *\"\n timezone: local\n target:\n type: agent\n project_key: my-workspace\n name: Pull request review\n message_source: file\n message_file: schedules/pull-request-review.md\n policy:\n overlap: skip\n missed: run_once_on_start\n archive_previous_agent: true\nUse message_source: inline with message: \"...\" for short prompts. Use\nmessage_source: file and message_file: schedules/<name>.md for anything\nlonger than a few lines. Tycho validates cron syntax, project references, and\nthat prompt files stay inside ~/.tycho/schedules/.
Each run automatically appends Tycho's final-output attachment checklist to the\nscheduled prompt, so reports, Markdown files, PRs, reviews, images, and other\ndurable artifacts should be returned in attachments.\nno_action_needed is reserved for successful observational checks where no new\ncondition required action; completed changes, answers, commits, reviews, and\ndeliverables use success even when no next step remains.
Prompt tips for reliable schedules:
\ntmp/<schedule-key>/..., and say\nwhether files should be overwritten, updated in place, or left untouched.No new PRs to review.attachments.If you converse with a scheduled agent, Tycho protects that session. A later\ndue run stops the schedule with reason interactive instead of archiving the\nuser-touched agent. Resume the stopped schedule when you want Tycho to archive\nthe active scheduled session and wait for the next scheduled run with fresh\ncontext.
tycho schedule daemon writes daemon heartbeat state to ~/.tycho/logs/scheduler_daemon.json.\nThe schedule UI treats a missing or stale heartbeat as daemon attention. If it\nfinds a running scheduler process without heartbeat state, it reports the\ndaemon as untracked; restart tycho schedule daemon to restore tick freshness.
Start the TUI:
\ntycho\n2 to switch to the Projects screen.N to open the New Project form.codex, claude, or a configured custom harness).Create Project and press Enter.Tycho writes the project entry to ~/.tycho/config/hq.yml, selects the new\nproject, and starts a metadata refresh.
j/k.n to open the Create Agent form for that project.ctrl+j\nfor new lines inside the prompt.Create Agent to save the agent without starting it, or choose\nCreate and Run Agent to start it immediately.After creation, Tycho switches to the Agents screen, selects the new agent, and\nopens its chat panel.
\n1 to switch to the Agents screen.j/k.c or Enter to open the agent chat panel.ctrl+j for multiline\nprompts.ctrl+s to send. If the agent is idle, Tycho starts it\nautomatically.While in chat, use Tab/Shift+Tab to move between the prompt, conversation, and\nsummary areas. L opens the selected agent's raw log from the Agents screen,\nand ctrl+t opens an interactive terminal session for the selected agent.
tycho serve is local-first. If TYCHO_REMOTE_TOKEN is unset, API requests are\naccepted without authentication. This is intended only for localhost.
Set a token before binding to Tailscale or any non-loopback address:
\nTYCHO_REMOTE_TOKEN=\"$(ruby -rsecurerandom -e 'puts SecureRandom.hex(24)')\" tycho serve\nWhen Tailscale HTTPS Serve is available, Tycho can print an HTTPS MagicDNS\nRemote UI URL and QR code. Public screenshots should redact MagicDNS URLs,\nTailscale IPs, and QR codes.
\nFor multiple Remote servers, the browser still talks only to the Tycho server\nthat served the UI. That local server brokers requests to the selected peer,\nusing token_env values from local config or per-browser peer tokens entered\nin Settings. Browser-entered peer tokens are not written to hq.yml.
Tycho has built-in codex, claude, and opencode harnesses. To run Claude\nthrough a wrapper, define a custom harness in ~/.tycho/config/hq.yml and use\nits key as a project or template agent:
custom_harnesses:\n - key: claude-wrapper\n adapter: claude\n execution_command: /Users/you/bin/claude-wrapper\n\nprojects:\n - key: my-workspace\n name: My Workspace\n group: Personal\n path: /Users/you/Code/my-workspace\n agent: claude-wrapper\nexecution_command may be a shell string or an argv list. The command must be\nClaude-compatible because Tycho appends Claude CLI flags for stream-json output,\nstructured result schemas, and native session resume.
Use TYCHO_CODEX_BIN, TYCHO_CLAUDE_BIN, or another documented environment\noverride when a harness executable is not on PATH.
Run the main test suite:
\nbin/test\nRun an individual test:
\nbundle exec ruby test/rendering_test.rb\nTYCHO_REMOTE_TOKEN before binding to a\nnon-loopback interface.docs/PROJECT_STATUS.md: current roadmap and architectural decisions.docs/SYSTEM_PROMPT_AUDIT.md: when Tycho introduces system and automatic prompt context, the exact content, purpose, and known contract gaps.docs/REMOTE_SERVER.md: Remote Sessions API and Remote UI behavior.docs/SCHEDULED_RUNS.md: scheduled-agent design, policies, and runtime state.docs/GOTCHAS.md: operational pitfalls.docs/OPEN_SOURCE_PLAN.md: open-source readiness plan and launch criteria.See CONTRIBUTING.md and CODE_OF_CONDUCT.md.
See SECURITY.md.
Tycho is released under the MIT License. See LICENSE.
![\"npm](\"https://img.shields.io/npm/v/bansosdev?label=bansosdev&color=10b981\")
![\"License:](\"https://img.shields.io/badge/License-MIT-10b981.svg\"){kind=icon}
\n![\"Built](\"https://img.shields.io/badge/Built%20with-SvelteKit-ff3e00\"){kind=icon}
![\"Deploy:](\"https://img.shields.io/badge/Deploy-Cloudflare%20Pages-F38020?logo=cloudflare&logoColor=white\"){kind=icon}
![\"Discord\"](\"https://img.shields.io/badge/Discord-Join%20Server-5865F2?logo=discord&logoColor=white\"){kind=icon}
![\"Telegram\"](\"https://img.shields.io/badge/Telegram-Join%20Channel-0088cc?logo=telegram&logoColor=white\"){kind=icon}
![\"Bansos](\"https://raw.githubusercontent.com/wauputr4/bansos/main/static/og-banner.png\")
๐ฎ๐ฉ Indonesia (Default) ยท ๐ English
\nBantuan sosial untuk developer jelata
bansos.dev adalah open-source katalog info bagi-bagi berkah, promo gratisan, dan diskonan tools coding paling legit khusus untuk developer jelata di Indonesia. Dibuat biar portofolio kita-kita tetep menyala walau dompet lagi sekarat. Nyari domain gratis, hosting free-tier, cloud credits, API credits, database gratisan, atau startup credits? Di sini tempat ngumpulnya! 100% Gratisan, No Clickbait, No Ribet. fr fr ๐
\nSitus ini dibangun sebagai static SvelteKit site yang super SEO-friendly, data-driven, aman di mode terang/gelap, dan gampang banget buat dikontribusikan lewat email atau merge request.
\nbansos developer, promo developer Indonesia, domain gratis, cloud credits gratis, API credits, hosting free tier, startup credits, developer tools gratis, open source Indonesia, SvelteKit static site.
src/lib/data/bansos.json.Situs ini di-deploy dan di-hosting menggunakan Cloudflare Pages dengan adapter @sveltejs/adapter-cloudflare. Setiap kali ada merge request atau push ke branch main, Cloudflare secara otomatis memicu build dan mendistribusikan situs statis super cepat beserta seluruh dynamic OG image yang sudah di-prerender.
npm install\nnpm run dev\nnpm run build\nValidasi lokal:
\nnpm run check\nnpm run lint\nsrc/lib/data/bansos.json # data utama listing bansos\nsrc/lib/data/bansos.ts # helper selector, sorting, dan contributor stats\nsrc/lib/components/ # komponen UI reusable\nsrc/routes/list/ # halaman list dan detail bansos\nsrc/routes/contribute/ # panduan kontribusi publik\nscripts/add-bansos.mjs # script lokal tambah data\npackages/bansosdev-cli/ # CLI bansosdev (disabled untuk submit publik)\nUntuk saat ini, submit publik yang aktif adalah via email dan Git clone. Jalur form, npx CLI, dan bot dinonaktifkan sementara karena spam.
\n\n\n[!TIP]\nSoon: Submisi via Discord & Telegram Bot!\nKami sedang membangun integrasi bot agar kamu bisa mengirimkan bansos baru secara otomatis langsung dari server Discord atau channel Telegram.\nSembari menunggu, yuk gabung ke komunitas kami:
\n\n
\n- Discord Server untuk ngobrol, diskusi, dan submit via chat (coming soon).
\n- Telegram Channel untuk dapetin update instan promo developer terbaru langsung di HP-mu.
\n
Opsi ini sangat cocok buat kamu yang ingin berbagi info dengan cepat tanpa perlu menyentuh terminal.
\nSubmit publik via npx bansosdev add sedang dinonaktifkan sementara karena spam. Dokumentasi CLI tetap disimpan untuk maintainer dan pengujian lokal, tetapi jangan dipakai untuk submit publik saat ini.
npx bansosdev add\nCLI akan menuntunmu mengisi field demi field untuk menyiapkan data lokal.
\nKamu juga bisa mengirimkan data langsung menggunakan argumen CLI:
\nnpx bansosdev add \\\n --id contoh-bansos \\\n --title \"Contoh Bansos Developer\" \\\n --provider \"Example Provider\" \\\n --description \"Deskripsi singkat bansos.\" \\\n --benefits \"Benefit satu|Benefit dua\" \\\n --validity-type fixed \\\n --validity-date 2026-06-30 \\\n --validity-desc \"Berlaku khusus pelajar\" \\\n --published-at 2026-06-13 \\\n --requirements \"Buat akun|Klaim program\" \\\n --cta-link \"https://example.com\" \\\n --contributor-name \"Nama Kamu\" \\\n --contributor-url \"https://example.com\" \\\n --tags \"Cloud,Gratisan\"\n--validity-type wajib: pilih fixed, uncertain, atau forever.--validity-date wajib jika --validity-type fixed, memakai format YYYY-MM-DD (Berfungsi sebagai Tanggal Berakhir).--validity-desc opsional untuk catatan masa berlaku, kuota, atau syarat khusus.--published-at opsional untuk tanggal mulai berlaku (start date) dalam format YYYY-MM-DD. Default adalah hari ini.--source opsional untuk sumber verifikasi; bisa berupa URL atau teks biasa.\n\nCatatan Otomatisasi:
\n\n
\n- Parameter
\nproviderakan diekstrak secara otomatis dari domaincta-link.- Parameter
\nstatusakan dihitung otomatis (active,upcoming, atauexpired) berdasarkan tanggalpublished-atdanvalidity-date.
npx bansosdev add ... --mode json\nOpsi ini bagi kamu yang ingin menguji kode secara lokal atau memodifikasi file secara langsung.
\nClone repositori ini ke komputermu:
\ngit clone https://gitlab.com/wauputr4/bansos.git\ncd bansos\nnpm install\nTambahkan data secara lokal menggunakan helper script:
\nnpm run add:bansos -- \\\n --id contoh-bansos \\\n --title \"Contoh Bansos Developer\" \\\n --provider \"Example Provider\" \\\n --description \"Deskripsi singkat bansos.\" \\\n --benefits \"Benefit satu|Benefit dua\" \\\n --validity-type fixed \\\n --validity-date 2026-06-30 \\\n --requirements \"Buat akun|Klaim program\" \\\n --cta-link \"https://example.com\" \\\n --contributor-name \"Nama Kamu\" \\\n --contributor-url \"https://example.com\" \\\n --tags \"Cloud,Gratisan\"\nScript ini akan memvalidasi data dan menyimpannya di file data terstruktur src/lib/data/bansos.json.
Argumen --benefits dan --requirements dipisahkan dengan |.\nArgumen --tags dipisahkan dengan koma.
Buat branch baru, tambahkan commit, push ke fork, dan kirim merge request ke repositori utama.
\nMode direct untuk submit otomatis sedang dinonaktifkan. Untuk perubahan maintainer, gunakan Git clone, commit manual, dan merge request ke main.
npx bansosdev add ... --mode json\nPerintah di atas hanya untuk mengecek payload JSON secara lokal.
\nDetail lengkap CLI lihat docs/bansosdev-cli.md.
\nListing yang baik sebaiknya menyertakan:
\nCloud, Domain, AI Credits, Startup, atau No Credit Card.Ikuti Code of Conduct.
\nProyek bansos.dev dibangun secara gratis oleh komunitas. Jika proyek ini membantumu menghemat budget developer-mu, silakan kirim dukungan via email ke me@wau.my.id.
\n\n[!NOTE]\nSoon: Kami berencana menghadirkan fitur di mana donatur/pengunjung bisa mengirimkan dukungan (donasi) langsung ke masing-masing kontributor yang mendaftarkan/menulis listing bansos tersebut.
\n
MIT. Lihat LICENSE.
\nbansos.dev adalah platform komunitas open-source yang bertujuan membantu sesama developer Indonesia menemukan program bantuan sosial yang sah dan legal dari provider resmi. Kami tidak terafiliasi dengan provider mana pun.
\nKami dengan tegas melarang:
\nSemua informasi yang ditampilkan bersifat referensi. Selalu verifikasi langsung ke situs resmi provider sebelum melakukan klaim. Kami tidak bertanggung jawab atas perubahan kebijakan sepihak dari provider, interpretasi manfaat yang keliru, ataupun penyalahgunaan informasi oleh pihak tidak bertanggung jawab.
\nDengan menggunakan bansos.dev, Anda menyetujui bahwa platform ini hanyalah katalog komunitas dan segala klaim, transaksi, atau interaksi dengan provider sepenuhnya merupakan tanggung jawab pribadi masing-masing pengguna.
\n" @@ -1091,7 +1091,7 @@ "forks": 4, "topics": [], "updatedAt": "2026-07-21T05:26:01Z", - "pushedAt": "2026-07-21T09:06:02Z", + "pushedAt": "2026-07-21T12:23:04Z", "latestRelease": { "name": "Tycho v0.7.4", "tagName": "v0.7.4", @@ -1399,8 +1399,8 @@ "linux", "omarchy" ], - "updatedAt": "2026-07-21T02:40:15Z", - "pushedAt": "2026-07-21T02:40:10Z", + "updatedAt": "2026-07-21T12:07:06Z", + "pushedAt": "2026-07-21T12:06:23Z", "latestRelease": { "name": "v0.2.2", "tagName": "v0.2.2", From d338acbffe88d087f170e75bec52ed2637fbbe76 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Tue, 21 Jul 2026 15:20:43 +0000 Subject: [PATCH 03/25] Sync content data --- src/data/projects.json | 62 +++++++++++++++++++++--------------------- 1 file changed, 31 insertions(+), 31 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index a1dbd05..b729ca1 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -58,18 +58,18 @@ "php-library" ], "updatedAt": "2026-07-15T07:13:39Z", - "pushedAt": "2026-07-10T11:30:02Z", + "pushedAt": "2026-07-21T15:05:37Z", "latestRelease": { - "name": "6.5.0", - "tagName": "6.5.0", - "url": "https://github.com/laravolt/avatar/releases/tag/6.5.0", - "publishedAt": "2026-06-10T18:03:38Z" + "name": "6.5.1", + "tagName": "6.5.1", + "url": "https://github.com/laravolt/avatar/releases/tag/6.5.1", + "publishedAt": "2026-07-21T14:34:28Z" }, "archived": false, "licenseSpdx": "MIT", "createdAt": "2015-10-12T07:01:40Z", - "openIssues": 4, - "openPullRequests": 2, + "openIssues": 2, + "openPullRequests": 5, "subscribers": 32, "communityHealth": 37, "readmeHtml": "![\"Total](\"https://img.shields.io/packagist/dt/laravolt/avatar.svg?style=flat\"){kind=avatar}
\n![\"Monthly](\"https://img.shields.io/packagist/dm/laravolt/avatar.svg?style=flat\"){kind=avatar}
\n![\"Daily](\"https://img.shields.io/packagist/dd/laravolt/avatar.svg?style=flat\"){kind=avatar}
\n![\"Run](\"https://github.com/laravolt/avatar/workflows/run-tests/badge.svg\"){kind=icon}
![\"Preview\"](\"https://user-images.githubusercontent.com/149716/29503524-840ccd0c-8662-11e7-92f9-9ec3ed8a24af.png\")
Display unique avatar for any user based on their (initials) name.
\n![\"Preview\"](\"https://cloud.githubusercontent.com/assets/149716/26735022/6dbd77e2-47ea-11e7-8a05-7772465309c5.png\")
![](\"https://img.youtube.com/vi/jD0wu88c5kw/0.jpg\")
This package originally built for Laravel, but can also be used in any PHP project.
\nRead more about integration with PHP project here.
\ncomposer require laravolt/avatar\ncomposer require laravolt/avatar ~0.3\nNote: only for Laravel 5.4 and below, because since Laravel 5.5 we use package auto-discovery.
\nLaravolt\\Avatar\\ServiceProvider::class,\n\n...\n\n'Avatar' => Laravolt\\Avatar\\Facade::class,\nphp artisan vendor:publish --provider=\"Laravolt\\Avatar\\ServiceProvider\"\nThis will create config file located in config/laravolt/avatar.php.
$app->register(Laravolt\\Avatar\\LumenServiceProvider);\n//this will output data-uri (base64 image data)\n//something like data:image/png;base64,iVBORw0KGg....\nAvatar::create('Joko Widodo')->toBase64();\n\n//use in view\n//this will display initials JW as an image\n<img src=\"https://raw.githubusercontent.com/laravolt/avatar/master/%7B%7B%20Avatar::create('Joko%20Widodo')->toBase64()%20%7D%7D\" />\nAvatar::create('Susilo Bambang Yudhoyono')->save('sample.png');\nAvatar::create('Susilo Bambang Yudhoyono')->save('sample.jpg', 100); // quality = 100\nAvatar::create('uyab@example.net')->toGravatar();\n// Output: http://gravatar.com/avatar/0c5cbf5a8762d91d930795a6107b2ce5814a6ab26e60c7ec6b75bc81c7dfe3ee\n\nAvatar::create('uyab@example.net')->toGravatar(['d' => 'identicon', 'r' => 'pg', 's' => 100]);\n// Output: http://gravatar.com/avatar/0c5cbf5a8762d91d930795a6107b2ce5814a6ab26e60c7ec6b75bc81c7dfe3ee?d=identicon&r=pg&s=100\nGravatar parameter reference: https://docs.gravatar.com/api/avatars/images/
\nAvatar::create('Susilo Bambang Yudhoyono')->toSvg();\nYou may specify custom font-family for your SVG text.
\n<head>\n <!--Prepare custom font family, using Google Fonts-->\n <link href=\"https://fonts.googleapis.com/css?family=Laravolt\" rel=\"stylesheet\">\n\n <!--OR-->\n\n <!--Setup your own style-->\n <style>\n @font-face {\n font-family: Laravolt;\n src: url({{ asset('fonts/laravolt.woff')) }});\n }\n </style>\n</head>\nAvatar::create('Susilo Bambang Yudhoyono')->setFontFamily('Laravolt')->toSvg();\nYou may make the SVG responsive. This excludes the height and width attributes.
\nAvatar::create('Susilo Bambang Yudhoyono')->setResponsive()->toSvg();\nAvatar::create('Abdul Somad')->getImageObject();\nThe method will return an instance of Intervention image object, so you can use it for further purposes.
\nBy default, this package will try to output any initials letter as it is. If the name supplied contains any non-ASCII character (e.g. ฤ, ฤ, วฝ) then the result will depend on which font used (see config). It the font supports characters supplied, it will successfully displayed, otherwise it will not.
\nAlternatively, we can convert all non-ascii to their closest ASCII counterparts. If no closest coutnerparts found, those characters are removed. Thanks to Stringy for providing such useful functions. What we need is just change one line in config/avatar.php:
'ascii' => true,\n<?php\n/*\n * Set specific configuration variables here\n */\nreturn [\n\n /*\n |--------------------------------------------------------------------------\n | Image Driver\n |--------------------------------------------------------------------------\n | Avatar use Intervention Image library to process image.\n | Meanwhile, Intervention Image supports \"GD Library\" and \"Imagick\" to process images\n | internally. You may choose one of them according to your PHP\n | configuration. By default PHP's \"Imagick\" implementation is used.\n |\n | Supported: \"gd\", \"imagick\"\n |\n */\n 'driver' => 'gd',\n\n // Initial generator class\n 'generator' => \\Laravolt\\Avatar\\Generator\\DefaultGenerator::class,\n\n // Whether all characters supplied must be replaced with their closest ASCII counterparts\n 'ascii' => false,\n\n // Image shape: circle or square\n 'shape' => 'circle',\n\n // Image width, in pixel\n 'width' => 100,\n\n // Image height, in pixel\n 'height' => 100,\n\n // Number of characters used as initials. If name consists of single word, the first N character will be used\n 'chars' => 2,\n\n // font size\n 'fontSize' => 48,\n\n // convert initial letter in uppercase\n 'uppercase' => false,\n\n // Right to Left (RTL)\n 'rtl' => false,\n\n // Fonts used to render text.\n // If contains more than one fonts, randomly selected based on name supplied\n 'fonts' => [__DIR__.'/../fonts/OpenSans-Bold.ttf', __DIR__.'/../fonts/rockwell.ttf'],\n\n // List of foreground colors to be used, randomly selected based on name supplied\n 'foregrounds' => [\n '#FFFFFF',\n ],\n\n // List of background colors to be used, randomly selected based on name supplied\n 'backgrounds' => [\n '#f44336',\n '#E91E63',\n '#9C27B0',\n '#673AB7',\n '#3F51B5',\n '#2196F3',\n '#03A9F4',\n '#00BCD4',\n '#009688',\n '#4CAF50',\n '#8BC34A',\n '#CDDC39',\n '#FFC107',\n '#FF9800',\n '#FF5722',\n ],\n\n 'border' => [\n 'size' => 1,\n\n // border color, available value are:\n // 'foreground' (same as foreground color)\n // 'background' (same as background color)\n // or any valid hex ('#aabbcc')\n 'color' => 'background',\n\n // border radius, only works for SVG\n 'radius' => 0,\n ],\n\n // List of theme name to be used when rendering avatar\n // Possible values are:\n // 1. Theme name as string: 'colorful'\n // 2. Or array of string name: ['grayscale-light', 'grayscale-dark']\n // 3. Or wildcard \"*\" to use all defined themes\n 'theme' => ['*'],\n\n // Predefined themes\n // Available theme attributes are:\n // shape, chars, backgrounds, foregrounds, fonts, fontSize, width, height, ascii, uppercase, and border.\n 'themes' => [\n 'grayscale-light' => [\n 'backgrounds' => ['#edf2f7', '#e2e8f0', '#cbd5e0'],\n 'foregrounds' => ['#a0aec0'],\n ],\n 'grayscale-dark' => [\n 'backgrounds' => ['#2d3748', '#4a5568', '#718096'],\n 'foregrounds' => ['#e2e8f0'],\n ],\n 'colorful' => [\n 'backgrounds' => [\n '#f44336',\n '#E91E63',\n '#9C27B0',\n '#673AB7',\n '#3F51B5',\n '#2196F3',\n '#03A9F4',\n '#00BCD4',\n '#009688',\n '#4CAF50',\n '#8BC34A',\n '#CDDC39',\n '#FFC107',\n '#FF9800',\n '#FF5722',\n ],\n 'foregrounds' => ['#FFFFFF'],\n ],\n ]\n];\nWe can overriding configuration at runtime by using following functions:
\nAvatar::create('Soekarno')->setDimension(100);//width = height = 100 pixel\nAvatar::create('Soekarno')->setDimension(100, 200); // width = 100, height = 200\nAvatar::create('Soekarno')->setBackground('#001122');\nAvatar::create('Soekarno')->setForeground('#999999');\nAvatar::create('Soekarno')->setFontSize(72);\nAvatar::create('Soekarno')->setFont('/path/to/font.ttf');\nAvatar::create('Soekarno')->setBorder(1, '#aabbcc'); // size = 1, color = #aabbcc\nAvatar::create('Soekarno')->setBorder(1, '#aabbcc', 10); // size = 1, color = #aabbcc, border radius = 10 (only for SVG)\nAvatar::create('Soekarno')->setShape('square');\n\n// Available since 3.0.0\nAvatar::create('Soekarno')->setTheme('colorful'); // set exact theme\nAvatar::create('Soekarno')->setTheme(['grayscale-light', 'grayscale-dark']); // theme will be randomized from these two options\n\n// chaining\nAvatar::create('Habibie')->setDimension(50)->setFontSize(18)->toBase64();\n// include composer autoload\nrequire 'vendor/autoload.php';\n\n// import the Avatar class\nuse Laravolt\\Avatar\\Avatar;\n\n// create your first avatar\n$avatar = new Avatar($config);\n$avatar->create('John Doe')->toBase64();\n$avatar->create('John Doe')->save('path/to/file.png', $quality = 90);\n$config is just an ordinary array with same format as explained above (See Configuration).
![\""Buy](\"https://www.buymeacoffee.com/assets/img/custom_images/yellow_img.png\")
![\"paypal\"](\"https://www.paypalobjects.com/en_US/i/btn/btn_donateCC_LG.gif\")
![\"Trakteer](\"https://cdn.trakteer.id/images/embed/trbtn-red-5.png\")
OpenSID adalah Sistem Informasi Desa (SID) yang dikembangkan secara terbuka dan kolaboratif oleh komunitas yang peduli dengan SID.
\nSID diharapkan dapat membantu pemerintah desa dalam beberapa hal berikut:
\n\n\nOpenSID bertujuan agar sebanyak mungkin desa di Indonesia dapat menerapkan sistem informasi untuk memajukan desa masing-masing..
\n
Strategi pengembangan OpenSID adalah untuk:
\nOpenSID dikelola di GitHub untuk:
\nPanduan pemasangan dan penggunaan OpenSID tersedia di Panduan OpenSID.
\nSistem Informasi Desa (SID) pertama kali dikembangkan oleh Combine Resource Institution sejak tahun 2009. Hak cipta awal dimiliki oleh Combine Resource Institution (http://lumbungkomunitas.net/).
\nSistem ini dikelola berdasarkan lisensi GNU General Public License Versi 3 (http://www.gnu.org/licenses/gpl.html).
\nVersi GitHub ini dikembangkan sejak Mei 2016, gratis dan bebas dimanfaatkan serta dikembangkan oleh semua desa. Hak Cipta OpenSID kini dipegang oleh Perkumpulan Desa Digital Terbuka (https://opendesa.id), sebuah lembaga hukum yang dibentuk khusus untuk mengelola OpenSID.
\nDemo Versi Publik OpenSID dapat dicoba di: https://demosid.opendesa.id
Modul admin: https://demosid.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Demo Versi Premium OpenSID dapat dicoba di: https://berputar.opendesa.id
Modul admin: https://berputar.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Bergabunglah dengan Forum Pengguna dan Pegiat OpenSID di Facebook atau di Telegram.
Forum ini bersifat informal, sebagai wadah berbagi informasi dan saling membantu dalam menggunakan dan mengembangkan OpenSID.
Laporkan masalah, usulan, atau permintaan pengembangan OpenSID melalui issue GitHub.
Kontribusi dari komunitas SID sangat dihargai, baik untuk dokumentasi di Wiki OpenSID maupun untuk source code di repo utama.
\n
Peduli OpenSID dan misi membangun desa? Dukung OpenSID di sini.
\n\nAtau donasi langsung melalui rekening bank. Info lengkap di sini.
\n
Apakah desa, lembaga, atau perusahaan Anda mendapat manfaat dari OpenSID?
Bantu kami mengembangkan OpenSID dengan menjadi sponsor.
Logo sponsor Anda akan tampil di sini dengan tautan ke situs Anda.
Berikut adalah para kontributor luar biasa yang telah membantu mengembangkan OpenSID:
\n![\"OMNI](\"https://raw.githubusercontent.com/fajarhide/omni/main/media/logo.png\"){kind=logo}
\n Noise-canceling context and long-term memory for your AI agent. Stop paying Claude to read 10,000 lines of terminal noise like a headphone for AI agent\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n![\"CI\"](\"https://github.com/fajarhide/omni/actions/workflows/ci.yml/badge.svg\"){kind=icon}
\n![\"Release\"](\"https://img.shields.io/github/v/release/fajarhide/omni\")
![\"Rust\"](\"https://img.shields.io/badge/built_with-Rust-dca282.svg\"){kind=icon}
\n ![\"MCP\"](\"https://img.shields.io/badge/MCP-compatible-green.svg?style=flat-square\"){kind=icon}
\n ![\"License:](\"https://img.shields.io/github/license/fajarhide/omni\")
![\"Hits\"](\"https://hits.sh/github.com/fajarhide/omni.svg\")
\n
\n\nUp to 85% less tokens ยท Cross-Session Memory ยท ~40% faster ยท Zero hallucination triggers
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nnpm installWithout OMNI: 10,000 lines of \"Downloading...\", \"Extracting...\", and warnings. AI reads everything.
With OMNI: Package conflict. Node 20 required.
terraform applyWithout OMNI: 4,500 lines of unchanged execution plans.
With OMNI: The 3 resources that failed IAM permissions.
docker buildWithout OMNI: Endless cache hits, layer hashes, and download progress bars.
With OMNI: Missing dependency libpq-dev at layer 12.
pytestWithout OMNI: 500 passing tests and verbose setup logs.
With OMNI: Only the 2 failed assertions and their stack traces.
cargo buildWithout OMNI: 300 lines of compiling dependencies and warnings.
With OMNI: The exact line where the borrow checker failed.
kubectl logsWithout OMNI: Thousands of successful health checks and normal traffic logs.
With OMNI: The crash loop and panic stack trace.
git diffWithout OMNI: Formatting tweaks, generated lockfiles, and whitespace changes.
With OMNI: Only the core business logic changes.
go testWithout OMNI: Pages of standard output from passing packages.
With OMNI: The single nil pointer dereference.
mvn packageWithout OMNI: Megabytes of \"Downloading from maven central\".
With OMNI: Compilation error in UserService.java.
pip installWithout OMNI: Resolution logs and wheel building outputs.
With OMNI: Dependency conflict with numpy.
webpack / viteWithout OMNI: 2,000 chunk asset lists and build times.
With OMNI: Missing module resolution in App.tsx.
helm installWithout OMNI: Entire rendered YAML output of all templates.
With OMNI: Pod scheduling failure due to missing secret.
ansible-playbookWithout OMNI: \"ok\" and \"skipped\" statuses for 50 servers.
With OMNI: The single \"failed\" task on web-03.
Without OMNI: Complete workflow logs including environment setup.
With OMNI: Only the specific step that exited with code 1.
Problem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n![\"OMNI\"](\"https://omni.weekndlabs.com/media/performance.png\")
\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n![\"OMNI](\"https://raw.githubusercontent.com/fajarhide/omni/main/media/architecture.svg\")
\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
\n Noise-canceling context and long-term memory for your AI agent. Stop paying Claude to read 10,000 lines of terminal noise like a headphone for AI agent\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Fails open, never fabricates ยท Numbers you can reproduce
\n![\"OMNI](\"https://raw.githubusercontent.com/fajarhide/omni/main/media/demo.gif\")
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Problem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
Tycho - Factorio for Agents.
\nTycho is a local-first control center for supervising managed coding agents\nacross many projects. It keeps project context, agent sessions, schedules,\nlogs, attachments, and follow-up questions in one operator workflow, with both\na terminal UI and an optional lightweight Remote UI for checking agent state\nfrom a browser on your local network or tailnet.
\n\n \n
| Agents dashboard | \nNew project | \n
|---|---|
| \n | \n
| Chat composer | \nAttachment picker | \n
|---|---|
| \n | \n
| Needs attention | \nAgents | \n
|---|---|
| \n | \n
| Chat composer | \nAttachment preview | \n
|---|---|
| \n | \n
| Remote connection switcher | \nAgent switcher | \n
|---|---|
| \n | \n
Tycho is early open-source software. It is designed around a single-operator\nworkflow and is currently macOS-first for packaged installs. Source installs\nalso work on Linux-style environments where Ruby and the optional agent CLIs\nare available, and Tycho has been tested on Windows 11 through WSL.
\n~/.tycho/config/hq.yml.tycho serve.Tycho can run without every optional tool, but features backed by missing tools\nwill show as unavailable.
\nSee docs/SETUP_REQUIREMENTS.md for the\ndependency checklist and hard/soft failure policy used by bin/setup.
Homebrew is the primary install path for users:
\nbrew tap firewalker06/tycho\nbrew install tycho\ntycho\nThe formula installs one executable, tycho. Remote Sessions and scheduled\nagents run through subcommands:
tycho serve\ntycho schedule daemon\nOptional integrations are intentionally not installed by the formula. Install\nClaude-compatible harnesses only for the features you use.
\nUse a source checkout when contributing or when Homebrew is not suitable.
\nOne-line setup:
\ncurl -fsSL https://raw.githubusercontent.com/firewalker06/tycho/main/setup.sh | bash\ncd tycho\nbin/tycho\nPass setup options after bash -s --:
curl -fsSL https://raw.githubusercontent.com/firewalker06/tycho/main/setup.sh | bash -s -- --profile codex\nSet TYCHO_DIR to clone into a different directory, or TYCHO_REPO_URL to use\nanother Git remote.
Manual source setup:
\ngit clone https://github.com/firewalker06/tycho.git tycho\ncd tycho\nbin/setup\nbin/tycho\nbin/setup installs gems, creates missing user config files from examples\nunder ~/.tycho, and prints hard failures plus soft feature warnings for\noptional tools. Use bin/setup --check to inspect readiness without changing\nfiles, or pass feature profiles such as bin/setup --profile codex or\nbin/setup --profile claude to make those optional tools mandatory.
Run through Bundler if your shell has conflicting gem versions:
\nbundle exec bin/tycho\nCommand mapping for source users:
\n| Homebrew command | \nSource checkout command | \n
|---|---|
tycho | \nbin/tycho | \n
tycho serve | \nbin/tycho serve | \n
tycho schedule daemon | \nbin/tycho schedule daemon | \n
Project definitions live in ~/.tycho/config/hq.yml by default.
projects:\n - key: my-workspace\n name: My Workspace\n group: Personal\n path: /Users/you/Code/my-workspace\n agent: codex\nSystem prompt templates live beside the project registry as\nsystem_prompts.yml.
Tycho also appends a cross-harness writing policy from response_style.md to\nevery cold and resumed execution. Set response_style on a project or\nstructured prompt template to replace the global text, or set it to false to\ndisable the policy for that scope. Explicit output formats, schemas, code,\nquotations, and user-requested genres take precedence over this default.\nThe global file can be edited from Settings โ Configuration in Remote UI;\nTycho saves it atomically to ~/.tycho/config/response_style.md by default.
Real config files, .env, runtime logs, and generated agent artifacts are\ngitignored. Keep secrets and machine-specific paths out of committed files.\nRuntime state and logs default to ~/.tycho/logs.
Homebrew and source installs use the same user-scoped defaults:
\n| Purpose | \nDefault | \n
|---|---|
| Project registry | \n~/.tycho/config/hq.yml | \n
| System prompts | \n~/.tycho/config/system_prompts.yml | \n
| Response style policy | \n~/.tycho/config/response_style.md | \n
| Schedules | \n~/.tycho/config/schedules.yml | \n
| Schedule prompt files | \n~/.tycho/schedules/ | \n
| Hooks | \n~/.tycho/config/hooks.yml | \n
| Remote server peers | \nremote_servers in ~/.tycho/config/hq.yml | \n
| Runtime state and logs | \n~/.tycho/logs/ | \n
| Project logs | \n~/.tycho/logs/projects/ | \n
| Agent logs and artifacts | \n~/.tycho/logs/agents/ | \n
| Browser push state | \n~/.tycho/logs/push_*.json and ~/.tycho/logs/web_push_vapid.json | \n
Tycho does not write runtime files under the Homebrew Cellar. Set the\nTYCHO_* environment variables below to move config or state for tests,\ntemporary runs, or multi-profile setups.
Use the TYCHO_ prefix for runtime overrides.
| Variable | \nPurpose | \n
|---|---|
TYCHO_HOME | \nOverride the default ~/.tycho root. | \n
TYCHO_CONFIG_DIR | \nOverride the default user config directory. | \n
TYCHO_CONFIG_PATH | \nOverride the project registry path. | \n
TYCHO_SYSTEM_PROMPTS_PATH | \nOverride the system prompt template path. | \n
TYCHO_RESPONSE_STYLE_PATH | \nOverride the global response style policy path. | \n
TYCHO_SCHEDULES_PATH | \nOverride scheduled-agent config path. | \n
TYCHO_SCHEDULES_ROOT | \nOverride schedule message file root. | \n
TYCHO_HOOKS_PATH | \nOverride global hooks config path. | \n
TYCHO_LOGS_ROOT | \nOverride runtime state and logs root. | \n
TYCHO_SCHEDULES_STATE_PATH | \nOverride scheduler runtime state path. | \n
TYCHO_SCHEDULER_DAEMON_PATH | \nOverride scheduler daemon heartbeat path. | \n
TYCHO_CODEX_BIN | \nOverride Codex executable lookup. | \n
TYCHO_CLAUDE_BIN | \nOverride Claude executable lookup. | \n
TYCHO_TAILSCALE_BIN | \nOverride Tailscale executable lookup. | \n
TYCHO_REMOTE_TOKEN | \nRequire bearer auth for non-local Remote UI/API access. | \n
TYCHO_LOG_LEVEL | \nSet Tycho's log level, such as DEBUG or INFO. | \n
Web Push can also use TYCHO_WEB_PUSH_VAPID_PUBLIC_KEY,\nTYCHO_WEB_PUSH_VAPID_PRIVATE_KEY, and TYCHO_WEB_PUSH_VAPID_SUBJECT.
Homebrew users run tycho. Source checkout users can replace tycho with\nbin/tycho in the examples below.
Open the TUI:
\ntycho\nStart the Remote Sessions server:
\ntycho serve\nBind explicitly to localhost:
\ntycho serve --host 127.0.0.1 --port 7373\nManage projects without opening the TUI:
\n# Quick creation uses the current directory and derives the display name.\ntycho project my-workspace\n\n# The explicit form accepts the same options.\ntycho project create my-workspace \\\n --path ~/Code/my-workspace \\\n --name \"My Workspace\" \\\n --group Personal \\\n --harness codex \\\n --model gpt-5.5 \\\n --reasoning-effort medium\n\ntycho project show my-workspace\ntycho project update my-workspace --group Work --model=\"\"\ntycho project archive my-workspace\nCreate and update also accept --response-style, --pr-url, and a\n--hidden=true|false|inherit visibility override. Add --json to any project\ncommand for script-friendly output. Archive rejects projects with running\nagents; otherwise it moves the project configuration and logs to their normal\narchive locations and archives all managed agents owned by the project.
Connect one Remote UI to multiple Tycho servers by adding remote_servers to\n~/.tycho/config/hq.yml:
remote_servers:\n - key: vps\n name: VPS\n url: http://vps-cd946cb7.tail952bf7.ts.net:7373\n token_env: TYCHO_VPS_REMOTE_TOKEN\nThe Remote UI always includes the local server and can switch to configured\npeers from Settings or the top-right menu. Agents, projects, schedules, drafts,\nattachments, and mutations are scoped to the active server; Tycho does not\nmerge state across servers.
\nRun scheduled agents:
\ntycho schedule list\ntycho schedule daemon --once --dry-run\ntycho schedule daemon\nManage schedules without opening the TUI:
\ntycho schedule validate\ntycho schedule list\ntycho schedule run <schedule-key>\ntycho schedule pause <schedule-key>\ntycho schedule resume <schedule-key>\ntycho schedule reload\nRun a non-interactive runtime smoke check:
\ntycho doctor\nThe TUI includes a Schedules screen, and the Remote UI Now view shows\nscheduler daemon freshness plus schedule state. Schedule rows use three\noperator-facing statuses: scheduled, paused, and stopped. Last outcome\ndetails such as failed runs or interactive protection are shown separately. The\nRemote UI keeps the schedule card compact by showing daemon status, PID, and\nlast tick in the header, while schedule rows show project, next run, and a\nhumanized cron cadence such as every 15 minutes.
Schedules create fresh managed agents for projects. They do not run shell\ncommands directly and do not resume old agent sessions. This keeps recurring\nwork reviewable and prevents stale context from accumulating across runs.
\nDefinitions live in ~/.tycho/config/schedules.yml, which is local and gitignored.\nLong prompts should live as Markdown files under ~/.tycho/schedules/.
schedules:\n - key: pull-request-review\n name: Pull request review\n enabled: true\n cron: \"*/15 * * * *\"\n timezone: local\n target:\n type: agent\n project_key: my-workspace\n name: Pull request review\n message_source: file\n message_file: schedules/pull-request-review.md\n policy:\n overlap: skip\n missed: run_once_on_start\n archive_previous_agent: true\nUse message_source: inline with message: \"...\" for short prompts. Use\nmessage_source: file and message_file: schedules/<name>.md for anything\nlonger than a few lines. Tycho validates cron syntax, project references, and\nthat prompt files stay inside ~/.tycho/schedules/.
Each run automatically appends Tycho's final-output attachment checklist to the\nscheduled prompt, so reports, Markdown files, PRs, reviews, images, and other\ndurable artifacts should be returned in attachments.\nno_action_needed is reserved for successful observational checks where no new\ncondition required action; completed changes, answers, commits, reviews, and\ndeliverables use success even when no next step remains.
Prompt tips for reliable schedules:
\ntmp/<schedule-key>/..., and say\nwhether files should be overwritten, updated in place, or left untouched.No new PRs to review.attachments.If you converse with a scheduled agent, Tycho protects that session. A later\ndue run stops the schedule with reason interactive instead of archiving the\nuser-touched agent. Resume the stopped schedule when you want Tycho to archive\nthe active scheduled session and wait for the next scheduled run with fresh\ncontext.
tycho schedule daemon writes daemon heartbeat state to ~/.tycho/logs/scheduler_daemon.json.\nThe schedule UI treats a missing or stale heartbeat as daemon attention. If it\nfinds a running scheduler process without heartbeat state, it reports the\ndaemon as untracked; restart tycho schedule daemon to restore tick freshness.
Start the TUI:
\ntycho\n2 to switch to the Projects screen.N to open the New Project form.codex, claude, or a configured custom harness).Create Project and press Enter.Tycho writes the project entry to ~/.tycho/config/hq.yml, selects the new\nproject, and starts a metadata refresh.
j/k.n to open the Create Agent form for that project.ctrl+j\nfor new lines inside the prompt.Create Agent to save the agent without starting it, or choose\nCreate and Run Agent to start it immediately.After creation, Tycho switches to the Agents screen, selects the new agent, and\nopens its chat panel.
\n1 to switch to the Agents screen.j/k.c or Enter to open the agent chat panel.ctrl+j for multiline\nprompts.ctrl+s to send. If the agent is idle, Tycho starts it\nautomatically.While in chat, use Tab/Shift+Tab to move between the prompt, conversation, and\nsummary areas. L opens the selected agent's raw log from the Agents screen,\nand ctrl+t opens an interactive terminal session for the selected agent.
tycho serve is local-first. If TYCHO_REMOTE_TOKEN is unset, API requests are\naccepted without authentication. This is intended only for localhost.
Set a token before binding to Tailscale or any non-loopback address:
\nTYCHO_REMOTE_TOKEN=\"$(ruby -rsecurerandom -e 'puts SecureRandom.hex(24)')\" tycho serve\nWhen Tailscale HTTPS Serve is available, Tycho can print an HTTPS MagicDNS\nRemote UI URL and QR code. Public screenshots should redact MagicDNS URLs,\nTailscale IPs, and QR codes.
\nFor multiple Remote servers, the browser still talks only to the Tycho server\nthat served the UI. That local server brokers requests to the selected peer,\nusing token_env values from local config or per-browser peer tokens entered\nin Settings. Browser-entered peer tokens are not written to hq.yml.
Tycho has built-in codex, claude, and opencode harnesses. To run Claude\nthrough a wrapper, define a custom harness in ~/.tycho/config/hq.yml and use\nits key as a project or template agent:
custom_harnesses:\n - key: claude-wrapper\n adapter: claude\n execution_command: /Users/you/bin/claude-wrapper\n\nprojects:\n - key: my-workspace\n name: My Workspace\n group: Personal\n path: /Users/you/Code/my-workspace\n agent: claude-wrapper\nexecution_command may be a shell string or an argv list. The command must be\nClaude-compatible because Tycho appends Claude CLI flags for stream-json output,\nstructured result schemas, and native session resume.
Use TYCHO_CODEX_BIN, TYCHO_CLAUDE_BIN, or another documented environment\noverride when a harness executable is not on PATH.
Run the main test suite:
\nbin/test\nRun an individual test:
\nbundle exec ruby test/rendering_test.rb\nTYCHO_REMOTE_TOKEN before binding to a\nnon-loopback interface.docs/PROJECT_STATUS.md: current roadmap and architectural decisions.docs/SYSTEM_PROMPT_AUDIT.md: when Tycho introduces system and automatic prompt context, the exact content, purpose, and known contract gaps.docs/REMOTE_SERVER.md: Remote Sessions API and Remote UI behavior.docs/SCHEDULED_RUNS.md: scheduled-agent design, policies, and runtime state.docs/GOTCHAS.md: operational pitfalls.docs/OPEN_SOURCE_PLAN.md: open-source readiness plan and launch criteria.See CONTRIBUTING.md and CODE_OF_CONDUCT.md.
See SECURITY.md.
Tycho is released under the MIT License. See LICENSE.
\n\nThis project was built almost entirely by AI (Claude, GPT, and other LLMs).\nThe code, architecture, and documentation were largely generated, reviewed, and iterated by AI agents with human oversight.\nUse at your own risk โ thorough review before production use is strongly recommended.
\n
![\"Artidor\"](\"https://raw.githubusercontent.com/Aofsnorth/Artidor/main/apps/web/public/logos/artidor/logo-particles.png\"){kind=logo}
\nLocal-first ยท MIT-licensed ยท No uploads ยท No paywalls ยท AI-native
\nWebsite ยท Quick start ยท Features ยท AI Co-Pilot ยท Issues ยท Discord
\n![\"MIT](\"https://img.shields.io/badge/license-MIT-white?style=flat-square\"){kind=icon}
![\"Built](\"https://img.shields.io/badge/fork-OpenCut-2a2a2a?style=flat-square\"){kind=icon}
![\"Bun\"](\"https://img.shields.io/badge/bun-1.2.18-f9f1e1?style=flat-square&logo=bun&logoColor=black\"){kind=icon}
![\"Next.js](\"https://img.shields.io/badge/next.js-16-black?style=flat-square&logo=next.js\"){kind=icon}
![\"React](\"https://img.shields.io/badge/react-19-149eca?style=flat-square&logo=react\"){kind=icon}
![\"Rust](\"https://img.shields.io/badge/rust-2024-f74c00?style=flat-square&logo=rust\"){kind=icon}
![\"wgpu\"](\"https://img.shields.io/badge/gpu-wgpu-005a9c?style=flat-square\"){kind=icon}
\n ![\"Artidor](\"https://raw.githubusercontent.com/Aofsnorth/Artidor/main/docs/progress/image-1.png\")
\n
\n ![\"Artidor](\"https://raw.githubusercontent.com/Aofsnorth/Artidor/main/docs/progress/image-2.png\")
\n ![\"Artidor](\"https://raw.githubusercontent.com/Aofsnorth/Artidor/main/docs/progress/image-3.png\")
\n
\n ![\"Artidor](\"https://raw.githubusercontent.com/Aofsnorth/Artidor/main/docs/progress/image-4.png\")
\n ![\"Artidor](\"https://raw.githubusercontent.com/Aofsnorth/Artidor/main/docs/progress/image-5.png\")
\n
Most \"free\" video editors are paywalled. The rest upload your footage to a server you don't control. The ones that don't are unusable.
\nArtidor does the obvious things:
\nNo manifesto. No \"rethinking the creative process.\" Just a tool that works.
\nTICKS_PER_SECOND = 120_000 for lossless time math at 24/25/29.97/30/50/59.94/60/120 fpsMediaRecorder chunks, no full downloadArtidor ships with an AI panel in the left bar (under Assets). The Co-Pilot speaks every command the editor speaks โ split, trim, retime, keyframe, transition, color-grade, import, export โ and dispatches them as tool calls against the live editor.
Three things set it apart from \"AI edits your video\" toys:
\nThe Co-Pilot doesn't transcribe your prompt and run a script. It has 40+ typed tools โ set_project_fps, insert_text_element, upsert_keyframe, apply_preset, export_project โ each one wraps a real EditorCore method. The LLM can't hallucinate outside the editor's surface.
Every command you fire (via mouse, keyboard, or the AI) is logged to a 500-event telemetry store. The Co-Pilot's system prompt includes your last 20 edits โ cut pattern, easing, pacing โ so its suggestions match your style instead of generic.
\nDrop a finished video into the AI panel. The style extractor runs entirely client-side:
\nStyleProfile injected into the promptThe Co-Pilot then imitates that pacing on your timeline.
\n# .env.local โ pick ONE\nOPENAI_API_KEY=sk-...\nANTHROPIC_API_KEY=sk-ant-...\nOLLAMA_BASE_URL=http://localhost:11434 # local\nIf no key is set, the panel still opens โ it just tells you on the first send.
\nPrerequisites: Bun โฅ 1.2.18. Docker is optional (for cloud features like collab).
\ngit clone https://github.com/Aofsnorth/Artidor.git\ncd Artidor\nbun install\nbun dev:web\nOpen http://localhost:3000. Projects live in IndexedDB; nothing leaves your machine.
\ngit clone https://github.com/Aofsnorth/Artidor.git\ncd Artidor\ndocker compose up -d db redis serverless-redis-http\ncp apps/web/.env.example apps/web/.env.local\nbun install\nbun dev:web\nThe default .env.example works out of the box โ Postgres + Redis are auto-created with dev credentials. The offline editor works without any of this.
# Build the WASM module once\nbun run build:wasm\ncd rust/wasm/pkg && bun link\ncd ../../apps/web && bun link artidor-wasm\n\n# Or: rebuild on every change\nbun dev:wasm # in a second terminal\nbun dev:web # in the first\napps/desktop uses GPUI. See apps/desktop/README.md for the Rust toolchain.
Artidor/\nโโ apps/\nโ โโ web/ Next.js 16 + React 19 frontend\nโ โ โโ src/\nโ โ โ โโ app/ Routes, layouts, server components\nโ โ โ โ โโ api/ API routes (ai, auth, drive, github, โฆ)\nโ โ โ โ โโ editor/ /editor/[project_id] โ the workspace\nโ โ โ โ โโ projects/ /projects โ the dashboard\nโ โ โ โโ components/ UI shell โ no domain logic\nโ โ โ โ โโ editor/panels/ Asset / properties / timeline\nโ โ โ โโ core/ EditorCore facade + 14 managers\nโ โ โ โโ hooks/ React bindings\nโ โ โ โโ lib/\nโ โ โ โ โโ ai/ AI Co-Pilot (provider, tools, telemetry, style)\nโ โ โ โ โโ timeline/ Timeline types\nโ โ โ โ โโ export/ MediaRecorder pipelines\nโ โ โ โโ stores/ Zustand stores\nโ โ โโ public/ Static assets (logos, fonts, screenshots)\nโ โโ desktop/ GPUI shell โ same Rust core\nโ\nโโ rust/\nโ โโ wasm/ Compiles to artidor-wasm npm package\nโ โโ crates/ Workspace crates\nโ โโ bridge/ #[export] proc-macro โ wasm_bindgen\nโ โโ time/ MediaTime, FrameRate, Easing, keyframes\nโ โโ gpu/ wgpu device + pipeline cache\nโ โโ compositor/ Scene graph + draw ordering\nโ โโ effects/ Effect definitions + parameter trees\nโ โโ masks/ Mask shapes + compositing\nโ\nโโ docs/ Architecture notes\nโโ .github/ CI, issue templates, contributing\nRule of thumb: if it's not a UI concern, it goes in rust/. Every line of business logic in apps/web/src/core/ is a migration in progress.
The app works fully offline with no environment variables. The defaults in apps/web/.env.example cover local dev. Cloud / AI features need these:
| Variable | \nRequired for | \nDefault | \n
|---|---|---|
OPENAI_API_KEY | \nAI Co-Pilot (GPT) | \nโ | \n
ANTHROPIC_API_KEY | \nAI Co-Pilot (Claude) | \nโ | \n
OLLAMA_BASE_URL | \nAI Co-Pilot (local) | \nhttp://localhost:11434 | \n
GITHUB_TOKEN | \nHigher GitHub API rate (5k/hr) | \nโ | \n
DATABASE_URL | \nPostgres (cloud features) | \npostgresql://artidor:artidor@localhost:5432/artidor | \n
BETTER_AUTH_SECRET | \nAuth | \ndev-only fallback | \n
UPSTASH_REDIS_REST_URL | \nRedis | \nhttp://localhost:8079 | \n
UPSTASH_REDIS_REST_TOKEN | \nRedis | \ndev-only fallback | \n
FREESOUND_CLIENT_ID | \nSound search | \nโ | \n
FREESOUND_KEY | \nSound search | \nโ | \n
rust/crates/time exposes MediaTime(i64) wrapping ticks at TICKS_PER_SECOND = 120_000. That divides cleanly into every supported framerate (24/25/29.97/30/50/59.94/60/120), so frame boundaries are never fractional.rust/. Frontends are replaceable; the core is not. #[export] on a Rust function turns into #[wasm_bindgen(js_name = \"camelCase\")] automatically.apps/web/src/core/ facade is a thin wrapper over the Rust core. Components own rendering, never domain rules.localStorage. The LLM sees your last 20 edits and matches them.motion/react only for orchestration; everything else is transition + CSS variables.| Script | \nWhat it does | \n
|---|---|
bun dev:web | \nNext.js dev server on :3000 | \n
bun dev:wasm | \ncargo watch rebuilds the Rust โ WASM package on every change | \n
bun run build:web | \nProduction build of the web app | \n
bun run build:wasm | \nOne-shot WASM build | \n
bun run lint:web | \nBiome lint | \n
bun run lint:web:fix | \nBiome lint with --write --unsafe | \n
bun run format:web | \nBiome format (renderer dir) | \n
bun run test | \nBun test runner | \n
bun run preview:web | \nNext.js production preview | \n
bun run publish:wasm | \nBuild + publish artidor-wasm to npm | \n
bun run generate:fonts | \nRegenerate the font sprite chunks in public/ | \n
Two rules:
\naria-* beats div. CSS transition beats an animation lib. Postgres constraints beat app code. A Rust iterator beats a JS one.rust/, UI goes in apps/. If you find yourself putting a domain rule in a React component, move it.Before opening a PR:
\nbun run lint:web (and lint:web:fix for what biome can repair)bun run test (Bun test runner)cd apps/web && bunx tsc --noEmit (no type errors)For larger changes, open an issue first so we can agree on direction. See .github/CONTRIBUTING.md for the rest.
MIT. Use it, fork it, ship a competitor, whatever.
\nBuilt on the foundation of OpenCut โ same MIT license, same DNA. All Rust core is original Artidor work.
\nBuilt in public ยท The repo is the brand
\n\n\n\n
Display unique avatar for any user based on their (initials) name.
\n
This package originally built for Laravel, but can also be used in any PHP project.
\nRead more about integration with PHP project here.
\ncomposer require laravolt/avatar\ncomposer require laravolt/avatar ~0.3\nNote: only for Laravel 5.4 and below, because since Laravel 5.5 we use package auto-discovery.
\nLaravolt\\Avatar\\ServiceProvider::class,\n\n...\n\n'Avatar' => Laravolt\\Avatar\\Facade::class,\nphp artisan vendor:publish --provider=\"Laravolt\\Avatar\\ServiceProvider\"\nThis will create config file located in config/laravolt/avatar.php.
$app->register(Laravolt\\Avatar\\LumenServiceProvider);\n//this will output data-uri (base64 image data)\n//something like data:image/png;base64,iVBORw0KGg....\nAvatar::create('Joko Widodo')->toBase64();\n\n//use in view\n//this will display initials JW as an image\n<img src=\"https://raw.githubusercontent.com/laravolt/avatar/master/%7B%7B%20Avatar::create('Joko%20Widodo')->toBase64()%20%7D%7D\" />\nAvatar::create('Susilo Bambang Yudhoyono')->save('sample.png');\nAvatar::create('Susilo Bambang Yudhoyono')->save('sample.jpg', 100); // quality = 100\nAvatar::create('uyab@example.net')->toGravatar();\n// Output: http://gravatar.com/avatar/0c5cbf5a8762d91d930795a6107b2ce5814a6ab26e60c7ec6b75bc81c7dfe3ee\n\nAvatar::create('uyab@example.net')->toGravatar(['d' => 'identicon', 'r' => 'pg', 's' => 100]);\n// Output: http://gravatar.com/avatar/0c5cbf5a8762d91d930795a6107b2ce5814a6ab26e60c7ec6b75bc81c7dfe3ee?d=identicon&r=pg&s=100\nGravatar parameter reference: https://docs.gravatar.com/api/avatars/images/
\nAvatar::create('Susilo Bambang Yudhoyono')->toSvg();\nYou may specify custom font-family for your SVG text.
\n<head>\n <!--Prepare custom font family, using Google Fonts-->\n <link href=\"https://fonts.googleapis.com/css?family=Laravolt\" rel=\"stylesheet\">\n\n <!--OR-->\n\n <!--Setup your own style-->\n <style>\n @font-face {\n font-family: Laravolt;\n src: url({{ asset('fonts/laravolt.woff')) }});\n }\n </style>\n</head>\nAvatar::create('Susilo Bambang Yudhoyono')->setFontFamily('Laravolt')->toSvg();\nYou may make the SVG responsive. This excludes the height and width attributes.
\nAvatar::create('Susilo Bambang Yudhoyono')->setResponsive()->toSvg();\nAvatar::create('Abdul Somad')->getImageObject();\nThe method will return an instance of Intervention image object, so you can use it for further purposes.
\nBy default, this package will try to output any initials letter as it is. If the name supplied contains any non-ASCII character (e.g. ฤ, ฤ, วฝ) then the result will depend on which font used (see config). It the font supports characters supplied, it will successfully displayed, otherwise it will not.
\nAlternatively, we can convert all non-ascii to their closest ASCII counterparts. If no closest coutnerparts found, those characters are removed. Thanks to Stringy for providing such useful functions. What we need is just change one line in config/avatar.php:
'ascii' => true,\n<?php\n/*\n * Set specific configuration variables here\n */\nreturn [\n\n /*\n |--------------------------------------------------------------------------\n | Image Driver\n |--------------------------------------------------------------------------\n | Avatar use Intervention Image library to process image.\n | Meanwhile, Intervention Image supports \"GD Library\" and \"Imagick\" to process images\n | internally. You may choose one of them according to your PHP\n | configuration. By default PHP's \"Imagick\" implementation is used.\n |\n | Supported: \"gd\", \"imagick\"\n |\n */\n 'driver' => 'gd',\n\n // Initial generator class\n 'generator' => \\Laravolt\\Avatar\\Generator\\DefaultGenerator::class,\n\n // Whether all characters supplied must be replaced with their closest ASCII counterparts\n 'ascii' => false,\n\n // Image shape: circle or square\n 'shape' => 'circle',\n\n // Image width, in pixel\n 'width' => 100,\n\n // Image height, in pixel\n 'height' => 100,\n\n // Number of characters used as initials. If name consists of single word, the first N character will be used\n 'chars' => 2,\n\n // font size\n 'fontSize' => 48,\n\n // convert initial letter in uppercase\n 'uppercase' => false,\n\n // Right to Left (RTL)\n 'rtl' => false,\n\n // Fonts used to render text.\n // If contains more than one fonts, randomly selected based on name supplied\n 'fonts' => [__DIR__.'/../fonts/OpenSans-Bold.ttf', __DIR__.'/../fonts/rockwell.ttf'],\n\n // List of foreground colors to be used, randomly selected based on name supplied\n 'foregrounds' => [\n '#FFFFFF',\n ],\n\n // List of background colors to be used, randomly selected based on name supplied\n 'backgrounds' => [\n '#f44336',\n '#E91E63',\n '#9C27B0',\n '#673AB7',\n '#3F51B5',\n '#2196F3',\n '#03A9F4',\n '#00BCD4',\n '#009688',\n '#4CAF50',\n '#8BC34A',\n '#CDDC39',\n '#FFC107',\n '#FF9800',\n '#FF5722',\n ],\n\n 'border' => [\n 'size' => 1,\n\n // border color, available value are:\n // 'foreground' (same as foreground color)\n // 'background' (same as background color)\n // or any valid hex ('#aabbcc')\n 'color' => 'background',\n\n // border radius, only works for SVG\n 'radius' => 0,\n ],\n\n // List of theme name to be used when rendering avatar\n // Possible values are:\n // 1. Theme name as string: 'colorful'\n // 2. Or array of string name: ['grayscale-light', 'grayscale-dark']\n // 3. Or wildcard \"*\" to use all defined themes\n 'theme' => ['*'],\n\n // Predefined themes\n // Available theme attributes are:\n // shape, chars, backgrounds, foregrounds, fonts, fontSize, width, height, ascii, uppercase, and border.\n 'themes' => [\n 'grayscale-light' => [\n 'backgrounds' => ['#edf2f7', '#e2e8f0', '#cbd5e0'],\n 'foregrounds' => ['#a0aec0'],\n ],\n 'grayscale-dark' => [\n 'backgrounds' => ['#2d3748', '#4a5568', '#718096'],\n 'foregrounds' => ['#e2e8f0'],\n ],\n 'colorful' => [\n 'backgrounds' => [\n '#f44336',\n '#E91E63',\n '#9C27B0',\n '#673AB7',\n '#3F51B5',\n '#2196F3',\n '#03A9F4',\n '#00BCD4',\n '#009688',\n '#4CAF50',\n '#8BC34A',\n '#CDDC39',\n '#FFC107',\n '#FF9800',\n '#FF5722',\n ],\n 'foregrounds' => ['#FFFFFF'],\n ],\n ]\n];\nWe can overriding configuration at runtime by using following functions:
\nAvatar::create('Soekarno')->setDimension(100);//width = height = 100 pixel\nAvatar::create('Soekarno')->setDimension(100, 200); // width = 100, height = 200\nAvatar::create('Soekarno')->setBackground('#001122');\nAvatar::create('Soekarno')->setForeground('#999999');\nAvatar::create('Soekarno')->setFontSize(72);\nAvatar::create('Soekarno')->setFont('/path/to/font.ttf');\nAvatar::create('Soekarno')->setBorder(1, '#aabbcc'); // size = 1, color = #aabbcc\nAvatar::create('Soekarno')->setBorder(1, '#aabbcc', 10); // size = 1, color = #aabbcc, border radius = 10 (only for SVG)\nAvatar::create('Soekarno')->setShape('square');\n\n// Available since 3.0.0\nAvatar::create('Soekarno')->setTheme('colorful'); // set exact theme\nAvatar::create('Soekarno')->setTheme(['grayscale-light', 'grayscale-dark']); // theme will be randomized from these two options\n\n// chaining\nAvatar::create('Habibie')->setDimension(50)->setFontSize(18)->toBase64();\n// include composer autoload\nrequire 'vendor/autoload.php';\n\n// import the Avatar class\nuse Laravolt\\Avatar\\Avatar;\n\n// create your first avatar\n$avatar = new Avatar($config);\n$avatar->create('John Doe')->toBase64();\n$avatar->create('John Doe')->save('path/to/file.png', $quality = 90);\n$config is just an ordinary array with same format as explained above (See Configuration).
\n\n\n
Display unique avatar for any user based on their (initials) name.
\n
| Version | \nPHP | \nLaravel | \nIntervention Image | \n
|---|---|---|---|
| 7.x | \n>= 8.3 | \n10.x โ 13.x | \n^4.0 | \n
| 6.x | \n>= 8.1 | \n10.x โ 12.x | \n^3.0 | \n
| 5.x | \n>= 8.0 | \n9.x โ 11.x | \n^2.0 | \n
This package originally built for Laravel, but can also be used in any PHP project.
\nRead more about integration with PHP project here.
\ncomposer require laravolt/avatar\nThe service provider and Avatar facade are registered automatically via package auto-discovery.
php artisan vendor:publish --provider=\"Laravolt\\Avatar\\ServiceProvider\"\nThis will create config file located in config/laravolt/avatar.php.
$app->register(Laravolt\\Avatar\\LumenServiceProvider);\n//this will output data-uri (base64 image data)\n//something like data:image/png;base64,iVBORw0KGg....\nAvatar::create('Joko Widodo')->toBase64();\n\n//use in view\n//this will display initials JW as an image\n<img src=\"https://raw.githubusercontent.com/laravolt/avatar/master/%7B%7B%20Avatar::create('Joko%20Widodo')->toBase64()%20%7D%7D\" />\nAvatar::create('Susilo Bambang Yudhoyono')->save('sample.png');\nAvatar::create('Susilo Bambang Yudhoyono')->save('sample.jpg', 100); // quality = 100\nAvatar::create('uyab@example.net')->toGravatar();\n// Output: http://gravatar.com/avatar/0c5cbf5a8762d91d930795a6107b2ce5814a6ab26e60c7ec6b75bc81c7dfe3ee\n\nAvatar::create('uyab@example.net')->toGravatar(['d' => 'identicon', 'r' => 'pg', 's' => 100]);\n// Output: http://gravatar.com/avatar/0c5cbf5a8762d91d930795a6107b2ce5814a6ab26e60c7ec6b75bc81c7dfe3ee?d=identicon&r=pg&s=100\nGravatar parameter reference: https://docs.gravatar.com/api/avatars/images/
\nAvatar::create('Susilo Bambang Yudhoyono')->toSvg();\nYou may specify custom font-family for your SVG text.
\n<head>\n <!--Prepare custom font family, using Google Fonts-->\n <link href=\"https://fonts.googleapis.com/css?family=Laravolt\" rel=\"stylesheet\">\n\n <!--OR-->\n\n <!--Setup your own style-->\n <style>\n @font-face {\n font-family: Laravolt;\n src: url({{ asset('fonts/laravolt.woff')) }});\n }\n </style>\n</head>\nAvatar::create('Susilo Bambang Yudhoyono')->setFontFamily('Laravolt')->toSvg();\nYou may make the SVG responsive. This excludes the height and width attributes.
\nAvatar::create('Susilo Bambang Yudhoyono')->setResponsive()->toSvg();\nAvatar::create('Abdul Somad')->getImageObject();\nThe method will return an instance of Intervention image object, so you can use it for further purposes.
\nBy default, this package will try to output any initials letter as it is. If the name supplied contains any non-ASCII character (e.g. ฤ, ฤ, วฝ) then the result will depend on which font used (see config). It the font supports characters supplied, it will successfully displayed, otherwise it will not.
\nAlternatively, we can convert all non-ascii to their closest ASCII counterparts. If no closest coutnerparts found, those characters are removed. Thanks to Stringy for providing such useful functions. What we need is just change one line in config/avatar.php:
'ascii' => true,\n<?php\n/*\n * Set specific configuration variables here\n */\nreturn [\n\n /*\n |--------------------------------------------------------------------------\n | Image Driver\n |--------------------------------------------------------------------------\n | Avatar use Intervention Image library to process image.\n | Meanwhile, Intervention Image supports \"GD Library\" and \"Imagick\" to process images\n | internally. You may choose one of them according to your PHP\n | configuration. By default PHP's \"Imagick\" implementation is used.\n |\n | Supported: \"gd\", \"imagick\"\n |\n */\n 'driver' => 'gd',\n\n // Initial generator class\n 'generator' => \\Laravolt\\Avatar\\Generator\\DefaultGenerator::class,\n\n // Whether all characters supplied must be replaced with their closest ASCII counterparts\n 'ascii' => false,\n\n // Image shape: circle or square\n 'shape' => 'circle',\n\n // Image width, in pixel\n 'width' => 100,\n\n // Image height, in pixel\n 'height' => 100,\n\n // Number of characters used as initials. If name consists of single word, the first N character will be used\n 'chars' => 2,\n\n // font size\n 'fontSize' => 48,\n\n // convert initial letter in uppercase\n 'uppercase' => false,\n\n // Right to Left (RTL)\n 'rtl' => false,\n\n // Fonts used to render text.\n // If contains more than one fonts, randomly selected based on name supplied\n 'fonts' => [__DIR__.'/../fonts/OpenSans-Bold.ttf', __DIR__.'/../fonts/rockwell.ttf'],\n\n // List of foreground colors to be used, randomly selected based on name supplied\n 'foregrounds' => [\n '#FFFFFF',\n ],\n\n // List of background colors to be used, randomly selected based on name supplied\n 'backgrounds' => [\n '#f44336',\n '#E91E63',\n '#9C27B0',\n '#673AB7',\n '#3F51B5',\n '#2196F3',\n '#03A9F4',\n '#00BCD4',\n '#009688',\n '#4CAF50',\n '#8BC34A',\n '#CDDC39',\n '#FFC107',\n '#FF9800',\n '#FF5722',\n ],\n\n 'border' => [\n 'size' => 1,\n\n // border color, available value are:\n // 'foreground' (same as foreground color)\n // 'background' (same as background color)\n // or any valid hex ('#aabbcc')\n 'color' => 'background',\n\n // border radius, only works for SVG\n 'radius' => 0,\n ],\n\n // List of theme name to be used when rendering avatar\n // Possible values are:\n // 1. Theme name as string: 'colorful'\n // 2. Or array of string name: ['grayscale-light', 'grayscale-dark']\n // 3. Or wildcard \"*\" to use all defined themes\n 'theme' => ['*'],\n\n // Predefined themes\n // Available theme attributes are:\n // shape, chars, backgrounds, foregrounds, fonts, fontSize, width, height, ascii, uppercase, and border.\n 'themes' => [\n 'grayscale-light' => [\n 'backgrounds' => ['#edf2f7', '#e2e8f0', '#cbd5e0'],\n 'foregrounds' => ['#a0aec0'],\n ],\n 'grayscale-dark' => [\n 'backgrounds' => ['#2d3748', '#4a5568', '#718096'],\n 'foregrounds' => ['#e2e8f0'],\n ],\n 'colorful' => [\n 'backgrounds' => [\n '#f44336',\n '#E91E63',\n '#9C27B0',\n '#673AB7',\n '#3F51B5',\n '#2196F3',\n '#03A9F4',\n '#00BCD4',\n '#009688',\n '#4CAF50',\n '#8BC34A',\n '#CDDC39',\n '#FFC107',\n '#FF9800',\n '#FF5722',\n ],\n 'foregrounds' => ['#FFFFFF'],\n ],\n ]\n];\nWe can overriding configuration at runtime by using following functions:
\nAvatar::create('Soekarno')->setDimension(100);//width = height = 100 pixel\nAvatar::create('Soekarno')->setDimension(100, 200); // width = 100, height = 200\nAvatar::create('Soekarno')->setBackground('#001122');\nAvatar::create('Soekarno')->setForeground('#999999');\nAvatar::create('Soekarno')->setFontSize(72);\nAvatar::create('Soekarno')->setFont('/path/to/font.ttf');\nAvatar::create('Soekarno')->setBorder(1, '#aabbcc'); // size = 1, color = #aabbcc\nAvatar::create('Soekarno')->setBorder(1, '#aabbcc', 10); // size = 1, color = #aabbcc, border radius = 10 (only for SVG)\nAvatar::create('Soekarno')->setShape('square');\n\n// Available since 3.0.0\nAvatar::create('Soekarno')->setTheme('colorful'); // set exact theme\nAvatar::create('Soekarno')->setTheme(['grayscale-light', 'grayscale-dark']); // theme will be randomized from these two options\n\n// chaining\nAvatar::create('Habibie')->setDimension(50)->setFontSize(18)->toBase64();\n// include composer autoload\nrequire 'vendor/autoload.php';\n\n// import the Avatar class\nuse Laravolt\\Avatar\\Avatar;\n\n// create your first avatar\n$avatar = new Avatar($config);\n$avatar->create('John Doe')->toBase64();\n$avatar->create('John Doe')->save('path/to/file.png', $quality = 90);\n$config is just an ordinary array with same format as explained above (See Configuration).
\n Noise-canceling context and long-term memory for your AI agent. Stop paying Claude to read 10,000 lines of terminal noise like a headphone for AI agent\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Problem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
NATS is a Next.js-based ERP system designed to handle various business functions ranging from accounting, inventory, sales, purchasing, POS, to payroll.
\n\n![\"1776302209274\"](\"https://raw.githubusercontent.com/maziyank/nats/app-main/image/README/1776302209274.png\")
\nMain Dashboard View
![\"1776302243874\"](\"https://raw.githubusercontent.com/maziyank/nats/app-main/image/README/1776302243874.png\")
\nAccounting Module
![\"1776302322989\"](\"https://raw.githubusercontent.com/maziyank/nats/app-main/image/README/1776302322989.png\")
\nFinancial Report
![\"1776302374473\"](\"https://raw.githubusercontent.com/maziyank/nats/app-main/image/README/1776302374473.png\")
\nPoint of Sale (POS)
Follow the steps below to run NATS in your local environment.
\nBefore starting, ensure your system has the following components:
\ngit clone <repository-url>\ncd nats\nnpm install\nCopy the .env.example file to .env and adjust its values:
cp .env.example .env\nEnsure the DATABASE_URL variable correctly points to your PostgreSQL instance:\nDATABASE_URL=\"postgresql://user:password@localhost:5432/nats\"
Create a database in PostgreSQL:
\npsql -U postgres -c \"CREATE DATABASE nats;\"\nPerform database migration and schema creation:
\nnpx prisma generate\nnpx prisma migrate dev --name init\nPopulate the database with initial data (roles, default users, etc.). Choose one of the following options:
\nOption A: Complete Seeding (Recommended for Testing)\nIncludes sample products, transactions, and bulk data:
\nnpm run prisma db seed\nOption B: Minimal Seeding (Clean Start)\nIncludes only essential data: Company Profile, Chart of Accounts, and Default Roles/Users:
\nnpm run prisma:seed:minimal\nDefault Credentials:
\npassword123 (for all default users)| Role | \nName | \n|
|---|---|---|
| Super Admin | \nadmin@example.com | \nAdmin User | \n
| Accountant | \naccountant@example.com | \nJohn Accountant | \n
| Cashier | \ncashier@example.com | \nJane Cashier | \n
| Manager | \nmanager@example.com | \nMike Manager | \n
| Merchant | \nmerchant@example.com | \nSample Merchant | \n
| Customer | \ncustomer@example.com | \nSample Customer | \n
Run the development server:
\nnpm run dev\nThe application can be accessed at http://localhost:3000.
\nIf you want to run the application using Docker Compose:
\ndocker-compose up -d\nAfter the containers are running, initialize the database:
\ndocker-compose exec app npx prisma migrate deploy\n# Run complete seed\ndocker-compose exec app npm run prisma db seed\n\n# OR run minimal seed\ndocker-compose exec app npm run prisma:seed:minimal\nThis project is licensed under LICENSE.
\n" }, - { - "fullName": "rayasabari/yntk-ts", - "name": "yntk-ts", - "owner": "rayasabari", - "ownerAvatarUrl": "https://avatars.githubusercontent.com/u/57746279?v=4", - "description": "You Need This Kit - Type-safe Starter: REST API boilerplate that ships with User Management, Role Management, and Authentication with JWTs, Prisma ORM, and PostgreSQL", - "metaDescription": "You Need This Kit - Type-safe Starter: REST API boilerplate that ships with User Management, Role Management, and Authentication with JWTs, Prisma ORMโฆ", - "url": "https://github.com/rayasabari/yntk-ts", - "homepage": "", - "language": "TypeScript", - "stars": 20, - "forks": 0, - "topics": [], - "updatedAt": "2026-06-25T07:10:46Z", - "pushedAt": "2026-04-08T09:29:42Z", - "latestRelease": null, - "archived": false, - "licenseSpdx": "", - "createdAt": "2026-04-08T09:27:49Z", - "openIssues": 0, - "openPullRequests": 0, - "subscribers": 0, - "communityHealth": 28, - "readmeHtml": "TypeScript/Express REST API starter kit that ships with User Management, Role Management, and Authentication with JWTs, Prisma ORM, and PostgreSQL. The project is organized by feature with explicit service/repository layers so business logic stays separated from transport concerns.
\nsrc/\nโโโ app.ts # Express app bootstrap\nโโโ server.ts # Starts HTTP server\nโโโ config/ # Environment loader & Prisma client wrapper\nโโโ controllers/ # HTTP handlers grouped by module + shared helpers\nโโโ services/ # Business logic (auth/user) & mappers\nโโโ repositories/ # Prisma data access per module\nโโโ middleware/ # Cross-cutting middleware (auth, validation)\nโโโ routes/ # Express routers mounted under /auth and /users\nโโโ validations/ # Zod schemas for request validation\nโโโ views/ # Email templates\nโโโ errors/ # Custom AppError type & error utilities\nโโโ utils/ # Shared utilities (password, string, Zod helpers)\nโโโ types/ # Shared TS types & Express module augmentation\npnpm install\nCreate .env (never commit it) with the required settings:
# Server Configuration\nNODE_ENV=development\nFRONTEND_URL=http://localhost:8080\nPORT=5050\nLOG_LEVEL=info\n\n# Database\nDATABASE_URL=postgresql://USER:PASSWORD@HOST:PORT/DATABASE\n\n# JWT Configuration\nJWT_SECRET=super-secret\n\n# Bcrypt Configuration\nSALT_ROUNDS=10\n\n# Email Configuration\nEMAIL_HOST=smtp.gmail.com\nEMAIL_PORT=587\nEMAIL_USER=your-email@gmail.com\nEMAIL_PASSWORD=your-app-password\nEMAIL_FROM=noreply@yourapp.com\n\n# Token Configuration\nACCESS_TOKEN_EXPIRY=\"1h\"\nENABLE_REFRESH_TOKEN=true\nREFRESH_TOKEN_IN_JSON=true\nREFRESH_TOKEN_IN_COOKIE=true\nREFRESH_TOKEN_EXPIRY=7 * 24 * 60 * 60 * 1000 # 7 days in miliseconds\nCOOKIE_SAME_SITE=\"strict\" # strict, lax, none\nRESET_PASSWORD_TOKEN_EXPIRY=1 * 60 * 60 * 1000 # 1 hour in miliseconds\nEMAIL_VERIFICATION_TOKEN_EXPIRY=24 * 60 * 60 * 10000 # 24 hours in miliseconds\n\n# CORS Configuration\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080,https://yourdomain.com\nCORS_CREDENTIALS=true\n\n\nNote for Gmail: Use an App Password instead of your regular password. Enable 2FA and generate an App Password in Google Account Settings โ Security โ App passwords.
\n
prisma/schema.prisma.pnpm prisma migrate dev (for local) or pnpm prisma db push for quick sync.pnpm prisma generate. Output lands in src/generated/prisma.pnpm prisma db seedpnpm dev\nRuns tsx in watch mode, recompiling on changes. The API listens on PORT from the env file (defaults to 5050).
| Command | \nPurpose | \n
|---|---|
pnpm dev | \nStart the API in watch mode with tsx | \n
pnpm prisma migrate dev | \nCreate/apply migrations and regenerate Prisma client | \n
pnpm prisma generate | \nRegenerate Prisma client manually | \n
pnpm prisma db push | \nQuick sync schema to database without migrations | \n
pnpm prisma db seed | \nSeed the database | \n
pnpm build | \nBuild the API for production (bundles with tsup) | \n
pnpm start | \nStart the production server from dist/ | \n
pnpm test | \nRun all tests in watch mode | \n
pnpm test:unit | \nRun unit tests only | \n
pnpm test:integration | \nRun integration tests only (sequential) | \n
\n\nโ Production build: The repo currently runs via tsx; add a
\ntscbuild +startscript before deploying to production environments like Vercel/Node runtime functions.
The API includes interactive documentation powered by Swagger UI and OpenAPI 3.0 (swagger-jsdoc and swagger-ui-express).
/api-docs (accessible when the server is running)All endpoints respond with { status, message, data? } JSON payloads.\nFor paginated endpoints (like GET /users and GET /roles), the response also includes meta and links objects containing paging data and HATEOAS navigational URLs. They optionally accept query parameters: ?page=1&limit=10&sortBy=createdAt&sortOrder=asc&search=value.
/auth)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
POST | \n/auth/register | \nRegister a new user and send verification email | \nPublic | \nregisterUserSchema | \n
POST | \n/auth/login | \nVerify credentials and return JWT token | \nPublic | \n- | \n
POST | \n/auth/logout | \nLogout and revoke refresh token | \nRequired | \nlogoutSchema | \n
POST | \n/auth/refresh-token | \nRefresh access token using refresh token | \nPublic | \nrefreshTokenSchema | \n
POST | \n/auth/verify-email | \nVerify email address using token from email | \nPublic | \nverifyEmailSchema | \n
POST | \n/auth/resend-verification | \nResend verification email (rate limited: 3/10min) | \nPublic | \nresendVerificationSchema | \n
POST | \n/auth/forgot-password | \nRequest password reset email (rate limited: 3/15min) | \nPublic | \nforgotPasswordSchema | \n
POST | \n/auth/reset-password | \nReset password using token from email | \nPublic | \nresetPasswordSchema | \n
/users)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
POST | \n/users | \nCreate a user (admin-style) | \nRequired (users:create) | \ncreateUserSchema | \n
GET | \n/users | \nList all users | \nRequired (users:read) | \n- | \n
GET | \n/users/:id | \nFetch a user by ID | \nRequired (users:read) | \n- | \n
PUT | \n/users/:id | \nUpdate user fields (username, email, displayName) | \nRequired (users:update) | \nupdateUserSchema | \n
PUT | \n/users/:id/roles | \nAssign roles to a user | \nRequired (roles:assign) | \nassignRolesSchema | \n
PATCH | \n/users/password | \nUpdate current user's password | \nRequired (users:update) | \nupdatePasswordSchema | \n
DELETE | \n/users/:id | \nRemove a user | \nRequired (users:delete) | \n- | \n
/roles & /permissions)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
GET | \n/roles | \nList all roles | \nRequired (roles:read) | \n- | \n
GET | \n/roles/:id | \nFetch a role by ID | \nRequired (roles:read) | \n- | \n
POST | \n/roles | \nCreate a new role | \nRequired (roles:create) | \ncreateRoleSchema | \n
PUT | \n/roles/:id | \nUpdate an existing role | \nRequired (roles:update) | \nupdateRoleSchema | \n
DELETE | \n/roles/:id | \nRemove a role | \nRequired (roles:delete) | \n- | \n
GET | \n/permissions | \nList all system permissions | \nRequired (roles:read) | \n- | \n
Auth Required: Endpoints require Authorization: Bearer <token> header.
The API uses Zod for request validation with the following schemas:
\nregisterUserSchema: Validates user registration (username, email, displayName, password)createUserSchema: Validates user creation (username, email, displayName)updateUserSchema: Validates user updates (partial fields)assignRolesSchema: Validates assigning role IDs to a userupdatePasswordSchema: Validates password changes (currentPassword, newPassword, confirmPassword)createRoleSchema: Validates new role details and array of permission IDsupdateRoleSchema: Validates partial updates to a roleverifyEmailSchema: Validates email verification tokenresendVerificationSchema: Validates email for resending verificationforgotPasswordSchema: Validates email for password reset requestsresetPasswordSchema: Validates reset token and new password (min 8 chars, uppercase, lowercase, number)refreshTokenSchema: Validates refresh token from body or cookieslogoutSchema: Validates optional refresh token for revocation on logoutAll schemas include:
\nauthMiddleware (src/middleware/auth.middleware.ts): JWT verification and user authenticationrequireVerified (src/middleware/require-verified.middleware.ts): Ensures user email is verified before accessvalidate (src/middleware/validation.middleware.ts): Zod schema validation for request body/params/queryforgotPasswordLimiter (src/middleware/rate-limit.middleware.ts): Rate limiting for password reset (3 requests per 15 minutes)resendVerificationLimiter (src/middleware/rate-limit.middleware.ts): Rate limiting for resending verification emails (3 requests per 10 minutes)src/validations/<module>.validation.ts for request validation.src/routes/<module>.routes.ts and mount them in src/routes/index.ts.src/controllers/<module>.controller.ts.src/services/<module>.service.ts and reuse AppError for controlled failures.src/repositories/<module>.repository.ts.The API uses a custom AppError class for controlled error handling:
The API implements a robust, secure Refresh Token Rotation mechanism to safely extend user sessions without compromising security.
\nRefresh tokens are configured via environment variables in .env:
ENABLE_REFRESH_TOKEN=true\nREFRESH_TOKEN_IN_JSON=true\nREFRESH_TOKEN_IN_COOKIE=true\nCOOKIE_SAME_SITE=\"strict\" # strict, lax, none\nREFRESH_TOKEN_EXPIRY=604800000 # 7 days in milliseconds\n/auth/refresh-token, their old refresh token is immediately revoked and a new pair is issued. This provides Replay Protection.HTTP-Only cookie (for web clients to prevent XSS) and/or directly in the JSON response body (for mobile apps or specialized clients).COOKIE_SAME_SITE variable. Use strict when the frontend and backend are on the exact same domain, or lax/none (along with secure: true) if operating across subdomains or entirely different domains.The API includes Cross-Origin Resource Sharing (CORS) support to allow requests from different origins (e.g., frontend applications).
\nCORS is configured via environment variables in .env:
# Comma-separated list of allowed origins\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080,https://yourdomain.com\n\n# Allow credentials (cookies, authorization headers)\nCORS_CREDENTIALS=true\nContent-Type, Authorization, X-Requested-With\n\n[!WARNING]\nProduction Security
\n\n
\n- Never use
\n*(wildcard) forALLOWED_ORIGINSin production- Only add trusted domains to the whitelist
\n- Use HTTPS for production origins (e.g.,
\nhttps://yourdomain.com)- Regularly audit the allowed origins list
\n
\n\n[!IMPORTANT]\nCredentials Configuration
\n\n
\n- Set
\nCORS_CREDENTIALS=trueonly if you're using cookies or need to send authorization headers- When credentials are enabled, you cannot use wildcard origins
\n- Frontend must include
\ncredentials: 'include'(fetch) orwithCredentials: true(axios)
Development:
\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080\nCORS_CREDENTIALS=true\nProduction:
\nALLOWED_ORIGINS=https://yourdomain.com,https://admin.yourdomain.com\nCORS_CREDENTIALS=true\nCORS Error: \"No 'Access-Control-Allow-Origin' header\"
\nALLOWED_ORIGINSsrc/app.tsCredentials Not Working:
\nCORS_CREDENTIALS=true in .envcredentials: 'include' or withCredentials: true*)The API uses Pino for structured JSON logging with comprehensive audit trails:
\nLogged Events:
\nLog Format:
\nExample Log:
\n{\n \"level\": 30,\n \"time\": 1702890637123,\n \"action\": \"user_login\",\n \"userId\": \"5ba52d7e-07f9-4b15-998f-fb1bf0885e7d\",\n \"email\": \"user@example.com\",\n \"msg\": \"User logged in successfully\"\n}\nThe project is configured to use tsup for efficient bundling.
pnpm builddist/src/server.ts and dependencies to dist/server.jsdist/views/emailspnpm startnode dist/server.jsbuild and start scripts in package.json.DATABASE_URL is set and migrations are run.Always run migrations in production before starting the app:
\npnpm prisma migrate deploy\nThe project uses Vitest for unit and integration testing.
\n# Run all tests (watch mode)\npnpm test\n\n# Run unit tests only\npnpm test:unit\n\n# Run integration tests only\npnpm test:integration\n\n# Run with coverage\npnpm exec vitest run --coverage\nTests are organized in tests/ with separate directories for unit and integration tests:
tests/\nโโโ unit/ # Unit tests (mocked dependencies)\nโ โโโ controllers/\nโ โโโ services/\nโ โโโ middleware/\nโ โโโ utils/\nโโโ integration/ # Integration tests (real database)\n โโโ helpers/ # Test utilities (DB reset)\n โโโ repositories/ # Repository tests\n โโโ routes/ # Route/endpoint tests\nIntegration tests run against a real PostgreSQL database. Ensure your DATABASE_URL points to a test database that can be safely cleared between tests.
\n\n" - }, { "fullName": "ammar-rasyidi/mandum-rimba", "name": "mandum-rimba", @@ -1345,7 +1320,7 @@ "url": "https://github.com/ammar-rasyidi/mandum-rimba", "homepage": "https://www.mandumrimba.org", "language": "TypeScript", - "stars": 19, + "stars": 20, "forks": 5, "topics": [ "conservation", @@ -1362,7 +1337,7 @@ "pmtiles", "wildlife" ], - "updatedAt": "2026-07-20T10:31:49Z", + "updatedAt": "2026-07-21T15:22:06Z", "pushedAt": "2026-07-20T10:31:33Z", "latestRelease": { "name": "Mandum Rimba v1.0.0", @@ -1379,6 +1354,31 @@ "communityHealth": 85, "readmeHtml": "[!WARNING]\nIntegration tests truncate all tables before each test. Do not run against a production database.
\n
\n ![\"Mandum](\"https://raw.githubusercontent.com/ammar-rasyidi/mandum-rimba/main/apps/web/public/images/hero_light.svg\")
\n
\n ![\"Support](\"https://img.shields.io/badge/Trakteer-Dukung-ff4d4d?style=for-the-badge\"){kind=icon}
![\"Support](\"https://img.shields.io/badge/PayPal-Support-003087?style=for-the-badge&logo=paypal&logoColor=white\"){kind=icon}
Prefer to scan directly? QRIS & GoPay
\n ![\"QRIS,](\"https://raw.githubusercontent.com/ammar-rasyidi/mandum-rimba/main/apps/web/public/images/qris_ammar.png\")
\n ย ย \n ![\"GoPay,](\"https://raw.githubusercontent.com/ammar-rasyidi/mandum-rimba/main/apps/web/public/images/gopay.png\")
\n
An independent, non-profit observatory for Indonesia's forests, land, and\nprotected wildlife. A map-first public-interest web app that distills credible\nsatellite and public data, deforestation, palm oil & mining expansion, linked\ndisasters, and the wildlife losing its home, into one open map anyone can check.
\n๐ณ Live at mandumrimba.org ยท bilingual (Indonesia / English)
\n๐ White paper, what Mandum Rimba is and why it exists:\nEnglish ยท\nBahasa Indonesia
\n\n\nEvidence over accusation. We gather and show the data as it is, and never\ndraw conclusions on anyone's behalf. We overlay official data against\nsatellite reality and let the gap speak. Every layer has a source, a date,\nand a methodology link.
\n
Live layers are checked; the rest are on the roadmap.
\nTwo tools turn a distant statistic into a neighbour: \"Yang Tinggal di\nDekatmu\" finds the nearest recorded threatened animal to your city (plus the\nnearest protected area) and renders a share card, and \"Kartu Penduduk Rimba\"\nissues a playful KTP-style resident card featuring that animal. Photos and\nlocation stay in the browser and are never uploaded or stored.
\nEvery dataset is public and independently verifiable; the in-app\nmethodology and\ndata-sources pages carry per-dataset\nlicenses, coverage, and update dates, plus an honest list of the gaps where\ncredible open data does not yet exist.
\nThe wildlife-distribution layer is an offline build: GBIF occurrence density for\nthreatened + flagship/endemic species, weighted by ESA WorldCover natural-habitat\ncover (city points dropped) and contoured per island so a species never bleeds\nonto an island it doesn't live on. Cryptic species with no public coordinates are\nshown as documented-range markers. Pre-1990 museum specimens are excluded so the\nmap reflects present-day presence. The full build is in\nscripts/species-distribution.
apps/web, Next.js 14 (App Router), MapLibre GL with vector tiles,\nnext-intl (Indonesian / English), Recharts.apps/api, NestJS: weekly ingest jobs that pull from the public sources\nabove, build vector tiles, and expose a public read-only REST API.packages/shared, shared TypeScript domain types.๐ DATA-FLOW.md explains how data moves from source to map ยท\nSETUP.md covers local setup, ingest jobs, and deployment.
\npnpm install\n\n# each app ships an .env.example, copy and fill in source API keys + service\n# connection strings (a database and an object store), then:\ncp apps/api/.env.example apps/api/.env\ncp apps/web/.env.example apps/web/.env.local\n\npnpm dev # web on :3000, api on :4000\nThe ingest jobs run on a weekly schedule; data sources without a stable machine\nendpoint are skipped cleanly when unconfigured, so the app runs with whatever\nsubset you have keys for.
\nThe code is open; the brand is the maintainer's. The two are licensed separately.
\nGovernance is a single-steward model (GOVERNANCE.md) and\ncontributions are under the Contributor License Agreement.
\n" }, + { + "fullName": "rayasabari/yntk-ts", + "name": "yntk-ts", + "owner": "rayasabari", + "ownerAvatarUrl": "https://avatars.githubusercontent.com/u/57746279?v=4", + "description": "You Need This Kit - Type-safe Starter: REST API boilerplate that ships with User Management, Role Management, and Authentication with JWTs, Prisma ORM, and PostgreSQL", + "metaDescription": "You Need This Kit - Type-safe Starter: REST API boilerplate that ships with User Management, Role Management, and Authentication with JWTs, Prisma ORMโฆ", + "url": "https://github.com/rayasabari/yntk-ts", + "homepage": "", + "language": "TypeScript", + "stars": 20, + "forks": 0, + "topics": [], + "updatedAt": "2026-06-25T07:10:46Z", + "pushedAt": "2026-04-08T09:29:42Z", + "latestRelease": null, + "archived": false, + "licenseSpdx": "", + "createdAt": "2026-04-08T09:27:49Z", + "openIssues": 0, + "openPullRequests": 0, + "subscribers": 0, + "communityHealth": 28, + "readmeHtml": "TypeScript/Express REST API starter kit that ships with User Management, Role Management, and Authentication with JWTs, Prisma ORM, and PostgreSQL. The project is organized by feature with explicit service/repository layers so business logic stays separated from transport concerns.
\nsrc/\nโโโ app.ts # Express app bootstrap\nโโโ server.ts # Starts HTTP server\nโโโ config/ # Environment loader & Prisma client wrapper\nโโโ controllers/ # HTTP handlers grouped by module + shared helpers\nโโโ services/ # Business logic (auth/user) & mappers\nโโโ repositories/ # Prisma data access per module\nโโโ middleware/ # Cross-cutting middleware (auth, validation)\nโโโ routes/ # Express routers mounted under /auth and /users\nโโโ validations/ # Zod schemas for request validation\nโโโ views/ # Email templates\nโโโ errors/ # Custom AppError type & error utilities\nโโโ utils/ # Shared utilities (password, string, Zod helpers)\nโโโ types/ # Shared TS types & Express module augmentation\npnpm install\nCreate .env (never commit it) with the required settings:
# Server Configuration\nNODE_ENV=development\nFRONTEND_URL=http://localhost:8080\nPORT=5050\nLOG_LEVEL=info\n\n# Database\nDATABASE_URL=postgresql://USER:PASSWORD@HOST:PORT/DATABASE\n\n# JWT Configuration\nJWT_SECRET=super-secret\n\n# Bcrypt Configuration\nSALT_ROUNDS=10\n\n# Email Configuration\nEMAIL_HOST=smtp.gmail.com\nEMAIL_PORT=587\nEMAIL_USER=your-email@gmail.com\nEMAIL_PASSWORD=your-app-password\nEMAIL_FROM=noreply@yourapp.com\n\n# Token Configuration\nACCESS_TOKEN_EXPIRY=\"1h\"\nENABLE_REFRESH_TOKEN=true\nREFRESH_TOKEN_IN_JSON=true\nREFRESH_TOKEN_IN_COOKIE=true\nREFRESH_TOKEN_EXPIRY=7 * 24 * 60 * 60 * 1000 # 7 days in miliseconds\nCOOKIE_SAME_SITE=\"strict\" # strict, lax, none\nRESET_PASSWORD_TOKEN_EXPIRY=1 * 60 * 60 * 1000 # 1 hour in miliseconds\nEMAIL_VERIFICATION_TOKEN_EXPIRY=24 * 60 * 60 * 10000 # 24 hours in miliseconds\n\n# CORS Configuration\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080,https://yourdomain.com\nCORS_CREDENTIALS=true\n\n\nNote for Gmail: Use an App Password instead of your regular password. Enable 2FA and generate an App Password in Google Account Settings โ Security โ App passwords.
\n
prisma/schema.prisma.pnpm prisma migrate dev (for local) or pnpm prisma db push for quick sync.pnpm prisma generate. Output lands in src/generated/prisma.pnpm prisma db seedpnpm dev\nRuns tsx in watch mode, recompiling on changes. The API listens on PORT from the env file (defaults to 5050).
| Command | \nPurpose | \n
|---|---|
pnpm dev | \nStart the API in watch mode with tsx | \n
pnpm prisma migrate dev | \nCreate/apply migrations and regenerate Prisma client | \n
pnpm prisma generate | \nRegenerate Prisma client manually | \n
pnpm prisma db push | \nQuick sync schema to database without migrations | \n
pnpm prisma db seed | \nSeed the database | \n
pnpm build | \nBuild the API for production (bundles with tsup) | \n
pnpm start | \nStart the production server from dist/ | \n
pnpm test | \nRun all tests in watch mode | \n
pnpm test:unit | \nRun unit tests only | \n
pnpm test:integration | \nRun integration tests only (sequential) | \n
\n\nโ Production build: The repo currently runs via tsx; add a
\ntscbuild +startscript before deploying to production environments like Vercel/Node runtime functions.
The API includes interactive documentation powered by Swagger UI and OpenAPI 3.0 (swagger-jsdoc and swagger-ui-express).
/api-docs (accessible when the server is running)All endpoints respond with { status, message, data? } JSON payloads.\nFor paginated endpoints (like GET /users and GET /roles), the response also includes meta and links objects containing paging data and HATEOAS navigational URLs. They optionally accept query parameters: ?page=1&limit=10&sortBy=createdAt&sortOrder=asc&search=value.
/auth)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
POST | \n/auth/register | \nRegister a new user and send verification email | \nPublic | \nregisterUserSchema | \n
POST | \n/auth/login | \nVerify credentials and return JWT token | \nPublic | \n- | \n
POST | \n/auth/logout | \nLogout and revoke refresh token | \nRequired | \nlogoutSchema | \n
POST | \n/auth/refresh-token | \nRefresh access token using refresh token | \nPublic | \nrefreshTokenSchema | \n
POST | \n/auth/verify-email | \nVerify email address using token from email | \nPublic | \nverifyEmailSchema | \n
POST | \n/auth/resend-verification | \nResend verification email (rate limited: 3/10min) | \nPublic | \nresendVerificationSchema | \n
POST | \n/auth/forgot-password | \nRequest password reset email (rate limited: 3/15min) | \nPublic | \nforgotPasswordSchema | \n
POST | \n/auth/reset-password | \nReset password using token from email | \nPublic | \nresetPasswordSchema | \n
/users)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
POST | \n/users | \nCreate a user (admin-style) | \nRequired (users:create) | \ncreateUserSchema | \n
GET | \n/users | \nList all users | \nRequired (users:read) | \n- | \n
GET | \n/users/:id | \nFetch a user by ID | \nRequired (users:read) | \n- | \n
PUT | \n/users/:id | \nUpdate user fields (username, email, displayName) | \nRequired (users:update) | \nupdateUserSchema | \n
PUT | \n/users/:id/roles | \nAssign roles to a user | \nRequired (roles:assign) | \nassignRolesSchema | \n
PATCH | \n/users/password | \nUpdate current user's password | \nRequired (users:update) | \nupdatePasswordSchema | \n
DELETE | \n/users/:id | \nRemove a user | \nRequired (users:delete) | \n- | \n
/roles & /permissions)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
GET | \n/roles | \nList all roles | \nRequired (roles:read) | \n- | \n
GET | \n/roles/:id | \nFetch a role by ID | \nRequired (roles:read) | \n- | \n
POST | \n/roles | \nCreate a new role | \nRequired (roles:create) | \ncreateRoleSchema | \n
PUT | \n/roles/:id | \nUpdate an existing role | \nRequired (roles:update) | \nupdateRoleSchema | \n
DELETE | \n/roles/:id | \nRemove a role | \nRequired (roles:delete) | \n- | \n
GET | \n/permissions | \nList all system permissions | \nRequired (roles:read) | \n- | \n
Auth Required: Endpoints require Authorization: Bearer <token> header.
The API uses Zod for request validation with the following schemas:
\nregisterUserSchema: Validates user registration (username, email, displayName, password)createUserSchema: Validates user creation (username, email, displayName)updateUserSchema: Validates user updates (partial fields)assignRolesSchema: Validates assigning role IDs to a userupdatePasswordSchema: Validates password changes (currentPassword, newPassword, confirmPassword)createRoleSchema: Validates new role details and array of permission IDsupdateRoleSchema: Validates partial updates to a roleverifyEmailSchema: Validates email verification tokenresendVerificationSchema: Validates email for resending verificationforgotPasswordSchema: Validates email for password reset requestsresetPasswordSchema: Validates reset token and new password (min 8 chars, uppercase, lowercase, number)refreshTokenSchema: Validates refresh token from body or cookieslogoutSchema: Validates optional refresh token for revocation on logoutAll schemas include:
\nauthMiddleware (src/middleware/auth.middleware.ts): JWT verification and user authenticationrequireVerified (src/middleware/require-verified.middleware.ts): Ensures user email is verified before accessvalidate (src/middleware/validation.middleware.ts): Zod schema validation for request body/params/queryforgotPasswordLimiter (src/middleware/rate-limit.middleware.ts): Rate limiting for password reset (3 requests per 15 minutes)resendVerificationLimiter (src/middleware/rate-limit.middleware.ts): Rate limiting for resending verification emails (3 requests per 10 minutes)src/validations/<module>.validation.ts for request validation.src/routes/<module>.routes.ts and mount them in src/routes/index.ts.src/controllers/<module>.controller.ts.src/services/<module>.service.ts and reuse AppError for controlled failures.src/repositories/<module>.repository.ts.The API uses a custom AppError class for controlled error handling:
The API implements a robust, secure Refresh Token Rotation mechanism to safely extend user sessions without compromising security.
\nRefresh tokens are configured via environment variables in .env:
ENABLE_REFRESH_TOKEN=true\nREFRESH_TOKEN_IN_JSON=true\nREFRESH_TOKEN_IN_COOKIE=true\nCOOKIE_SAME_SITE=\"strict\" # strict, lax, none\nREFRESH_TOKEN_EXPIRY=604800000 # 7 days in milliseconds\n/auth/refresh-token, their old refresh token is immediately revoked and a new pair is issued. This provides Replay Protection.HTTP-Only cookie (for web clients to prevent XSS) and/or directly in the JSON response body (for mobile apps or specialized clients).COOKIE_SAME_SITE variable. Use strict when the frontend and backend are on the exact same domain, or lax/none (along with secure: true) if operating across subdomains or entirely different domains.The API includes Cross-Origin Resource Sharing (CORS) support to allow requests from different origins (e.g., frontend applications).
\nCORS is configured via environment variables in .env:
# Comma-separated list of allowed origins\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080,https://yourdomain.com\n\n# Allow credentials (cookies, authorization headers)\nCORS_CREDENTIALS=true\nContent-Type, Authorization, X-Requested-With\n\n[!WARNING]\nProduction Security
\n\n
\n- Never use
\n*(wildcard) forALLOWED_ORIGINSin production- Only add trusted domains to the whitelist
\n- Use HTTPS for production origins (e.g.,
\nhttps://yourdomain.com)- Regularly audit the allowed origins list
\n
\n\n[!IMPORTANT]\nCredentials Configuration
\n\n
\n- Set
\nCORS_CREDENTIALS=trueonly if you're using cookies or need to send authorization headers- When credentials are enabled, you cannot use wildcard origins
\n- Frontend must include
\ncredentials: 'include'(fetch) orwithCredentials: true(axios)
Development:
\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080\nCORS_CREDENTIALS=true\nProduction:
\nALLOWED_ORIGINS=https://yourdomain.com,https://admin.yourdomain.com\nCORS_CREDENTIALS=true\nCORS Error: \"No 'Access-Control-Allow-Origin' header\"
\nALLOWED_ORIGINSsrc/app.tsCredentials Not Working:
\nCORS_CREDENTIALS=true in .envcredentials: 'include' or withCredentials: true*)The API uses Pino for structured JSON logging with comprehensive audit trails:
\nLogged Events:
\nLog Format:
\nExample Log:
\n{\n \"level\": 30,\n \"time\": 1702890637123,\n \"action\": \"user_login\",\n \"userId\": \"5ba52d7e-07f9-4b15-998f-fb1bf0885e7d\",\n \"email\": \"user@example.com\",\n \"msg\": \"User logged in successfully\"\n}\nThe project is configured to use tsup for efficient bundling.
pnpm builddist/src/server.ts and dependencies to dist/server.jsdist/views/emailspnpm startnode dist/server.jsbuild and start scripts in package.json.DATABASE_URL is set and migrations are run.Always run migrations in production before starting the app:
\npnpm prisma migrate deploy\nThe project uses Vitest for unit and integration testing.
\n# Run all tests (watch mode)\npnpm test\n\n# Run unit tests only\npnpm test:unit\n\n# Run integration tests only\npnpm test:integration\n\n# Run with coverage\npnpm exec vitest run --coverage\nTests are organized in tests/ with separate directories for unit and integration tests:
tests/\nโโโ unit/ # Unit tests (mocked dependencies)\nโ โโโ controllers/\nโ โโโ services/\nโ โโโ middleware/\nโ โโโ utils/\nโโโ integration/ # Integration tests (real database)\n โโโ helpers/ # Test utilities (DB reset)\n โโโ repositories/ # Repository tests\n โโโ routes/ # Route/endpoint tests\nIntegration tests run against a real PostgreSQL database. Ensure your DATABASE_URL points to a test database that can be safely cleared between tests.
\n\n" + }, { "fullName": "rizukirr/hyprsimple", "name": "hyprsimple", @@ -1634,7 +1634,7 @@ "windows-desktop" ], "updatedAt": "2026-07-21T05:40:00Z", - "pushedAt": "2026-07-21T05:44:30Z", + "pushedAt": "2026-07-21T17:29:02Z", "latestRelease": { "name": "v0.3.2", "tagName": "v0.3.2", @@ -1645,7 +1645,7 @@ "licenseSpdx": "MIT", "createdAt": "2026-02-23T13:59:12Z", "openIssues": 1, - "openPullRequests": 0, + "openPullRequests": 1, "subscribers": 0, "communityHealth": 75, "readmeHtml": "[!WARNING]\nIntegration tests truncate all tables before each test. Do not run against a production database.
\n
Muslimtify keeps you consistent with your daily prayers by delivering accurate prayer times and timely desktop notifications. Designed for Linux and Windows, it automatically calculates prayer schedules and reminds you 30, 15, and 5 minutes before the Adhan โ or at your own custom intervals โ and when it's time to pray. All calculations run locally, requiring no internet connection or external services.
\nMuslimtify supports 21 international calculation methods including MWL, ISNA, Umm al-Qura (Makkah), Egyptian General Authority, Kemenag (Indonesia), JAKIM (Malaysia), Diyanet (Turkey), and more. The default method is Kemenag. With persistent configuration and minimal setup, Muslimtify integrates seamlessly into your daily routine without interrupting your workflow.
\n\n\n[!Note]\nPrayer time calculations are powered by libmuslim, a portable library extracted from this project to enable a more flexible and reusable ecosystem for Muslim developers.
\n
| Linux | \nWindows | \n
|---|---|
\n\nRoadmap
\n\n
\n- \n
Merge commandlocation autoandmethod autointo (only)config autoand optimize auto detection per (249) country- \n
Refactor from timer-driven into a portable long-running loop- \n
Add custom adzan sound notifications- \n
Re-design command-line (BREAKING CHANGES)- Add read lat/long from user GPS
\n- Add GUI (see branch gui to see a progress)
\n- Distribute to Flatpak
\n- MacOS support (if devices is available)
\n- Wearable Device support
\n- Embedded Device Support
\n
\n\n[!Important]\nThis project is available for Linux and Windows users, but not yet for Mac users because we need a Mac device to make Muslimtify run on macOS. We are looking for brothers and sisters who have a Mac and experience in low-level C programming to contribute to the project and help bring Muslimtify to macOS. Alternatively, you can support us via GitHub Sponsors in the sponsor section.
\n
Every release ships ready-to-run binaries for Linux and Windows on the\nReleases page.
\nLinux (x86_64 or aarch64) โ the binaries are dynamically linked, so\ninstall the runtime libraries first, then extract and install:
# Ubuntu/Debian\nsudo apt install libnotify4 libcurl4\n# Fedora/RHEL\nsudo dnf install libnotify libcurl\n# Arch\nsudo pacman -S libnotify curl\n\ntar xzf muslimtify-<version>-linux-<arch>.tar.gz\nsudo cp -r muslimtify-<version>-linux-<arch>/{bin,lib,share} /usr/local/\nmuslimtify daemon install\nWindows (x64 or arm64) โ download and run the matching installer:
muslimtify-<version>-setup-x64.exe # Intel/AMD\nmuslimtify-<version>-setup-arm64.exe # ARM\nVerify any download against the published checksums:
\nsha256sum -c SHA256SUMS\nyay -S muslimtify\nsudo dnf copr enable rizukirr/muslimtify\nsudo dnf install muslimtify\nsudo add-apt-repository ppa:rizukirr/muslimtify\nsudo apt update\nsudo apt install muslimtify\nInstall dependencies:
\n# Ubuntu/Debian\nsudo apt install git build-essential cmake pkg-config libnotify-dev libcurl4-openssl-dev\n\n# Fedora/RHEL\nsudo dnf install git gcc cmake pkgconfig libnotify-devel libcurl-devel\n\n# Arch Linux\nsudo pacman -S git base-devel cmake pkgconfig libnotify curl\nClone, install, and enable background checks:
\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\nsudo ./install.sh\nmuslimtify daemon install\nwinget install muslimtify\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\n.\\install.ps1\nmuslimtify daemon install\nTo remove the Windows install later, run .\\uninstall.ps1.
If you prefer building manually first:
\ncmake -S . -B build\ncmake --build build --config Release\ncmake --install build --config Release\nmuslimtify daemon install\nRun muslimtify daemon status to check if Muslimtify is registered with systemd. If no status is found, run muslimtify daemon install to register the service and ensure it runs as expected.
Muslimtify automatically selects the standard prayer time calculation method based on your country and location. Run muslimtify to verify that your configuration is correct. If the automatic selection does not meet your needs, you can set it manually using muslimtify method <key-method>. A full list of available methods is documented here.
Muslimtify can be configured with CLI commands or by editing config.json\nmanually.
Config paths:
\n~/.config/muslimtify/config.json~/.cache/muslimtify%APPDATA%\\muslimtify\\config.json%LOCALAPPDATA%\\muslimtifyCommon setup commands:
\nmuslimtify location set --auto # detect location from IP\nmuslimtify location set --auto --city=Mansoura # auto-detect but use your own city label\nmuslimtify method --auto # select method from the detected country\nmuslimtify location set --lat=-6.175 --long=106.82 # set location manually (uses system timezone)\nmuslimtify location set --timezone=Asia/Jakarta # override timezone\nmuslimtify location set --city=Jakarta # add a city label\nmuslimtify location set --refresh-interval=21600 # re-check location every 6h (0=off, min 3600)\nmuslimtify method --list # list all available calculation methods\nmuslimtify method mwl # set calculation method\nmuslimtify madzhab hanafi # set madzhab (shafi/hanafi)\nmuslimtify notification --reminder --all 30 15 5 # set every prayer's reminders (minutes before adhan)\nmuslimtify notification --reminder fajr 30 15 5 # set reminders for a single prayer\nmuslimtify notification # show current notification settings\nmuslimtify location # show current location\nResetting the configuration is done by deleting config.json. Muslimtify falls\nback to built-in defaults when the file is missing, and rewrites it the next\ntime you change a setting. Validation runs automatically every time the config\nis loaded.
Muslimtify supports the following calculation methods:
\n| Key | \nMethod | \nRegion | \n
|---|---|---|
mwl | \nMuslim World League | \nEurope, Far East | \n
makkah | \nUmm al-Qura, Makkah | \nArabian Peninsula | \n
isna | \nISNA | \nNorth America | \n
egypt | \nEgyptian General Authority | \nAfrica, Middle East | \n
karachi | \nUniv. Islamic Sciences, Karachi | \nPakistan, India, Bangladesh | \n
turkey | \nDiyanet, Turkey | \nTurkey | \n
singapore | \nMUIS, Singapore | \nSingapore | \n
jakim | \nJAKIM, Malaysia | \nMalaysia | \n
kemenag | \nKEMENAG, Indonesia | \nIndonesia (default) | \n
france | \nUOIF, France | \nFrance | \n
russia | \nSpiritual Admin., Russia | \nRussia | \n
dubai | \nGAIAE, Dubai | \nUAE | \n
qatar | \nMin. of Awqaf, Qatar | \nQatar | \n
kuwait | \nMin. of Awqaf, Kuwait | \nKuwait | \n
jordan | \nMin. of Awqaf, Jordan | \nJordan | \n
gulf | \nGulf Region | \nGulf states | \n
tunisia | \nMin. of Religious Affairs | \nTunisia | \n
algeria | \nMin. of Religious Affairs | \nAlgeria | \n
morocco | \nMin. of Habous, Morocco | \nMorocco | \n
portugal | \nComunidade Islamica de Lisboa | \nPortugal | \n
moonsighting | \nMoonsighting Committee | \nWorldwide | \n
You can also use a custom method by setting \"method\": \"custom\" in config.json with your own fajr_angle and isha_angle values.
Manual JSON editing is useful when you want precise control over enabled\nprayers, reminder offsets, notification settings, or location data.
\n\nDefault config.json{\n \"location\": {\n \"latitude\": 0.0,\n \"longitude\": 0.0,\n \"timezone\": \"UTC\",\n \"timezone_offset\": 0.0,\n \"auto_detect\": true,\n \"city\": \"\",\n \"country\": \"\"\n },\n \"prayers\": {\n \"fajr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"sunrise\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuha\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuhr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"asr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"maghrib\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"isha\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n }\n },\n \"notification\": {\n \"timeout\": 5000,\n \"urgency\": \"critical\",\n \"sound\": \"adhan\",\n \"sound_alarm\": \"alarm\",\n \"sound_reminder\": \"reminder\",\n \"icon\": \"muslimtify\"\n },\n \"calculation\": {\n \"method\": \"kemenag\",\n \"madhab\": \"shafi\"\n }\n}\nmuslimtify daemon status to confirm the background service is running.notify-send \"Test\" \"Hello\".muslimtify location set --auto again.muslimtify location set --lat=<latitude> --long=<longitude>.\nIf the host machine is in a different region than the coordinates, override\nthe timezone with --timezone=<iana>, e.g.\nmuslimtify location set --lat=-6.21 --long=106.84 --timezone=Asia/Jakarta.ipinfo.io if auto detection keeps failing.Contributions are welcome. See CONTRIBUTING.md for workflow,\nstyle, and testing guidance.
\nMuslimtify is released under the MIT License. See the repository license files\nfor details.
\n\n \n
Give your AI a memory that never leaves your machine.
\n\n Your AI forgets everything between sessions. Uteke fixes that โ one binary, fully offline, ~45ms recall.\n
\n \n \n \n \n \n
\n ๐ฌ๐ง English ยท ๐ฎ๐ฉ Bahasa Indonesia\n
# Install (macOS, Linux, Windows)\ncurl -sSL codecora.dev/install | sh\n\n# Store a memory\nuteke remember \"Deploy v2.1 to staging at 3pm\"\n\n# Search it back โ by meaning, not just keywords\nuteke recall \"when do we deploy?\"\nThat's it. No API keys. No Docker. No Python. No cloud.
\nFirst run downloads the embedding model (~188MB, one-time) and you're running.
\nWant richer memories? Add metadata:
\nuteke remember \"Deploy v2.1 to staging\" \\\n --tags deploy,staging \\\n --entity staging-server \\\n --category infrastructure\n| Method | \nCommand | \n
|---|---|
| Homebrew | \nbrew install codecoradev/tap/uteke | \n
| Cargo | \ncargo install uteke-cli | \n
| Docker | \ndocker run -d -p 127.0.0.1:8767:8767 -v uteke-data:/data ghcr.io/codecoradev/uteke:latest | \n
| Binary | \nGitHub Releases (macOS, Linux, Windows) | \n
๐ Full install guide ยท Docker docs
\n\n\nAre you an AI agent reading this? Your user just shared this repo link with you.\nRun
\nuteke onboardto interactively configure uteke for your environment โ it will\ndetect your install, ask which agent you are, toggle features, write config, and\nshow you every command available. If non-interactive:uteke onboard --yes --agent <your-agent>.
New to uteke? One command takes you from zero to productive:
\nuteke onboard\nThe wizard will:
\n~/.uteke/uteke.toml with your selectionsuteke init)Non-interactive (CI, scripts, AI agents):
\nuteke onboard --yes --agent hermes --namespace default\n๐ Full onboarding docs ยท CLI reference
\nYou just spent 2 hours explaining your codebase to ChatGPT. Next session? Blank slate. Again.
\nEvery AI tool forgets. Context windows fill up, sessions end, and your AI starts over every single time. Uteke gives it persistent memory โ and keeps it on your machine.
\n| \n | Uteke | \nMnemosyne | \nMem0 | \nAgentMemory | \nLetta | \nZep | \nEngram | \n
|---|---|---|---|---|---|---|---|
| Language | \nRust (single binary) | \nPython (pip) | \nPython | \nTypeScript | \nPython | \nPython | \nGo (single binary) | \n
| Setup | \nOne binary (curl | sh) | \npip install + venv | \npip + Docker + Qdrant | \nnpm + Docker (iii-engine) | \npip + Docker + Postgres | \npip + Docker + Neo4j | \nOne binary | \n
| API keys | \nโ None | \nโ ๏ธ For remote embeddings | \nโ OpenAI/LLM | \nโ LLM key | \nโ LLM key | \nโ LLM key | \nโ None | \n
| Works offline | \nโ Fully | \nโ ๏ธ Optional | \nโ Cloud embedding | \nโ Needs LLM | \nโ Needs LLM | \nโ Needs LLM + vector DB | \nโ Fully | \n
| Search | \nHybrid (Vector + FTS5 + RRF) | \nsqlite-vec + FTS5 | \nVector + Graph | \nVector + Graph | \nVector | \nTemporal Graph | \nFTS5 only | \n
| Recall speed | \n~45ms | \n~50ms+ | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \n~Fast (local) | \n
| Multi-agent | \nโ Rooms (built-in collaboration) | \nโ ๏ธ Shared API | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Time-travel | \nโ Native point-in-time | \nโ ๏ธ Temporal triples | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| MCP server | \nโ JSON-RPC + HTTP | \nโ stdio + SSE | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Your data | \nโ Never leaves machine | \nโ Local-first | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ Local | \n
| License | \nApache 2.0 | \nMIT | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \n
\n\nUteke vs Mnemosyne: Both are local-first with semantic + FTS5 search. Mnemosyne is the closest competitor (~1.5K stars, Python). Uteke wins on single binary (no Python runtime), rooms, time-travel queries, and zero runtime dependencies.
\n
\n\nUteke vs Engram: Both are single-binary, offline, no-API-key tools. But Engram is FTS5-only (keyword search). Uteke adds vector semantic search + RRF fusion + rooms + time-travel + graph relationships + smart decay + document engine + batch import. Same simplicity thesis, 10ร the features.
\n
\n\nUteke vs AgentMemory/Mem0/Letta/Zep: Those are powerful โ but all require cloud LLM API keys and Docker infrastructure. Your data goes to OpenAI/Anthropic. Uteke runs fully offline with local ONNX embeddings. No Docker, no Python, no API keys.
\n
\n \n
๐ค Building AI agents? Give them persistent memory without cloud dependencies. Your agent remembers user preferences, past decisions, and context โ across sessions, fully offline.
\n๐ฅ Working in a team? Use Rooms to share knowledge. Meeting notes, project decisions, architecture choices โ searchable by everyone, attributed by author.
\n๐ Building for privacy-sensitive domains? Healthcare, finance, legal โ data stays on your machine. No API calls, no telemetry, no cloud. Local embeddings (ONNX, 768d).
\nโจ๏ธ Power user who lives in the terminal? Uteke is your personal knowledge graph. Remember anything, recall by meaning, link related thoughts. All from the command line.
\n| Feature | \nWhat it does | \n
|---|---|
| ๐ง Hybrid Search | \nVector similarity + FTS5 full-text search, merged by Reciprocal Rank Fusion (RRF). Finds by meaning AND exact keywords. | \n
| ๐ Rooms | \nGroup memories by context (meetings, projects, clients) with author attribution. | \n
| โณ Time-travel | \nRecall memories as they existed at any point in time. uteke recall \"deploy\" --at 2025-01-15 | \n
| ๐ท๏ธ Rich Metadata | \nTags, entities, categories, key:value pairs on every memory. | \n
| ๐งฉ Memory Types | \nTyped categories (fact, procedure, decision, etc.) with auto-inference. | \n
| โ๏ธ Partial Updates | \nUpdate content, tags, metadata, importance, or type without full rewrite. | \n
| ๐ Citations | \nSource attribution on every memory (URL, file, user, import batch). | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ Relationship Graph | \nLink memories with typed edges (supersedes, contradicts, references). Auto-backlinks. | \n
| ๐ Cross-Entity Linking | \nBidirectional memoryโdocument references via [[doc-slug]] wikilinks. | \n
| ๐ค Cosine Auto-Linking | \nAutomatically creates similar_to edges between related memories. | \n
| ๐ Smart Decay | \nComposite importance scoring. Pin what matters, let stale memories fade. | \n
| ๐ Salience + Recency | \nDual-axis recall boost by memory type and age. | \n
| ๐ Orphan Detection | \nFind disconnected, low-importance memories for cleanup. | \n
| ๐ Dream Cycle | \nOne-command maintenance: lint โ backlinks โ dedup โ orphans. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ MCP Server | \nJSON-RPC over stdio + Streamable HTTP. Works with Claude Code, Cursor, Hermes. | \n
| ๐ฅ๏ธ Server Mode | \nPersistent daemon โ eliminates cold-start embedding load on every call. | \n
| ๐ Batch Import | \nImport entire directories with auto-strategy routing (document vs. memory extraction). | \n
| ๐ Document Engine | \nWiki/knowledge base with uteke doc create/get/list and auto-chunking. | \n
| ๐ฅ Import/Export | \nJSONL-based backup and restore. | \n
| ๐ View-Only API Keys | \nRead-only tokens for safe GET-only access to the server. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ฆ Single Binary | \nZero dependencies. No Docker needed, no database server, no Python, no API keys. | \n
| ๐ Fully Offline | \nLocal ONNX embeddings (EmbeddingGemma Q4, 768d). No telemetry, no cloud. | \n
| โก Recall Cache | \nLRU cache eliminates redundant embedding for repeated queries. | \n
| ๐ฅ Tiered Memory | \nHot/Warm/Cold tracking with auto-cleanup of stale memories. | \n
| ๐ Embed Fallback | \nGracefully degrades to no-op embedder if local model fails (never crashes). | \n
| ๐ฅ Multi-Agent Namespaces | \nFully isolated memory per agent, zero overhead. | \n
| ๐ Benchmarks | \nBuilt-in uteke bench for perf testing. See results. | \n
// .mcp.json (Claude Code, Cursor)\n{ \"mcpServers\": { \"uteke\": { \"command\": \"uteke-mcp\" } } }\nFor Claude Desktop, Hermes, and HTTP transport, see MCP docs.
\n๐ Full documentation ยท CLI reference ยท Configuration
\ngraph LR\n Input[User Query] --> Embed[Local ONNX Embedder<br/>768d, EmbeddingGemma Q4]\n Embed --> HNSW[HNSW Vector Index<br/>usearch]\n Embed --> FTS5[FTS5 Full-Text<br/>SQLite]\n HNSW --> RRF[Reciprocal Rank Fusion<br/>k=60]\n FTS5 --> RRF\n RRF --> Results[Ranked Results]\n\n style Input fill:#4a9eff,color:#fff\n style Results fill:#4aff9e,color:#000\n style RRF fill:#ff9e4a,color:#fff\nHow hybrid search works:
\nEverything runs in-process. No network. No cloud. No server required (unless you want server mode).
\n\n \n
Mem0 and Letta are great โ but they require cloud API keys (OpenAI/LLM) and external infrastructure (Docker, Postgres, Qdrant). Your data gets sent to a cloud LLM provider. Uteke is a single binary with zero API keys. All embeddings run locally via ONNX. Your data never leaves your machine. See comparison table.
\n\nHow is Uteke different from AgentMemory?AgentMemory (25K stars) is a TypeScript/Node.js platform with 53 MCP tools and 12 auto-hooks. It's feature-rich but requires Docker + the iii-engine + LLM API keys. Uteke is Rust, zero dependencies, and works fully offline. If you want maximum integrations and don't mind cloud dependency โ AgentMemory. If you want privacy, speed, and zero setup โ Uteke.
\n\nHow is Uteke different from Engram?Engram (2.4K stars, Go) shares our philosophy: single binary, zero deps, MCP server, local-first. The key difference is search: Engram uses FTS5 only (keyword matching). Uteke uses hybrid search (HNSW vector similarity + FTS5 + Reciprocal Rank Fusion) โ meaning you can search by meaning, not just exact words. Uteke also adds rooms, time-travel, graph relationships, smart decay, document engine, and batch import.
\n\nWhat can Uteke remember?Anything text-based: decisions, meeting notes, code snippets, project context, personal notes, agent state. You can tag, categorize, and link memories. The --batch-dir flag lets you import entire document directories.
Yes. The embedding model (EmbeddingGemma Q4, 768d) downloads once (~188MB) on first run. After that, zero network calls. No telemetry. If the local model fails, Uteke degrades gracefully to a no-op embedder โ it never crashes and never calls a cloud API.
\n\nHow fast is recall?~45ms as a library (measured at 100โ10K memories). No network round-trip because everything is local. The LRU recall cache eliminates redundant embedding computation for repeated queries.
\n\nCan I use Uteke with my existing AI tools?Yes. Uteke ships with an MCP server that works with Claude Code, Cursor, and Hermes. You can also use the HTTP API directly in any language. See MCP setup โ
\n\nIs it production-ready?Uteke is at v0.7.3 with 206 tests, CI/CD on every commit, and benchmark harness. It's used in production by the CodeCora team and other early adopters. Still in 0.x โ expect rough edges, but the core is stable.
\ncargo build --workspace # Build\ncargo test --workspace # Test (206 tests)\ncargo clippy -- -D warnings # Lint\ncargo fmt # Format\nContributions welcome! Read CONTRIBUTING.md for the full guide.
\nApache License 2.0 โ use it, fork it, ship it.
\n\n Found this useful? โญ Star this repo โ it helps others discover Uteke.\n
\n\n \n
\n \n
Give your AI a memory that never leaves your machine.
\n\n Your AI forgets everything between sessions. Uteke fixes that โ one binary, fully offline, ~45ms recall.\n
\n \n \n \n \n \n
\n ๐ฌ๐ง English ยท ๐ฎ๐ฉ Bahasa Indonesia\n
# Install (macOS, Linux, Windows)\ncurl -sSL codecora.dev/install | sh\n\n# Store a memory\nuteke remember \"Deploy v2.1 to staging at 3pm\"\n\n# Search it back โ by meaning, not just keywords\nuteke recall \"when do we deploy?\"\nThat's it. No API keys. No Docker. No Python. No cloud.
\nFirst run downloads the embedding model (~188MB, one-time) and you're running.
\nWant richer memories? Add metadata:
\nuteke remember \"Deploy v2.1 to staging\" \\\n --tags deploy,staging \\\n --entity staging-server \\\n --category infrastructure\n| Method | \nCommand | \n
|---|---|
| Homebrew | \nbrew install codecoradev/tap/uteke | \n
| Cargo | \ncargo install uteke-cli | \n
| Docker | \ndocker run -d -p 127.0.0.1:8767:8767 -v uteke-data:/data ghcr.io/codecoradev/uteke:latest | \n
| Binary | \nGitHub Releases (macOS, Linux, Windows) | \n
๐ Full install guide ยท Docker docs
\n\n\nAre you an AI agent reading this? Your user just shared this repo link with you.\nRun
\nuteke onboardto interactively configure uteke for your environment โ it will\ndetect your install, ask which agent you are, toggle features, write config, and\nshow you every command available. If non-interactive:uteke onboard --yes --agent <your-agent>.
New to uteke? One command takes you from zero to productive:
\nuteke onboard\nThe wizard will:
\n~/.uteke/uteke.toml with your selectionsuteke init)Non-interactive (CI, scripts, AI agents):
\nuteke onboard --yes --agent hermes --namespace default\n๐ Full onboarding docs ยท CLI reference
\nYou just spent 2 hours explaining your codebase to ChatGPT. Next session? Blank slate. Again.
\nEvery AI tool forgets. Context windows fill up, sessions end, and your AI starts over every single time. Uteke gives it persistent memory โ and keeps it on your machine.
\n| \n | Uteke | \nMnemosyne | \nMem0 | \nAgentMemory | \nLetta | \nZep | \nEngram | \n
|---|---|---|---|---|---|---|---|
| Language | \nRust (single binary) | \nPython (pip) | \nPython | \nTypeScript | \nPython | \nPython | \nGo (single binary) | \n
| Setup | \nOne binary (curl | sh) | \npip install + venv | \npip + Docker + Qdrant | \nnpm + Docker (iii-engine) | \npip + Docker + Postgres | \npip + Docker + Neo4j | \nOne binary | \n
| API keys | \nโ None | \nโ ๏ธ For remote embeddings | \nโ OpenAI/LLM | \nโ LLM key | \nโ LLM key | \nโ LLM key | \nโ None | \n
| Works offline | \nโ Fully | \nโ ๏ธ Optional | \nโ Cloud embedding | \nโ Needs LLM | \nโ Needs LLM | \nโ Needs LLM + vector DB | \nโ Fully | \n
| Search | \nHybrid (Vector + FTS5 + RRF) | \nsqlite-vec + FTS5 | \nVector + Graph | \nVector + Graph | \nVector | \nTemporal Graph | \nFTS5 only | \n
| Recall speed | \n~45ms | \n~50ms+ | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \n~Fast (local) | \n
| Multi-agent | \nโ Rooms (built-in collaboration) | \nโ ๏ธ Shared API | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Time-travel | \nโ Native point-in-time | \nโ ๏ธ Temporal triples | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| MCP server | \nโ JSON-RPC + HTTP | \nโ stdio + SSE | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Your data | \nโ Never leaves machine | \nโ Local-first | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ Local | \n
| License | \nApache 2.0 | \nMIT | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \n
\n\nUteke vs Mnemosyne: Both are local-first with semantic + FTS5 search. Mnemosyne is the closest competitor (~1.5K stars, Python). Uteke wins on single binary (no Python runtime), rooms, time-travel queries, and zero runtime dependencies.
\n
\n\nUteke vs Engram: Both are single-binary, offline, no-API-key tools. But Engram is FTS5-only (keyword search). Uteke adds vector semantic search + RRF fusion + rooms + time-travel + graph relationships + smart decay + document engine + batch import. Same simplicity thesis, 10ร the features.
\n
\n\nUteke vs AgentMemory/Mem0/Letta/Zep: Those are powerful โ but all require cloud LLM API keys and Docker infrastructure. Your data goes to OpenAI/Anthropic. Uteke runs fully offline with local ONNX embeddings. No Docker, no Python, no API keys.
\n
\n \n
๐ค Building AI agents? Give them persistent memory without cloud dependencies. Your agent remembers user preferences, past decisions, and context โ across sessions, fully offline.
\n๐ฅ Working in a team? Use Rooms to share knowledge. Meeting notes, project decisions, architecture choices โ searchable by everyone, attributed by author.
\n๐ Building for privacy-sensitive domains? Healthcare, finance, legal โ data stays on your machine. No API calls, no telemetry, no cloud. Local embeddings (ONNX, 768d).
\nโจ๏ธ Power user who lives in the terminal? Uteke is your personal knowledge graph. Remember anything, recall by meaning, link related thoughts. All from the command line.
\n| Feature | \nWhat it does | \n
|---|---|
| ๐ง Hybrid Search | \nVector similarity + FTS5 full-text search, merged by Reciprocal Rank Fusion (RRF). Finds by meaning AND exact keywords. | \n
| ๐ Rooms | \nGroup memories by context (meetings, projects, clients) with author attribution. | \n
| โณ Time-travel | \nRecall memories as they existed at any point in time. uteke recall \"deploy\" --at 2025-01-15 | \n
| ๐ท๏ธ Rich Metadata | \nTags, entities, categories, key:value pairs on every memory. | \n
| ๐งฉ Memory Types | \nTyped categories (fact, procedure, decision, etc.) with auto-inference. | \n
| โ๏ธ Partial Updates | \nUpdate content, tags, metadata, importance, or type without full rewrite. | \n
| ๐ Citations | \nSource attribution on every memory (URL, file, user, import batch). | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ Relationship Graph | \nLink memories with typed edges (supersedes, contradicts, references). Auto-backlinks. | \n
| ๐ Cross-Entity Linking | \nBidirectional memoryโdocument references via [[doc-slug]] wikilinks. | \n
| ๐ค Cosine Auto-Linking | \nAutomatically creates similar_to edges between related memories. | \n
| ๐ Smart Decay | \nComposite importance scoring. Pin what matters, let stale memories fade. | \n
| ๐ Salience + Recency | \nDual-axis recall boost by memory type and age. | \n
| ๐ Orphan Detection | \nFind disconnected, low-importance memories for cleanup. | \n
| ๐ Dream Cycle | \nOne-command maintenance: lint โ backlinks โ dedup โ orphans. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ MCP Server | \nJSON-RPC over stdio + Streamable HTTP. Works with Claude Code, Cursor, Hermes. | \n
| ๐ฅ๏ธ Server Mode | \nPersistent daemon โ eliminates cold-start embedding load on every call. | \n
| ๐ Batch Import | \nImport entire directories with auto-strategy routing (document vs. memory extraction). | \n
| ๐ Document Engine | \nWiki/knowledge base with uteke doc create/get/list and auto-chunking. | \n
| ๐ฅ Import/Export | \nJSONL-based backup and restore. | \n
| ๐ View-Only API Keys | \nRead-only tokens for safe GET-only access to the server. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ฆ Single Binary | \nZero dependencies. No Docker needed, no database server, no Python, no API keys. | \n
| ๐ Fully Offline | \nLocal ONNX embeddings (EmbeddingGemma Q4, 768d). No telemetry, no cloud. | \n
| โก Recall Cache | \nLRU cache eliminates redundant embedding for repeated queries. | \n
| ๐ฅ Tiered Memory | \nHot/Warm/Cold tracking with auto-cleanup of stale memories. | \n
| ๐ Embed Fallback | \nGracefully degrades to no-op embedder if local model fails (never crashes). | \n
| ๐ฅ Multi-Agent Namespaces | \nFully isolated memory per agent, zero overhead. | \n
| ๐ Benchmarks | \nBuilt-in uteke bench for perf testing. See results. | \n
// .mcp.json (Claude Code, Cursor)\n{ \"mcpServers\": { \"uteke\": { \"command\": \"uteke-mcp\" } } }\nFor Claude Desktop, Hermes, and HTTP transport, see MCP docs.
\n๐ Full documentation ยท CLI reference ยท Configuration
\ngraph LR\n Input[User Query] --> Embed[Local ONNX Embedder<br/>768d, EmbeddingGemma Q4]\n Embed --> HNSW[HNSW Vector Index<br/>usearch]\n Embed --> FTS5[FTS5 Full-Text<br/>SQLite]\n HNSW --> RRF[Reciprocal Rank Fusion<br/>k=60]\n FTS5 --> RRF\n RRF --> Results[Ranked Results]\n\n style Input fill:#4a9eff,color:#fff\n style Results fill:#4aff9e,color:#000\n style RRF fill:#ff9e4a,color:#fff\nHow hybrid search works:
\nEverything runs in-process. No network. No cloud. No server required (unless you want server mode).
\n\n \n
Mem0 and Letta are great โ but they require cloud API keys (OpenAI/LLM) and external infrastructure (Docker, Postgres, Qdrant). Your data gets sent to a cloud LLM provider. Uteke is a single binary with zero API keys. All embeddings run locally via ONNX. Your data never leaves your machine. See comparison table.
\n\nHow is Uteke different from AgentMemory?AgentMemory (25K stars) is a TypeScript/Node.js platform with 53 MCP tools and 12 auto-hooks. It's feature-rich but requires Docker + the iii-engine + LLM API keys. Uteke is Rust, zero dependencies, and works fully offline. If you want maximum integrations and don't mind cloud dependency โ AgentMemory. If you want privacy, speed, and zero setup โ Uteke.
\n\nHow is Uteke different from Engram?Engram (2.4K stars, Go) shares our philosophy: single binary, zero deps, MCP server, local-first. The key difference is search: Engram uses FTS5 only (keyword matching). Uteke uses hybrid search (HNSW vector similarity + FTS5 + Reciprocal Rank Fusion) โ meaning you can search by meaning, not just exact words. Uteke also adds rooms, time-travel, graph relationships, smart decay, document engine, and batch import.
\n\nWhat can Uteke remember?Anything text-based: decisions, meeting notes, code snippets, project context, personal notes, agent state. You can tag, categorize, and link memories. The --batch-dir flag lets you import entire document directories.
Yes. The embedding model (EmbeddingGemma Q4, 768d) downloads once (~188MB) on first run. After that, zero network calls. No telemetry. If the local model fails, Uteke degrades gracefully to a no-op embedder โ it never crashes and never calls a cloud API.
\n\nHow fast is recall?~45ms as a library (measured at 100โ10K memories). No network round-trip because everything is local. The LRU recall cache eliminates redundant embedding computation for repeated queries.
\n\nCan I use Uteke with my existing AI tools?Yes. Uteke ships with an MCP server that works with Claude Code, Cursor, and Hermes. You can also use the HTTP API directly in any language. See MCP setup โ
\n\nIs it production-ready?Uteke is at v0.7.3 with 206 tests, CI/CD on every commit, and benchmark harness. It's used in production by the CodeCora team and other early adopters. Still in 0.x โ expect rough edges, but the core is stable.
\ncargo build --workspace # Build\ncargo test --workspace # Test (206 tests)\ncargo clippy -- -D warnings # Lint\ncargo fmt # Format\nContributions welcome! Read CONTRIBUTING.md for the full guide.
\nApache License 2.0 โ use it, fork it, ship it.
\n\n Found this useful? โญ Star this repo โ it helps others discover Uteke.\n
\n\n \n
\n \n
Give your AI a memory that never leaves your machine.
\n\n Your AI forgets everything between sessions. Uteke fixes that โ one binary, fully offline, ~45ms recall.\n
\n \n \n \n \n \n
\n ๐ฌ๐ง English ยท ๐ฎ๐ฉ Bahasa Indonesia\n
# Install (macOS, Linux, Windows)\ncurl -sSL codecora.dev/install | sh\n\n# Store a memory\nuteke remember \"Deploy v2.1 to staging at 3pm\"\n\n# Search it back โ by meaning, not just keywords\nuteke recall \"when do we deploy?\"\nThat's it. No API keys. No Docker. No Python. No cloud.
\nFirst run downloads the embedding model (~188MB, one-time) and you're running.
\nWant richer memories? Add metadata:
\nuteke remember \"Deploy v2.1 to staging\" \\\n --tags deploy,staging \\\n --entity staging-server \\\n --category infrastructure\n| Method | \nCommand | \n
|---|---|
| Homebrew | \nbrew install codecoradev/tap/uteke | \n
| Cargo | \ncargo install uteke-cli | \n
| Docker | \ndocker run -d -p 127.0.0.1:8767:8767 -v uteke-data:/data ghcr.io/codecoradev/uteke:latest | \n
| Binary | \nGitHub Releases (macOS, Linux, Windows) | \n
๐ Full install guide ยท Docker docs
\n\n\nAre you an AI agent reading this? Your user just shared this repo link with you.\nRun
\nuteke onboardto interactively configure uteke for your environment โ it will\ndetect your install, ask which agent you are, toggle features, write config, and\nshow you every command available. If non-interactive:uteke onboard --yes --agent <your-agent>.
New to uteke? One command takes you from zero to productive:
\nuteke onboard\nThe wizard will:
\n~/.uteke/uteke.toml with your selectionsuteke init)Non-interactive (CI, scripts, AI agents):
\nuteke onboard --yes --agent hermes --namespace default\n๐ Full onboarding docs ยท CLI reference
\nYou just spent 2 hours explaining your codebase to ChatGPT. Next session? Blank slate. Again.
\nEvery AI tool forgets. Context windows fill up, sessions end, and your AI starts over every single time. Uteke gives it persistent memory โ and keeps it on your machine.
\n| \n | Uteke | \nMnemosyne | \nMem0 | \nAgentMemory | \nLetta | \nZep | \nEngram | \n
|---|---|---|---|---|---|---|---|
| Language | \nRust (single binary) | \nPython (pip) | \nPython | \nTypeScript | \nPython | \nPython | \nGo (single binary) | \n
| Setup | \nOne binary (curl | sh) | \npip install + venv | \npip + Docker + Qdrant | \nnpm + Docker (iii-engine) | \npip + Docker + Postgres | \npip + Docker + Neo4j | \nOne binary | \n
| API keys | \nโ None | \nโ ๏ธ For remote embeddings | \nโ OpenAI/LLM | \nโ LLM key | \nโ LLM key | \nโ LLM key | \nโ None | \n
| Works offline | \nโ Fully | \nโ ๏ธ Optional | \nโ Cloud embedding | \nโ Needs LLM | \nโ Needs LLM | \nโ Needs LLM + vector DB | \nโ Fully | \n
| Search | \nHybrid (Vector + FTS5 + RRF) | \nsqlite-vec + FTS5 | \nVector + Graph | \nVector + Graph | \nVector | \nTemporal Graph | \nFTS5 only | \n
| Recall speed | \n~45ms | \n~50ms+ | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \n~Fast (local) | \n
| Multi-agent | \nโ Rooms (built-in collaboration) | \nโ ๏ธ Shared API | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Time-travel | \nโ Native point-in-time | \nโ ๏ธ Temporal triples | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| MCP server | \nโ JSON-RPC + HTTP | \nโ stdio + SSE | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Your data | \nโ Never leaves machine | \nโ Local-first | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ Local | \n
| License | \nApache 2.0 | \nMIT | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \n
\n\nUteke vs Mnemosyne: Both are local-first with semantic + FTS5 search. Mnemosyne is the closest competitor (~1.5K stars, Python). Uteke wins on single binary (no Python runtime), rooms, time-travel queries, and zero runtime dependencies.
\n
\n\nUteke vs Engram: Both are single-binary, offline, no-API-key tools. But Engram is FTS5-only (keyword search). Uteke adds vector semantic search + RRF fusion + rooms + time-travel + graph relationships + smart decay + document engine + batch import. Same simplicity thesis, 10ร the features.
\n
\n\nUteke vs AgentMemory/Mem0/Letta/Zep: Those are powerful โ but all require cloud LLM API keys and Docker infrastructure. Your data goes to OpenAI/Anthropic. Uteke runs fully offline with local ONNX embeddings. No Docker, no Python, no API keys.
\n
\n \n
๐ค Building AI agents? Give them persistent memory without cloud dependencies. Your agent remembers user preferences, past decisions, and context โ across sessions, fully offline.
\n๐ฅ Working in a team? Use Rooms to share knowledge. Meeting notes, project decisions, architecture choices โ searchable by everyone, attributed by author.
\n๐ Building for privacy-sensitive domains? Healthcare, finance, legal โ data stays on your machine. No API calls, no telemetry, no cloud. Local embeddings (ONNX, 768d).
\nโจ๏ธ Power user who lives in the terminal? Uteke is your personal knowledge graph. Remember anything, recall by meaning, link related thoughts. All from the command line.
\n| Feature | \nWhat it does | \n
|---|---|
| ๐ง Hybrid Search | \nVector similarity + FTS5 full-text search, merged by Reciprocal Rank Fusion (RRF). Finds by meaning AND exact keywords. | \n
| ๐ Rooms | \nGroup memories by context (meetings, projects, clients) with author attribution. | \n
| โณ Time-travel | \nRecall memories as they existed at any point in time. uteke recall \"deploy\" --at 2025-01-15 | \n
| ๐ท๏ธ Rich Metadata | \nTags, entities, categories, key:value pairs on every memory. | \n
| ๐งฉ Memory Types | \nTyped categories (fact, procedure, decision, etc.) with auto-inference. | \n
| โ๏ธ Partial Updates | \nUpdate content, tags, metadata, importance, or type without full rewrite. | \n
| ๐ Citations | \nSource attribution on every memory (URL, file, user, import batch). | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ Relationship Graph | \nLink memories with typed edges (supersedes, contradicts, references). Auto-backlinks. | \n
| ๐ Cross-Entity Linking | \nBidirectional memoryโdocument references via [[doc-slug]] wikilinks. | \n
| ๐ค Cosine Auto-Linking | \nAutomatically creates similar_to edges between related memories. | \n
| ๐ Smart Decay | \nComposite importance scoring. Pin what matters, let stale memories fade. | \n
| ๐ Salience + Recency | \nDual-axis recall boost by memory type and age. | \n
| ๐ Orphan Detection | \nFind disconnected, low-importance memories for cleanup. | \n
| ๐ Dream Cycle | \nOne-command maintenance: lint โ backlinks โ dedup โ orphans. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ MCP Server | \nJSON-RPC over stdio + Streamable HTTP. Works with Claude Code, Cursor, Hermes. | \n
| ๐ฅ๏ธ Server Mode | \nPersistent daemon โ eliminates cold-start embedding load on every call. | \n
| ๐ Batch Import | \nImport entire directories with auto-strategy routing (document vs. memory extraction). | \n
| ๐ Document Engine | \nWiki/knowledge base with uteke doc create/get/list and auto-chunking. | \n
| ๐ฅ Import/Export | \nJSONL-based backup and restore. | \n
| ๐ View-Only API Keys | \nRead-only tokens for safe GET-only access to the server. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ฆ Single Binary | \nZero dependencies. No Docker needed, no database server, no Python, no API keys. | \n
| ๐ Fully Offline | \nLocal ONNX embeddings (EmbeddingGemma Q4, 768d). No telemetry, no cloud. | \n
| โก Recall Cache | \nLRU cache eliminates redundant embedding for repeated queries. | \n
| ๐ฅ Tiered Memory | \nHot/Warm/Cold tracking with auto-cleanup of stale memories. | \n
| ๐ Embed Fallback | \nGracefully degrades to no-op embedder if local model fails (never crashes). | \n
| ๐ฅ Multi-Agent Namespaces | \nFully isolated memory per agent, zero overhead. | \n
| ๐ Benchmarks | \nBuilt-in uteke bench for perf testing. See results. | \n
// .mcp.json (Claude Code, Cursor)\n{ \"mcpServers\": { \"uteke\": { \"command\": \"uteke-mcp\" } } }\nFor Claude Desktop, Hermes, and HTTP transport, see MCP docs.
\n๐ Full documentation ยท CLI reference ยท Configuration
\ngraph LR\n Input[User Query] --> Embed[Local ONNX Embedder<br/>768d, EmbeddingGemma Q4]\n Embed --> HNSW[HNSW Vector Index<br/>usearch]\n Embed --> FTS5[FTS5 Full-Text<br/>SQLite]\n HNSW --> RRF[Reciprocal Rank Fusion<br/>k=60]\n FTS5 --> RRF\n RRF --> Results[Ranked Results]\n\n style Input fill:#4a9eff,color:#fff\n style Results fill:#4aff9e,color:#000\n style RRF fill:#ff9e4a,color:#fff\nHow hybrid search works:
\nEverything runs in-process. No network. No cloud. No server required (unless you want server mode).
\n\n \n
Mem0 and Letta are great โ but they require cloud API keys (OpenAI/LLM) and external infrastructure (Docker, Postgres, Qdrant). Your data gets sent to a cloud LLM provider. Uteke is a single binary with zero API keys. All embeddings run locally via ONNX. Your data never leaves your machine. See comparison table.
\n\nHow is Uteke different from AgentMemory?AgentMemory (25K stars) is a TypeScript/Node.js platform with 53 MCP tools and 12 auto-hooks. It's feature-rich but requires Docker + the iii-engine + LLM API keys. Uteke is Rust, zero dependencies, and works fully offline. If you want maximum integrations and don't mind cloud dependency โ AgentMemory. If you want privacy, speed, and zero setup โ Uteke.
\n\nHow is Uteke different from Engram?Engram (2.4K stars, Go) shares our philosophy: single binary, zero deps, MCP server, local-first. The key difference is search: Engram uses FTS5 only (keyword matching). Uteke uses hybrid search (HNSW vector similarity + FTS5 + Reciprocal Rank Fusion) โ meaning you can search by meaning, not just exact words. Uteke also adds rooms, time-travel, graph relationships, smart decay, document engine, and batch import.
\n\nWhat can Uteke remember?Anything text-based: decisions, meeting notes, code snippets, project context, personal notes, agent state. You can tag, categorize, and link memories. The --batch-dir flag lets you import entire document directories.
Yes. The embedding model (EmbeddingGemma Q4, 768d) downloads once (~188MB) on first run. After that, zero network calls. No telemetry. If the local model fails, Uteke degrades gracefully to a no-op embedder โ it never crashes and never calls a cloud API.
\n\nHow fast is recall?~45ms as a library (measured at 100โ10K memories). No network round-trip because everything is local. The LRU recall cache eliminates redundant embedding computation for repeated queries.
\n\nCan I use Uteke with my existing AI tools?Yes. Uteke ships with an MCP server that works with Claude Code, Cursor, and Hermes. You can also use the HTTP API directly in any language. See MCP setup โ
\n\nIs it production-ready?Uteke is at v0.7.3 with 206 tests, CI/CD on every commit, and benchmark harness. It's used in production by the CodeCora team and other early adopters. Still in 0.x โ expect rough edges, but the core is stable.
\ncargo build --workspace # Build\ncargo test --workspace # Test (206 tests)\ncargo clippy -- -D warnings # Lint\ncargo fmt # Format\nContributions welcome! Read CONTRIBUTING.md for the full guide.
\nApache License 2.0 โ use it, fork it, ship it.
\n\n Found this useful? โญ Star this repo โ it helps others discover Uteke.\n
\n\n \n
\n \n
Give your AI a memory that never leaves your machine.
\n\n Your AI forgets everything between sessions. Uteke fixes that โ one binary, fully offline, ~45ms recall.\n
\n \n \n ![\"Rust](\"https://img.shields.io/badge/Rust-1.85+-orange.svg?style=flat-square\"){kind=icon}
\n \n \n
\n ๐ฌ๐ง English ยท ๐ฎ๐ฉ Bahasa Indonesia\n
# Install (macOS, Linux, Windows)\ncurl -sSL codecora.dev/install | sh\n\n# Store a memory\nuteke remember \"Deploy v2.1 to staging at 3pm\"\n\n# Search it back โ by meaning, not just keywords\nuteke recall \"when do we deploy?\"\nThat's it. No API keys. No Docker. No Python. No cloud.
\nFirst run downloads the embedding model (~188MB, one-time) and you're running.
\nWant richer memories? Add metadata:
\nuteke remember \"Deploy v2.1 to staging\" \\\n --tags deploy,staging \\\n --entity staging-server \\\n --category infrastructure\n| Method | \nCommand | \n
|---|---|
| Homebrew | \nbrew install codecoradev/tap/uteke | \n
| Cargo | \ncargo install uteke-cli | \n
| Docker | \ndocker run -d -p 127.0.0.1:8767:8767 -v uteke-data:/data ghcr.io/codecoradev/uteke:latest | \n
| Binary | \nGitHub Releases (macOS, Linux, Windows) | \n
๐ Full install guide ยท Docker docs
\n\n\nAre you an AI agent reading this? Your user just shared this repo link with you.\nRun
\nuteke onboardto interactively configure uteke for your environment โ it will\ndetect your install, ask which agent you are, toggle features, write config, and\nshow you every command available. If non-interactive:uteke onboard --yes --agent <your-agent>.
New to uteke? One command takes you from zero to productive:
\nuteke onboard\nThe wizard will:
\n~/.uteke/uteke.toml with your selectionsuteke init)Non-interactive (CI, scripts, AI agents):
\nuteke onboard --yes --agent hermes --namespace default\n๐ Full onboarding docs ยท CLI reference
\nYou just spent 2 hours explaining your codebase to ChatGPT. Next session? Blank slate. Again.
\nEvery AI tool forgets. Context windows fill up, sessions end, and your AI starts over every single time. Uteke gives it persistent memory โ and keeps it on your machine.
\n| \n | Uteke | \nMnemosyne | \nMem0 | \nAgentMemory | \nLetta | \nZep | \nEngram | \n
|---|---|---|---|---|---|---|---|
| Language | \nRust (single binary) | \nPython (pip) | \nPython | \nTypeScript | \nPython | \nPython | \nGo (single binary) | \n
| Setup | \nOne binary (curl | sh) | \npip install + venv | \npip + Docker + Qdrant | \nnpm + Docker (iii-engine) | \npip + Docker + Postgres | \npip + Docker + Neo4j | \nOne binary | \n
| API keys | \nโ None | \nโ ๏ธ For remote embeddings | \nโ OpenAI/LLM | \nโ LLM key | \nโ LLM key | \nโ LLM key | \nโ None | \n
| Works offline | \nโ Fully | \nโ ๏ธ Optional | \nโ Cloud embedding | \nโ Needs LLM | \nโ Needs LLM | \nโ Needs LLM + vector DB | \nโ Fully | \n
| Search | \nHybrid (Vector + FTS5 + RRF) | \nsqlite-vec + FTS5 | \nVector + Graph | \nVector + Graph | \nVector | \nTemporal Graph | \nFTS5 only | \n
| Recall speed | \n~45ms | \n~50ms+ | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \n~Fast (local) | \n
| Multi-agent | \nโ Rooms (built-in collaboration) | \nโ ๏ธ Shared API | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Time-travel | \nโ Native point-in-time | \nโ ๏ธ Temporal triples | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| MCP server | \nโ JSON-RPC + HTTP | \nโ stdio + SSE | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Your data | \nโ Never leaves machine | \nโ Local-first | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ Local | \n
| License | \nApache 2.0 | \nMIT | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \n
\n\nUteke vs Mnemosyne: Both are local-first with semantic + FTS5 search. Mnemosyne is the closest competitor (~1.5K stars, Python). Uteke wins on single binary (no Python runtime), rooms, time-travel queries, and zero runtime dependencies.
\n
\n\nUteke vs Engram: Both are single-binary, offline, no-API-key tools. But Engram is FTS5-only (keyword search). Uteke adds vector semantic search + RRF fusion + rooms + time-travel + graph relationships + smart decay + document engine + batch import. Same simplicity thesis, 10ร the features.
\n
\n\nUteke vs AgentMemory/Mem0/Letta/Zep: Those are powerful โ but all require cloud LLM API keys and Docker infrastructure. Your data goes to OpenAI/Anthropic. Uteke runs fully offline with local ONNX embeddings. No Docker, no Python, no API keys.
\n
\n \n
๐ค Building AI agents? Give them persistent memory without cloud dependencies. Your agent remembers user preferences, past decisions, and context โ across sessions, fully offline.
\n๐ฅ Working in a team? Use Rooms to share knowledge. Meeting notes, project decisions, architecture choices โ searchable by everyone, attributed by author.
\n๐ Building for privacy-sensitive domains? Healthcare, finance, legal โ data stays on your machine. No API calls, no telemetry, no cloud. Local embeddings (ONNX, 768d).
\nโจ๏ธ Power user who lives in the terminal? Uteke is your personal knowledge graph. Remember anything, recall by meaning, link related thoughts. All from the command line.
\n| Feature | \nWhat it does | \n
|---|---|
| ๐ง Hybrid Search | \nVector similarity + FTS5 full-text search, merged by Reciprocal Rank Fusion (RRF). Finds by meaning AND exact keywords. | \n
| ๐ Rooms | \nGroup memories by context (meetings, projects, clients) with author attribution. | \n
| โณ Time-travel | \nRecall memories as they existed at any point in time. uteke recall \"deploy\" --at 2025-01-15 | \n
| ๐ท๏ธ Rich Metadata | \nTags, entities, categories, key:value pairs on every memory. | \n
| ๐งฉ Memory Types | \nTyped categories (fact, procedure, decision, etc.) with auto-inference. | \n
| โ๏ธ Partial Updates | \nUpdate content, tags, metadata, importance, or type without full rewrite. | \n
| ๐ Citations | \nSource attribution on every memory (URL, file, user, import batch). | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ Relationship Graph | \nLink memories with typed edges (supersedes, contradicts, references). Auto-backlinks. | \n
| ๐ Cross-Entity Linking | \nBidirectional memoryโdocument references via [[doc-slug]] wikilinks. | \n
| ๐ค Cosine Auto-Linking | \nAutomatically creates similar_to edges between related memories. | \n
| ๐ Smart Decay | \nComposite importance scoring. Pin what matters, let stale memories fade. | \n
| ๐ Salience + Recency | \nDual-axis recall boost by memory type and age. | \n
| ๐ Orphan Detection | \nFind disconnected, low-importance memories for cleanup. | \n
| ๐ Dream Cycle | \nOne-command maintenance: lint โ backlinks โ dedup โ orphans. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ MCP Server | \nJSON-RPC over stdio + Streamable HTTP. Works with Claude Code, Cursor, Hermes. | \n
| ๐ฅ๏ธ Server Mode | \nPersistent daemon โ eliminates cold-start embedding load on every call. | \n
| ๐ Batch Import | \nImport entire directories with auto-strategy routing (document vs. memory extraction). | \n
| ๐ Document Engine | \nWiki/knowledge base with uteke doc create/get/list and auto-chunking. | \n
| ๐ฅ Import/Export | \nJSONL-based backup and restore. | \n
| ๐ View-Only API Keys | \nRead-only tokens for safe GET-only access to the server. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ฆ Single Binary | \nZero dependencies. No Docker needed, no database server, no Python, no API keys. | \n
| ๐ Fully Offline | \nLocal ONNX embeddings (EmbeddingGemma Q4, 768d). No telemetry, no cloud. | \n
| โก Recall Cache | \nLRU cache eliminates redundant embedding for repeated queries. | \n
| ๐ฅ Tiered Memory | \nHot/Warm/Cold tracking with auto-cleanup of stale memories. | \n
| ๐ Embed Fallback | \nGracefully degrades to no-op embedder if local model fails (never crashes). | \n
| ๐ฅ Multi-Agent Namespaces | \nFully isolated memory per agent, zero overhead. | \n
| ๐ Benchmarks | \nBuilt-in uteke bench for perf testing. See results. | \n
// .mcp.json (Claude Code, Cursor)\n{ \"mcpServers\": { \"uteke\": { \"command\": \"uteke-mcp\" } } }\nFor Claude Desktop, Hermes, and HTTP transport, see MCP docs.
\n๐ Full documentation ยท CLI reference ยท Configuration
\ngraph LR\n Input[User Query] --> Embed[Local ONNX Embedder<br/>768d, EmbeddingGemma Q4]\n Embed --> HNSW[HNSW Vector Index<br/>usearch]\n Embed --> FTS5[FTS5 Full-Text<br/>SQLite]\n HNSW --> RRF[Reciprocal Rank Fusion<br/>k=60]\n FTS5 --> RRF\n RRF --> Results[Ranked Results]\n\n style Input fill:#4a9eff,color:#fff\n style Results fill:#4aff9e,color:#000\n style RRF fill:#ff9e4a,color:#fff\nHow hybrid search works:
\nEverything runs in-process. No network. No cloud. No server required (unless you want server mode).
\n\n \n
Mem0 and Letta are great โ but they require cloud API keys (OpenAI/LLM) and external infrastructure (Docker, Postgres, Qdrant). Your data gets sent to a cloud LLM provider. Uteke is a single binary with zero API keys. All embeddings run locally via ONNX. Your data never leaves your machine. See comparison table.
\n\nHow is Uteke different from AgentMemory?AgentMemory (25K stars) is a TypeScript/Node.js platform with 53 MCP tools and 12 auto-hooks. It's feature-rich but requires Docker + the iii-engine + LLM API keys. Uteke is Rust, zero dependencies, and works fully offline. If you want maximum integrations and don't mind cloud dependency โ AgentMemory. If you want privacy, speed, and zero setup โ Uteke.
\n\nHow is Uteke different from Engram?Engram (2.4K stars, Go) shares our philosophy: single binary, zero deps, MCP server, local-first. The key difference is search: Engram uses FTS5 only (keyword matching). Uteke uses hybrid search (HNSW vector similarity + FTS5 + Reciprocal Rank Fusion) โ meaning you can search by meaning, not just exact words. Uteke also adds rooms, time-travel, graph relationships, smart decay, document engine, and batch import.
\n\nWhat can Uteke remember?Anything text-based: decisions, meeting notes, code snippets, project context, personal notes, agent state. You can tag, categorize, and link memories. The --batch-dir flag lets you import entire document directories.
Yes. The embedding model (EmbeddingGemma Q4, 768d) downloads once (~188MB) on first run. After that, zero network calls. No telemetry. If the local model fails, Uteke degrades gracefully to a no-op embedder โ it never crashes and never calls a cloud API.
\n\nHow fast is recall?~45ms as a library (measured at 100โ10K memories). No network round-trip because everything is local. The LRU recall cache eliminates redundant embedding computation for repeated queries.
\n\nCan I use Uteke with my existing AI tools?Yes. Uteke ships with an MCP server that works with Claude Code, Cursor, and Hermes. You can also use the HTTP API directly in any language. See MCP setup โ
\n\nIs it production-ready?Uteke is at v0.7.3 with 206 tests, CI/CD on every commit, and benchmark harness. It's used in production by the CodeCora team and other early adopters. Still in 0.x โ expect rough edges, but the core is stable.
\ncargo build --workspace # Build\ncargo test --workspace # Test (206 tests)\ncargo clippy -- -D warnings # Lint\ncargo fmt # Format\nContributions welcome! Read CONTRIBUTING.md for the full guide.
\nApache License 2.0 โ use it, fork it, ship it.
\n\n Found this useful? โญ Star this repo โ it helps others discover Uteke.\n
\n\n \n
\n Noise-canceling context and long-term memory for your AI agent. Stop paying Claude to read 10,000 lines of terminal noise like a headphone for AI agent\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Problem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
A modern, project-aware terminal manager built with Tauri
\nTermul treats workspaces as first-class citizens, allowing you to organize terminals by project with persistent sessions, snapshots, and a clean tabbed interface.
\n![\"GitHub](\"https://img.shields.io/github/stars/gnoviawan/termul?style=social\")
![\"GitHub](\"https://img.shields.io/github/issues/gnoviawan/termul\")
![\"License\"](\"https://img.shields.io/github/license/gnoviawan/termul\")
![\"Platform\"](\"https://img.shields.io/badge/Platform-Windows%20%7C%20macOS%20%7C%20Linux-blue\"){kind=icon}
![\"Tauri\"](\"https://img.shields.io/badge/Tauri-2.0-FFC131?logo=tauri&logoColor=white\"){kind=icon}
![\"React\"](\"https://img.shields.io/badge/React-18-61DAFB?logo=react&logoColor=white\"){kind=icon}
Getting Started ยท Features ยท Documentation ยท Contributing ยท Report Bug ยท Request Feature
\n| Feature | \nDescription | \n
|---|---|
| Project-Based Workspaces | \nOrganize terminals by project with dedicated workspace directories, separate state, and per-project configuration | \n
| Pane-Based Split Layout | \nSplit your workspace into resizable panes and arrange terminals, editors, and browser tabs side by side | \n
| Tabbed Interface | \nWindows Terminal-style tab bar with drag-and-drop reordering, rename, and context menu | \n
| Multiple Shell Support | \nAuto-detects PowerShell, CMD, Git Bash, WSL, fish, zsh, and more; switch shells per tab | \n
| Feature | \nDescription | \n
|---|---|
| Code Editor | \nBuilt-in code editor with syntax highlighting, file buffers, dirty-state tracking, and save/reload | \n
| Markdown Editor | \nRich markdown editing powered by BlockNote with live preview, table of contents, and heading navigation | \n
| Mermaid Diagrams | \nRender Mermaid diagrams inline within your markdown documents | \n
| File Explorer | \nFull file tree with create, rename, delete, clipboard operations, drag-and-drop, and context menus | \n
| File Watching | \nLive file watching for real-time updates as files change on disk | \n
| Feature | \nDescription | \n
|---|---|
| Embedded Browser Tabs | \nBrowse the web directly inside your workspace using child webview tabs โ no app switching | \n
| Annotation Workflow | \nCapture browser states, annotate with severity and intent labels, review, and export | \n
| Annotation Export | \nPackage annotations with metadata into structured export formats | \n
| Feature | \nDescription | \n
|---|---|
| Command Palette | \nGlobal command launcher (Ctrl+K / Ctrl+Shift+P) for project switching, workspace actions, and more | \n
| Command History | \nPer-project and aggregate command history viewer with search | \n
| Keyboard Shortcuts | \nFully customizable shortcut bindings for every action | \n
| Git Integration | \nStatus bar shows current branch, working directory, git status, and exit code | \n
| Custom Title Bar | \nDesktop-native title bar with window controls, sidebar toggles, and settings navigation | \n
| Feature | \nDescription | \n
|---|---|
| Auto-Updater | \nBuilt-in update infrastructure with signed artifacts โ get notified and update without leaving the app | \n
| State Management | \nZustand-powered reactive stores for projects, terminals, workspace layout, editor buffers, browser sessions, and settings | \n
| Configurable Settings | \nTerminal and UI preferences, color picker, theme customization, and shell configuration | \n
| Cross-Platform | \nWorks on Windows, macOS, and Linux with native platform packaging | \n
| Error Boundaries | \nGraceful error handling with runtime error boundaries and user-friendly fallback UI | \n
| Domain | \nKey Components | \nZustand Store | \n
|---|---|---|
| Workspace | \nWorkspaceLayout, PaneRenderer, PaneContent, WorkspaceTabBar | \nworkspace-store | \n
| Terminal | \nConnectedTerminal, XTerminal, TerminalSearchBar, ActivityIndicator | \nterminal-store | \n
| Editor | \nEditorPanel, CodeEditor, MarkdownEditor, EditorToolbar, MermaidBlock | \neditor-store | \n
| Browser | \nBrowserPanel, BrowserControls, AnnotationPanel, AnnotationExportModal | \nbrowser-session-store, annotation-store | \n
| File Explorer | \nFileExplorer, FileTreeNode, FileTreeContextMenu | \nโ | \n
| Snapshots | \nCreateSnapshotModal, RestoreSnapshotModal, DeleteSnapshotModal | \nsnapshot-store | \n
| Projects | \nProjectSidebar, NewProjectModal | \nproject-store | \n
| Settings | \nShortcutRecorder, ColorPickerPopover, ContextBarSettingsPopover | \napp-settings-store, context-bar-settings-store | \n
| Updates | \nUpdateAvailableToast, UpdateReadyModal | \nupdater-store | \n
| Shared | \nCommandPalette, ContextMenu, ConfirmDialog, ShellSelector, ErrorBoundary | \nโ | \n
![\"Termul](\"https://raw.githubusercontent.com/gnoviawan/termul/dev/img/termul.png\")
brew tap gnoviawan/termul\nbrew install --cask termul\ncurl -fsSL https://raw.githubusercontent.com/gnoviawan/termul/main/scripts/install.sh | bash\nWindows users should install the .exe or .msi from GitHub Releases. Manual DMG downloads in a browser may still hit Gatekeeper, so macOS users should prefer Homebrew or curl.
| Dependency | \nVersion | \nNotes | \n
|---|---|---|
| Bun | \n1.3+ | \nJavaScript runtime and package manager | \n
| Rust | \nLatest stable | \nRequired for Tauri builds | \n
xcode-select --install\nsudo apt update\nsudo apt install libwebkit2gtk-4.1-dev \\\n build-essential curl wget file \\\n libxdo-dev libssl-dev \\\n libayatana-appindicator3-dev \\\n librsvg2-dev patchelf\nsudo dnf install webkit2gtk4.1-devel \\\n gcc gcc-c++ libopenssl-devel \\\n appindicator-devel librsvg2-devel \\\n patchelf\ncurl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh\nrustc --version && cargo --version\n# Clone the repository\ngit clone https://github.com/gnoviawan/termul.git\ncd termul\n\n# Install dependencies\nbun install\n\n# Launch in development mode\nbun run dev\nThis repository also includes a standalone Vite landing page under landing/.
# Install landing page dependencies (from landing/)\ncd landing && bun install\n\n# Start the landing page dev server\nbun run landing:dev\n\n# Lint the landing page\nbun run landing:lint\n\n# Build the landing page for production\nbun run landing:build\n# Build for your current platform\nbun run build\n\n# Platform-specific builds\nbun run build:tauri:win # Windows (x64)\nbun run build:tauri:mac-arm # macOS (Apple Silicon)\nbun run build:tauri:mac-x64 # macOS (Intel)\nbun run build:tauri:linux # Linux (x64)\n\n# Debug build (faster compilation, larger binary)\nbun run build:tauri:debug\nBuild output: src-tauri/target/release/bundle/
| Action | \nHow | \n
|---|---|
| New terminal | \nClick + next to tabs | \n
| Select specific shell | \nClick the dropdown arrow | \n
| Reorder tabs | \nDrag and drop | \n
| Rename tab | \nDouble-click the tab | \n
| Context menu | \nRight-click (rename, close, kill process) | \n
| Action | \nDefault Shortcut | \n
|---|---|
| New Terminal | \nCtrl+T | \n
| Next Tab | \nCtrl+PageDown | \n
| Previous Tab | \nCtrl+PageUp | \n
| Command Palette | \nCtrl+K / Ctrl+Shift+P | \n
\n\nShortcuts are customizable in Settings. On Tauri/WebView2, browser-reserved shortcuts such as
\nCtrl+Tabare not used as defaults because they are not reliably interceptable.
| Layer | \nTechnology | \n
|---|---|
| Desktop Runtime | \nTauri 2.0 | \n
| Backend | \nRust | \n
| UI Framework | \nReact 18 | \n
| Type System | \nTypeScript | \n
| Build Tool | \nVite | \n
| Styling | \nTailwind CSS + shadcn/ui | \n
| State Management | \nZustand | \n
| Terminal Emulation | \ntauri-pty + xterm.js | \n
| Animations | \nFramer Motion | \n
| Plugin | \nPurpose | \n
|---|---|
@tauri-apps/plugin-fs | \nFilesystem access | \n
@tauri-apps/plugin-store | \nConfiguration persistence | \n
@tauri-apps/plugin-os | \nOS information | \n
@tauri-apps/plugin-dialog | \nNative dialogs | \n
@tauri-apps/plugin-clipboard-manager | \nClipboard operations | \n
@tauri-apps/plugin-updater | \nAutomatic updates | \n
@tauri-apps/plugin-process | \nProcess management | \n
src/\nโโโ renderer/ # React frontend\nโ โโโ components/ # UI components\nโ โโโ hooks/ # Custom React hooks\nโ โโโ lib/ # Runtime adapters & desktop integration\nโ โโโ pages/ # Page components\nโ โโโ stores/ # Zustand stores\nโโโ shared/ # Shared types (main/renderer)\nsrc-tauri/ # Rust backend, config & bundling\ndocs/electron-old/ # Archived Electron docs & migration history\nThe renderer uses an adapter/service layer to keep desktop integrations isolated from UI code:
\nsrc/renderer/lib/\nโโโ tauri-*.ts # Tauri-native integrations\nโโโ *.ts # Runtime-safe facades & helpers\nโโโ __tests__/ # Regression & parity coverage\nbun run dev # Development mode with hot reload\nbun run test # Run tests\nbun run test:watch # Tests in watch mode\nbun run typecheck # Type checking\nbun run lint # Linting\nbun run tauri <command> # Direct Tauri CLI access\nssh-profiles.json.StrictHostKeyChecking=accept-new; do not override UserKnownHostsFile to /dev/null/NUL because that disables persistent host-key verification.ssh2 channel_direct_tcpip over the active SSH session; remote/reverse forwarding is not supported by the MVP command path yet.ssh2 intentionally stays on system OpenSSL for now. Enabling vendored-openssl forces a local OpenSSL source build that can fail in Windows/MSYS environments without a complete Perl module setup.Contributions are welcome! Please read the Contributing Guide for details on our code of conduct and the process for submitting pull requests.
\nThis project is licensed under the MIT License โ see the LICENSE file for details.
\nBuilt with โค๏ธ by gnoviawan
\n- Ruang kerja terbuka untuk ekosistem open source Indonesia โ dibikin + Ruang kerja terbuka untuk ekosistem open source Indonesia, dibikin bareng, dirawat bareng.
diff --git a/src/pages/events.astro b/src/pages/events.astro index b71ef2e..51dacc4 100644 --- a/src/pages/events.astro +++ b/src/pages/events.astro @@ -11,6 +11,6 @@ const formats = [ ---Halaman ini menampung format kegiatan yang ingin kami jalankan. Saat ada agenda, ia akan dicatat terbuka di sini dan di ruang diskusi.
Halaman ini menampung format kegiatan yang ingin kami jalankan. Saat ada agenda, ia akan dicatat terbuka di sini dan di ruang diskusi.
Saat kegiatan tersedia, detail waktu, format, dan catatan pasca-acara akan dicatat di halaman ini agar mudah diikuti ulang.
{body}
- Teknologi tumbuh kalau dirawat bersama โ + Teknologi tumbuh kalau dirawat bersama, dibuka, dibaca, dipakai ulang, dan diwariskan.
@@ -149,7 +149,7 @@ const references = [- Kalau bisa dipelajari, dipakai ulang, dan dirawat bersama โ itu sudah + Kalau bisa dipelajari, dipakai ulang, dan dirawat bersama, itu sudah open source.
IndopenSource berkomitmen menyediakan lingkungan yang ramah, aman, dan inklusif bagi semua orang. Kode ini berlaku untuk semua ruang komunitas โ GitHub, Discussions, event, dan interaksi lainnya.
+IndopenSource berkomitmen menyediakan lingkungan yang ramah, aman, dan inklusif bagi semua orang. Kode ini berlaku untuk semua ruang komunitas, GitHub, Discussions, event, dan interaksi lainnya.
Kami sengaja memulai dari daftar kosong agar setiap entri punya sumber, tautan aktif, dan diajukan secara terbukaโbukan hasil asumsi.
+Kami sengaja memulai dari daftar kosong agar setiap entri punya sumber, tautan aktif, dan diajukan secara terbuka, bukan hasil asumsi.
\n Noise-canceling context and long-term memory for your AI agent. Stop paying Claude to read 10,000 lines of terminal noise like a headphone for AI agent\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Problem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
\n Noise-canceling context and long-term memory for your AI agent โ lossy, but always reversible, and it never fabricates a result. Stop paying Claude to read 10,000 lines of terminal noise.\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Always reversible ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Every other compressor asks you to trust that what it cut didn't matter. OMNI doesn't ask โ it guarantees, and each guarantee is backed by code you can read:
\n| Guarantee | \nHow | \nProof | \n
|---|---|---|
| Get the original back, byte-for-byte | \neverything cut is archived in a local SQLite RewindStore (SHA-256 โ content); the agent gets a hash and calls omni_retrieve | \nHow it works | \n
| Never fabricates a result | \na distiller that parsed no signal returns the raw output, never a green no errors / passed string | \n#143 | \n
| Failures are never masked | \na command that exits non-zero passes through verbatim | \n#120 | \n
| Structured data is never touched | \nJSON / YAML / NDJSON / CSV pass through byte-for-byte | \npipeline::format | \n
| Numbers are measured, not aspirational | \n1,810 real traces replayed on the release binary โ and 63.6% of calls net zero, which we publish too | \nBenchmarks | \n
That is the one thing a bigger compression number can't buy: you can always recover the original, and it will never lie to your agent.
\nProblem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
A modern, project-aware terminal manager built with Tauri
\nTermul treats workspaces as first-class citizens, allowing you to organize terminals by project with persistent sessions, snapshots, and a clean tabbed interface.
\n
Getting Started ยท Features ยท Documentation ยท Contributing ยท Report Bug ยท Request Feature
\n| Feature | \nDescription | \n
|---|---|
| Project-Based Workspaces | \nOrganize terminals by project with dedicated workspace directories, separate state, and per-project configuration | \n
| Pane-Based Split Layout | \nSplit your workspace into resizable panes and arrange terminals, editors, and browser tabs side by side | \n
| Tabbed Interface | \nWindows Terminal-style tab bar with drag-and-drop reordering, rename, and context menu | \n
| Multiple Shell Support | \nAuto-detects PowerShell, CMD, Git Bash, WSL, fish, zsh, and more; switch shells per tab | \n
| Feature | \nDescription | \n
|---|---|
| Code Editor | \nBuilt-in code editor with syntax highlighting, file buffers, dirty-state tracking, and save/reload | \n
| Markdown Editor | \nRich markdown editing powered by BlockNote with live preview, table of contents, and heading navigation | \n
| Mermaid Diagrams | \nRender Mermaid diagrams inline within your markdown documents | \n
| File Explorer | \nFull file tree with create, rename, delete, clipboard operations, drag-and-drop, and context menus | \n
| File Watching | \nLive file watching for real-time updates as files change on disk | \n
| Feature | \nDescription | \n
|---|---|
| Embedded Browser Tabs | \nBrowse the web directly inside your workspace using child webview tabs โ no app switching | \n
| Annotation Workflow | \nCapture browser states, annotate with severity and intent labels, review, and export | \n
| Annotation Export | \nPackage annotations with metadata into structured export formats | \n
| Feature | \nDescription | \n
|---|---|
| Command Palette | \nGlobal command launcher (Ctrl+K / Ctrl+Shift+P) for project switching, workspace actions, and more | \n
| Command History | \nPer-project and aggregate command history viewer with search | \n
| Keyboard Shortcuts | \nFully customizable shortcut bindings for every action | \n
| Git Integration | \nStatus bar shows current branch, working directory, git status, and exit code | \n
| Custom Title Bar | \nDesktop-native title bar with window controls, sidebar toggles, and settings navigation | \n
| Feature | \nDescription | \n
|---|---|
| Auto-Updater | \nBuilt-in update infrastructure with signed artifacts โ get notified and update without leaving the app | \n
| State Management | \nZustand-powered reactive stores for projects, terminals, workspace layout, editor buffers, browser sessions, and settings | \n
| Configurable Settings | \nTerminal and UI preferences, color picker, theme customization, and shell configuration | \n
| Cross-Platform | \nWorks on Windows, macOS, and Linux with native platform packaging | \n
| Error Boundaries | \nGraceful error handling with runtime error boundaries and user-friendly fallback UI | \n
| Domain | \nKey Components | \nZustand Store | \n
|---|---|---|
| Workspace | \nWorkspaceLayout, PaneRenderer, PaneContent, WorkspaceTabBar | \nworkspace-store | \n
| Terminal | \nConnectedTerminal, XTerminal, TerminalSearchBar, ActivityIndicator | \nterminal-store | \n
| Editor | \nEditorPanel, CodeEditor, MarkdownEditor, EditorToolbar, MermaidBlock | \neditor-store | \n
| Browser | \nBrowserPanel, BrowserControls, AnnotationPanel, AnnotationExportModal | \nbrowser-session-store, annotation-store | \n
| File Explorer | \nFileExplorer, FileTreeNode, FileTreeContextMenu | \nโ | \n
| Snapshots | \nCreateSnapshotModal, RestoreSnapshotModal, DeleteSnapshotModal | \nsnapshot-store | \n
| Projects | \nProjectSidebar, NewProjectModal | \nproject-store | \n
| Settings | \nShortcutRecorder, ColorPickerPopover, ContextBarSettingsPopover | \napp-settings-store, context-bar-settings-store | \n
| Updates | \nUpdateAvailableToast, UpdateReadyModal | \nupdater-store | \n
| Shared | \nCommandPalette, ContextMenu, ConfirmDialog, ShellSelector, ErrorBoundary | \nโ | \n
brew tap gnoviawan/termul\nbrew install --cask termul\ncurl -fsSL https://raw.githubusercontent.com/gnoviawan/termul/main/scripts/install.sh | bash\nWindows users should install the .exe or .msi from GitHub Releases. Manual DMG downloads in a browser may still hit Gatekeeper, so macOS users should prefer Homebrew or curl.
| Dependency | \nVersion | \nNotes | \n
|---|---|---|
| Bun | \n1.3+ | \nJavaScript runtime and package manager | \n
| Rust | \nLatest stable | \nRequired for Tauri builds | \n
xcode-select --install\nsudo apt update\nsudo apt install libwebkit2gtk-4.1-dev \\\n build-essential curl wget file \\\n libxdo-dev libssl-dev \\\n libayatana-appindicator3-dev \\\n librsvg2-dev patchelf\nsudo dnf install webkit2gtk4.1-devel \\\n gcc gcc-c++ libopenssl-devel \\\n appindicator-devel librsvg2-devel \\\n patchelf\ncurl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh\nrustc --version && cargo --version\n# Clone the repository\ngit clone https://github.com/gnoviawan/termul.git\ncd termul\n\n# Install dependencies\nbun install\n\n# Launch in development mode\nbun run dev\nThis repository also includes a standalone Vite landing page under landing/.
# Install landing page dependencies (from landing/)\ncd landing && bun install\n\n# Start the landing page dev server\nbun run landing:dev\n\n# Lint the landing page\nbun run landing:lint\n\n# Build the landing page for production\nbun run landing:build\n# Build for your current platform\nbun run build\n\n# Platform-specific builds\nbun run build:tauri:win # Windows (x64)\nbun run build:tauri:mac-arm # macOS (Apple Silicon)\nbun run build:tauri:mac-x64 # macOS (Intel)\nbun run build:tauri:linux # Linux (x64)\n\n# Debug build (faster compilation, larger binary)\nbun run build:tauri:debug\nBuild output: src-tauri/target/release/bundle/
| Action | \nHow | \n
|---|---|
| New terminal | \nClick + next to tabs | \n
| Select specific shell | \nClick the dropdown arrow | \n
| Reorder tabs | \nDrag and drop | \n
| Rename tab | \nDouble-click the tab | \n
| Context menu | \nRight-click (rename, close, kill process) | \n
| Action | \nDefault Shortcut | \n
|---|---|
| New Terminal | \nCtrl+T | \n
| Next Tab | \nCtrl+PageDown | \n
| Previous Tab | \nCtrl+PageUp | \n
| Command Palette | \nCtrl+K / Ctrl+Shift+P | \n
\n\nShortcuts are customizable in Settings. On Tauri/WebView2, browser-reserved shortcuts such as
\nCtrl+Tabare not used as defaults because they are not reliably interceptable.
| Layer | \nTechnology | \n
|---|---|
| Desktop Runtime | \nTauri 2.0 | \n
| Backend | \nRust | \n
| UI Framework | \nReact 18 | \n
| Type System | \nTypeScript | \n
| Build Tool | \nVite | \n
| Styling | \nTailwind CSS + shadcn/ui | \n
| State Management | \nZustand | \n
| Terminal Emulation | \ntauri-pty + xterm.js | \n
| Animations | \nFramer Motion | \n
| Plugin | \nPurpose | \n
|---|---|
@tauri-apps/plugin-fs | \nFilesystem access | \n
@tauri-apps/plugin-store | \nConfiguration persistence | \n
@tauri-apps/plugin-os | \nOS information | \n
@tauri-apps/plugin-dialog | \nNative dialogs | \n
@tauri-apps/plugin-clipboard-manager | \nClipboard operations | \n
@tauri-apps/plugin-updater | \nAutomatic updates | \n
@tauri-apps/plugin-process | \nProcess management | \n
src/\nโโโ renderer/ # React frontend\nโ โโโ components/ # UI components\nโ โโโ hooks/ # Custom React hooks\nโ โโโ lib/ # Runtime adapters & desktop integration\nโ โโโ pages/ # Page components\nโ โโโ stores/ # Zustand stores\nโโโ shared/ # Shared types (main/renderer)\nsrc-tauri/ # Rust backend, config & bundling\ndocs/electron-old/ # Archived Electron docs & migration history\nThe renderer uses an adapter/service layer to keep desktop integrations isolated from UI code:
\nsrc/renderer/lib/\nโโโ tauri-*.ts # Tauri-native integrations\nโโโ *.ts # Runtime-safe facades & helpers\nโโโ __tests__/ # Regression & parity coverage\nbun run dev # Development mode with hot reload\nbun run test # Run tests\nbun run test:watch # Tests in watch mode\nbun run typecheck # Type checking\nbun run lint # Linting\nbun run tauri <command> # Direct Tauri CLI access\nssh-profiles.json.StrictHostKeyChecking=accept-new; do not override UserKnownHostsFile to /dev/null/NUL because that disables persistent host-key verification.ssh2 channel_direct_tcpip over the active SSH session; remote/reverse forwarding is not supported by the MVP command path yet.ssh2 intentionally stays on system OpenSSL for now. Enabling vendored-openssl forces a local OpenSSL source build that can fail in Windows/MSYS environments without a complete Perl module setup.Contributions are welcome! Please read the Contributing Guide for details on our code of conduct and the process for submitting pull requests.
\nThis project is licensed under the MIT License โ see the LICENSE file for details.
\nBuilt with โค๏ธ by gnoviawan
\n![\"Diskus](\"https://raw.githubusercontent.com/fadhilbarkah/diskus/main/public/assets/logo.png\"){kind=logo}
\n A lightweight, self-hosted comments system built for modern web applications.
\n![\"License:](\"https://img.shields.io/badge/License-GPLv3-blue.svg\"){kind=icon}
\n![\"Bun\"](\"https://img.shields.io/badge/Bun-%23000000.svg?logo=bun&logoColor=white\"){kind=icon}
\n
Diskus is designed to be a fast, privacy-respecting alternative to Disqus and other bloated third-party commenting services.
\n![\"Live](\"https://img.shields.io/badge/%F0%9F%9A%80_Try_the_Live_Demo-2ea44f?style=for-the-badge\"){kind=icon}
![\"Diskus](\"https://raw.githubusercontent.com/fadhilbarkah/diskus/main/public/assets/Diskus_Dashboard.png\")
\n Diskus Centralized Admin Dashboard
![\"Diskus](\"https://raw.githubusercontent.com/fadhilbarkah/diskus/main/public/assets/Diskus_Widget.png\")
\n Diskus Lightweight Widget
\n![\"Preact\"](\"https://img.shields.io/badge/Preact-673AB8?logo=preact&logoColor=white\"){kind=icon}
![\"Documentation\"](\"https://img.shields.io/badge/%F0%9F%93%96_Read_the_Docs-0366d6?style=for-the-badge\"){kind=icon}
embed.js), with a total footprint of ~28KB (gzipped), ensuring zero impact on your Core Web Vitals.Shadow DOM, ensuring your website's CSS never conflicts with the widget's design and vice-versa, without the heavy performance overhead of traditional iframes.isomorphic-dompurify) and Markdown parsing are offloaded to the server to maintain a minimal client bundle.Diskus operates as a monorepo containing three core packages:
\n/backend): A REST API built with Hono and running on Bun. Uses SQLite via Drizzle ORM for data persistence./dashboard): A Preact-based Single Page Application (SPA) for administrators to manage sites, moderate comments, and view users./widget): A highly optimized Preact component. The lightweight embed script (embed.js) dynamically injects the widget using a native Shadow DOM, guaranteeing 100% CSS isolation and zero style bleeding with the host website, while maintaining a featherlight ~28KB (gzipped) footprint containing full Tailwind CSS v4 logic.Clone the repository:
\ngit clone https://github.com/fadhilbarkah/diskus.git\ncd diskus\nInstall dependencies:
\nbun install\nSetup environment variables:\nCopy .env.example to .env in all three workspace directories (backend, dashboard, widget).
/backend): Configure the required JWT_SECRET. You can also configure DASHBOARD_ORIGIN (for CORS restrictions) and DATABASE_PATH (custom SQLite path).VITE_API_URL to point to your backend API URL.Initialize the database schema and optionally seed initial data:
\ncd backend\nbun run db:push\n# Optional: populate the database with test data and a default admin account\nbun run src/db/seed.ts\nStart the development server (runs backend, dashboard, and widget concurrently):
\n# From the project root\nbun dev\n\n\nNote: When you open the Dashboard (
\nhttp://localhost:5173) for the first time, you will be automatically prompted to create your initial Admin account. No manual seeding is required!
![\"Deploy](\"https://railway.com/button.svg\")
Click the button above to instantly deploy both the Backend API and the Dashboard Frontend. All environment variables, volumes, and start commands are pre-configured in this official template.
\nDiskus is fully containerized for easy production deployment using Docker Compose. We provide a one-click startup script that automatically handles secure secret generation.
\nBuild and start the services:
\n# Run the start script\n./start.sh\n\n\nNote: The script will automatically generate a secure
\n.envfile with a strongJWT_SECRETif one does not exist, and then rundocker-compose up -d --build.
The services will be available at:
\nhttp://localhost:5173 (or your domain)http://localhost:3000\n\nNote: The database uses a Docker Volume (
\ndiskus-data), so your comments will persist even if you restart the containers.
Instead of manually seeding the database, simply open your frontend domain in the browser. You will be greeted with a \"Create Admin Account\" screen. Register your account immediately to secure your deployment, as the setup screen will permanently disappear once the first admin is created.
\nIf you ever need to completely wipe your production database (e.g., to resolve severe migration conflicts or start fresh), you must destroy the Docker Named Volume. WARNING: This will permanently delete all comments and user data.
\n# Bring down containers and DESTROY the database volume (-v)\ndocker-compose down -v\n\n# Restart the services (a fresh database will be created)\n./start.sh\nOpen the Dashboard (http://localhost:5173), navigate to Websites, and register a new domain. You will receive an App ID.
Paste the following HTML snippet into your target website, replacing the data attributes with your specific keys:
\n<!-- Diskus Embed -->\n<div id=\"diskus-thread\" \n data-app-id=\"YOUR_APP_ID\" \n data-thread-key=\"your-unique-page-identifier\"\n data-api-url=\"http://localhost:3000/api/v1\">\n</div>\n<script src=\"http://localhost:5173/widget/dist/embed.js\" async defer></script>\n\n\nNote: The
\ndata-thread-keyshould be unique per page (e.g., the article slug or ID) so that comments remain tied to their specific content.
If you are locked out of the dashboard, you can reset your password securely via the command line. This requires direct access to your server terminal or Railway console.
\nbackend directory (if not already there).bun run reset-password admin@example.com\nPull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
\nPlease make sure to update tests as appropriate.
\n๏ปฟ> This repository is frozen at v2.0.0-community (LTS).
\n\n\nNo further releases will be tagged. Self-compiling this source still\nproduces a working Community binary under GPL v3.\nThe Premium edition (Web Print Premium, Email-to-Print, Driver\nAuto-Install) is distributed separately as an official signed binary.\nSee docs/RELEASE.md for the full release strategy.
\n
![\"ShaPrint](\"https://raw.githubusercontent.com/ardli-firman/sha-print/main/ShaPrint.WpfApp/Assets/Icons/Logo_rounded.png\"){kind=icon}
\n The Simplest & Most Reliable LAN / Cross-VLAN Virtual Printer Sharing Solution for Windows
\nShaPrint is an advanced, .NET 8-based application designed to reliably share physical printers across local networks (LAN) and cross-subnet/VLAN environments. It serves as a robust alternative when native Windows SMB Printer Sharing fails, struggles with network credential conflicts, or is obstructed by strict Windows security policies.
\nBy utilizing a Virtual Printer Port (Named Pipes) architecture and direct TCP/UDP transmission, ShaPrint guarantees that documents are printed with 100% fidelity and native quality.
\nSwitch Mode![](\"https://raw.githubusercontent.com/ardli-firman/sha-print/main/img/1.png\") | \n ||||
Server Mode![](\"https://raw.githubusercontent.com/ardli-firman/sha-print/main/img/2.png\") | \n Client Mode![](\"https://raw.githubusercontent.com/ardli-firman/sha-print/main/img/3.png\") | \n Monitoring![](\"https://raw.githubusercontent.com/ardli-firman/sha-print/main/img/6.png\") | \n Settings![](\"https://raw.githubusercontent.com/ardli-firman/sha-print/main/img/4.png\") | \n Update Manager![](\"https://raw.githubusercontent.com/ardli-firman/sha-print/main/img/5.png\") | \n
Server Mode\nRunning on the computer directly connected to the physical printer via USB or LAN, the Server scans for local printers and listens for raw print spool data on TCP Port 9877. It also broadcasts its presence using UDP Port 9876 for auto-discovery.
\nClient Mode\nThe application intercepts print jobs by creating a Virtual Printer Port within the Windows Spooler. Any document printed from standard applications (Word, Chrome, Acrobat) to this virtual printer is instantly intercepted and streamed directly to the Server.
\nMonitoring Mode\nOperators or users can monitor all active ShaPrint servers across the network channel. It discovers active servers via UDP and queries their status on TCP Port 9878 to retrieve real-time encrypted details.
\nBefore installing ShaPrint, please ensure your system meets the following requirements:
\nx64 architecture)10.0.17763) or newer (released November 2018)ShaPrint implements defense-in-depth security to protect your local network, computer performance, and hardware.
\n| Layer | \nMechanism | \nDescription | \n
|---|---|---|
| TCP Data (Print/Scan/Monitor) | \nAES-256-GCM | \nAll TCP payloads (print jobs, scan data, monitoring status) are encrypted with AES-256 in Galois/Counter Mode โ providing both confidentiality and tamper detection. Each encryption uses a fresh random 96-bit nonce. | \n
| UDP Discovery | \nHMAC-SHA256 | \nDiscovery responses are signed with HMAC-SHA256. Clients verify the signature before trusting any server response, preventing spoofing and man-in-the-middle attacks. | \n
| Key Derivation | \nPBKDF2 (100k iterations) | \nAll cryptographic keys are derived from the Network Channel shared secret using PBKDF2 with 100,000 iterations and unique salts per purpose (AES, HMAC, local config). | \n
| Config Integrity | \nHMAC-wrapped JSON | \nServer configuration files are stored with an embedded HMAC to detect tampering. Corrupted or modified configs are rejected on load. | \n
' \" ; $ \\ | & < > \\n \\r \\t \\0`) are explicitly blocked.skipUnicastSweep=true).ShaPrint interacts with hardware exclusively through standard Windows APIs:
\nAddJob / spool file injection)| Concern | \nVerdict | \nNotes | \n
|---|---|---|
| Local network security | \n๐ข Safe (with configuration) | \nAES-256-GCM + HMAC-SHA256 provide strong protection. Must customize Network Channel from default for multi-tenant environments. | \n
| Computer performance | \n๐ข Safe | \nCPU/RAM impact is negligible for monitoring. Print/scan load is temporary and on-demand. | \n
| Hardware damage | \n๐ข Safe | \nSoftware-only โ uses only standard Windows APIs (Spooler, WIA). No risk of physical damage. | \n
๐ด Customize Your Network Channel
\n\"DefaultChannel\" to a unique, random string known only to your devicesCoordinate with IT
\nUDP 9876 โ Service discoveryTCP 9877 โ Print/scan data transferTCP 9878 โ Server monitoring statusUse Only on Trusted Networks
\nShaPrint is packaged as a fully self-contained Standalone Setup. You do not need to install the .NET Runtime manually.
\nShaPrint_Setup_vX.Y.Z.exe from the GitHub Releases page.\n\n[!IMPORTANT]
\n
Native Driver Requirement: To guarantee print fidelity, you must install the official printer driver on the Client PC. For example, if the Server is hosting an Epson L3210, you must install the Epson L3210 driver on the Client PC beforehand.
Ctrl + P, select ShaPrint - [Printer Name], and Print!To compile the source code and generate the installer yourself, ensure you have the .NET 8 SDK and Inno Setup 6 installed.
\nOpen a terminal in the root directory and run the following commands to publish the binaries:
\n# Publish the main WPF Application\ndotnet publish ShaPrint.WpfApp/ShaPrint.WpfApp.csproj -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -p:IncludeNativeLibrariesForSelfExtract=true\n\n# Publish the Background Updater\ndotnet publish ShaPrint.Updater/ShaPrint.Updater.csproj -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -p:IncludeNativeLibrariesForSelfExtract=true\nUsing PowerShell, compile the .iss script:
& 'C:\\Program Files (x86)\\Inno Setup 6\\ISCC.exe' installer.iss\nYour compiled installer (ShaPrint_Setup_v1.0.x.exe) will be generated inside the Output\\ directory.
9876/UDP and 9877/TCP are open on the Server's Windows Firewall. If you are on a different subnet, auto-discovery will not workโuse the \"Specific Server IP\" feature.Turn your AI assistant into a tutor. It plans, hints, reviews and adapts while you write every line.
\n\n\nPair with vibekit: vibekit when you want speed, no-vibe when you want to learn.
\n
Vibe-coding produces output without producing understanding and copy-typing what the AI shows you produces the same hollow result one keystroke at a time. The thing that actually transfers is the thought process and manual code writing: deciding what to do, predicting what will happen and naming what broke, no-vibe is built so you contribute that, not just keystrokes.
\nhint / analogy / pseudo / show / less), so the code you write comes from a decision you made, not a block you transcribed.--ref <url> and the AI quotes actual source with file:line citations.PROFILE.md (global, stable identity) and a SUMMARY.md (per-project, running journey) it writes itself โ observed strengths, known gaps, style notes, current focus, open questions. You can edit either, or layer explicit overrides via user/*.md..no-vibe/**. Codex/Gemini enforce the same rule via instruction.no-vibe uses a four-layer stack, split by write cadence and scope:
\n| Layer | \nOwner | \nWhat lives in it | \n
|---|---|---|
| Default teaching style | \nPlugin โ defined in skills/no-vibe/SKILL.md | \nPlain words first, concrete-before-abstract, hint-before-answer, run + verify after every layer. The floor. | \n
~/.no-vibe/PROFILE.md (global, stable identity) | \nAI โ created on first /no-vibe, rewritten rarely when cross-project identity / style shifts | \nIdentity & expertise, learning style, disclosure mode, observed strengths, known gaps | \n
.no-vibe/SUMMARY.md (project, running journey) | \nAI โ created at the first layer close worth recording, rewritten often (every closed layer is a candidate) | \nCurrent Focus, Accomplishments, Open Questions in this project | \n
~/.no-vibe/user/*.md and .no-vibe/user/*.md | \nYou โ AI never creates, edits, or deletes anything inside | \nExplicit overrides: instructions you want the AI to follow without inferring them | \n
Why the split. Stable identity (the things that wouldn't change if you opened a different project tomorrow) and running journey (the things that only make sense inside this project) update on totally different cadences. Keeping them in one file forces the AI to decide on every rewrite whether this fact is stable or transient โ and gets it wrong. PROFILE only holds cross-project-durable facts; SUMMARY only holds project-bound state.
\nPROFILE.md is the AI's global progression file. On your first /no-vibe activation, the AI creates it with empty section headings (Identity & expertise, Learning style, Disclosure mode, Observed strengths, Known gaps). It's rewritten only when something durable about how you learn shifts โ most layers produce no PROFILE update.
SUMMARY.md is the AI's per-project journey file. It's not seeded on activation โ the AI creates it the first time a layer close produces an outcome worth recording, then keeps it tight by pruning resolved Open Questions and old Accomplishments. The most valuable section is Open Questions โ things you dodged with a workaround or didn't fully integrate, surfaced so the next session can revisit them.
The silent-default + NO_CHANGE rule. Both files follow two disciplines: most layer-closes produce no write (silent default), and the AI never rewrites a file with content equivalent to what's already there (NO_CHANGE). A no-op write is treated as a bug. Read either file any time to see what the AI has learned; edit them yourself if something looks wrong.
\nuser/*.md is your override layer. Drop any .md file into ~/.no-vibe/user/ (global) or .no-vibe/user/ (project) and the AI loads it sorted by filename. Anything in user/ wins on conflict with PROFILE.md or the default style. The AI is forbidden from writing to user/ โ when it notices a pattern that belongs there, it shows you the exact line and lets you add it.
Practical examples โ anything in this style works in user/*.md:
Per-session cycle state (current phase, layer, resume hints) lives separately in .no-vibe/data/sessions/<slug>.json โ you generally don't touch that.
/plugin marketplace add rizukirr/no-vibe\n/plugin install no-vibe@no-vibe\nRestart Claude Code.
\ncodex plugin marketplace add rizukirr/no-vibe\ncodex plugin add no-vibe --marketplace no-vibe\n(Requires a Codex CLI build with plugin marketplace support. For older Codex builds, see INSTALL.codex.md for the manual symlink path โ skills only, soft block.)
pi install npm:no-vibe\nOr pin to git: pi install git:github.com/rizukirr/no-vibe. See INSTALL.pi.md for verification steps.
gemini extensions install https://github.com/rizukirr/no-vibe\nPin a version with --ref=v2.0.3. See INSTALL.gemini.md for the legacy manual-symlink path.
Add to ~/.config/opencode/opencode.json:
{\n \"$schema\": \"https://opencode.ai/config.json\",\n \"plugin\": [\"no-vibe\"]\n}\nOpenCode has no plugin install CLI, so commands also need to be fetched once:
mkdir -p ~/.config/opencode/commands\nfor c in no-vibe no-vibe-challenge no-vibe-btw; do\n curl -fsSL \"https://raw.githubusercontent.com/rizukirr/no-vibe/refs/heads/main/.opencode/commands/$c.md\" \\\n -o \"$HOME/.config/opencode/commands/$c.md\"\ndone\nSee INSTALL.opencode.md for verification steps and the cache-refresh tip.
/no-vibe build a linear layer like pytorch's\nCodex uses $ instead of /.
| Command | \nEffect | \n
|---|---|
/no-vibe on / off | \npersistent mode toggle | \n
/no-vibe <topic> | \none-shot lesson | \n
/no-vibe --ref <url> <topic> | \nattach a reference project | \n
/no-vibe --mode concept|skill|debug <topic> | \nset voice mode | \n
/no-vibe-btw <task> | \none-shot escape hatch โ AI may write for this task only | \n
/no-vibe-challenge [<focus>] | \nget a coding challenge | \n
Flags combine: /no-vibe --ref pytorch --mode concept how does autograd work.
| Mode | \nBest for | \nStyle | \n
|---|---|---|
| concept (default) | \n\"teach me how X works\" | \nmore prose, deeper check-ins | \n
| skill | \n\"I want to practice writing Y\" | \nmuscle-memory repetition | \n
| debug | \n\"why does Z behave like this\" | \nstart from symptom, descend | \n
Voice modes control how AI talks. A separate axis, disclosure modes (guided write vs. showcase), controls how much AI reveals before the user writes code in a Phase 3 layer โ guided is the default and walks the user toward the code with English + graded hints on request; showcase shows the full code block upfront. Both default to running a one-question prediction gate before the user runs the code each layer, so the run becomes a self-test rather than passive verification. See skills/no-vibe/SKILL.md for the full disclosure-mode contract and the help verbs (hint / analogy / pseudo / show / less).
| Feature | \nClaude Code | \nOpenCode | \nPi | \nCodex | \nGemini CLI | \n
|---|---|---|---|---|---|
| File-write guard (hook) | \nโ | \nโ | \nโ | \nโ * | \nsoft | \n
| Bash-write guard (hook) | \nโ | \nโ | \nโ | \nโ * | \nsoft | \n
| Status + resume hint | \nโ | \nโ | \nโ | \nโ * | \nsoft | \n
| Commands | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| PROFILE.md + SUMMARY.md + user/ overrides | \nโ | \nโ | \nโ | \nโ | \nโ | \n
* Codex hooks fire under the marketplace install (codex plugin add no-vibe --marketplace no-vibe). The legacy manual-symlink install path is soft-only.
\"soft\" = instruction-enforced (no hook surface available); the rule still binds.
\nMIT. Issues and PRs welcome at github.com/rizukirr/no-vibe/issues.
\n" }, + { + "fullName": "ardli-firman/sha-print", + "name": "sha-print", + "owner": "ardli-firman", + "ownerAvatarUrl": "https://avatars.githubusercontent.com/u/48121202?v=4", + "description": "The Simplest & Most Reliable LAN / Cross-VLAN Virtual Printer Sharing Solution for Windows", + "metaDescription": "The Simplest & Most Reliable LAN / Cross-VLAN Virtual Printer Sharing Solution for Windows", + "url": "https://github.com/ardli-firman/sha-print", + "homepage": "", + "language": "C#", + "stars": 34, + "forks": 13, + "topics": [ + "desktop-app", + "printer", + "sharing", + "sharing-printer", + "windows" + ], + "updatedAt": "2026-07-18T03:45:52Z", + "pushedAt": "2026-07-18T03:40:27Z", + "latestRelease": { + "name": "ShaPrint v1.6.0 (Stable)", + "tagName": "v1.6.0-stable", + "url": "https://github.com/ardli-firman/sha-print/releases/tag/v1.6.0-stable", + "publishedAt": "2026-07-13T04:40:11Z" + }, + "archived": true, + "licenseSpdx": "", + "createdAt": "2026-05-23T05:37:35Z", + "openIssues": 0, + "openPullRequests": 0, + "subscribers": 0, + "communityHealth": 28, + "readmeHtml": "๏ปฟ> This repository is frozen at v2.0.0-community (LTS).
\n\n\nNo further releases will be tagged. Self-compiling this source still\nproduces a working Community binary under GPL v3.\nThe Premium edition (Web Print Premium, Email-to-Print, Driver\nAuto-Install) is distributed separately as an official signed binary.\nSee docs/RELEASE.md for the full release strategy.
\n
\n The Simplest & Most Reliable LAN / Cross-VLAN Virtual Printer Sharing Solution for Windows
\nShaPrint is an advanced, .NET 8-based application designed to reliably share physical printers across local networks (LAN) and cross-subnet/VLAN environments. It serves as a robust alternative when native Windows SMB Printer Sharing fails, struggles with network credential conflicts, or is obstructed by strict Windows security policies.
\nBy utilizing a Virtual Printer Port (Named Pipes) architecture and direct TCP/UDP transmission, ShaPrint guarantees that documents are printed with 100% fidelity and native quality.
\n| Switch Mode | \n ||||
| Server Mode | \n Client Mode | \n Monitoring | \n Settings | \n Update Manager | \n
Server Mode\nRunning on the computer directly connected to the physical printer via USB or LAN, the Server scans for local printers and listens for raw print spool data on TCP Port 9877. It also broadcasts its presence using UDP Port 9876 for auto-discovery.
\nClient Mode\nThe application intercepts print jobs by creating a Virtual Printer Port within the Windows Spooler. Any document printed from standard applications (Word, Chrome, Acrobat) to this virtual printer is instantly intercepted and streamed directly to the Server.
\nMonitoring Mode\nOperators or users can monitor all active ShaPrint servers across the network channel. It discovers active servers via UDP and queries their status on TCP Port 9878 to retrieve real-time encrypted details.
\nBefore installing ShaPrint, please ensure your system meets the following requirements:
\nx64 architecture)10.0.17763) or newer (released November 2018)ShaPrint implements defense-in-depth security to protect your local network, computer performance, and hardware.
\n| Layer | \nMechanism | \nDescription | \n
|---|---|---|
| TCP Data (Print/Scan/Monitor) | \nAES-256-GCM | \nAll TCP payloads (print jobs, scan data, monitoring status) are encrypted with AES-256 in Galois/Counter Mode โ providing both confidentiality and tamper detection. Each encryption uses a fresh random 96-bit nonce. | \n
| UDP Discovery | \nHMAC-SHA256 | \nDiscovery responses are signed with HMAC-SHA256. Clients verify the signature before trusting any server response, preventing spoofing and man-in-the-middle attacks. | \n
| Key Derivation | \nPBKDF2 (100k iterations) | \nAll cryptographic keys are derived from the Network Channel shared secret using PBKDF2 with 100,000 iterations and unique salts per purpose (AES, HMAC, local config). | \n
| Config Integrity | \nHMAC-wrapped JSON | \nServer configuration files are stored with an embedded HMAC to detect tampering. Corrupted or modified configs are rejected on load. | \n
' \" ; $ \\ | & < > \\n \\r \\t \\0`) are explicitly blocked.skipUnicastSweep=true).ShaPrint interacts with hardware exclusively through standard Windows APIs:
\nAddJob / spool file injection)| Concern | \nVerdict | \nNotes | \n
|---|---|---|
| Local network security | \n๐ข Safe (with configuration) | \nAES-256-GCM + HMAC-SHA256 provide strong protection. Must customize Network Channel from default for multi-tenant environments. | \n
| Computer performance | \n๐ข Safe | \nCPU/RAM impact is negligible for monitoring. Print/scan load is temporary and on-demand. | \n
| Hardware damage | \n๐ข Safe | \nSoftware-only โ uses only standard Windows APIs (Spooler, WIA). No risk of physical damage. | \n
๐ด Customize Your Network Channel
\n\"DefaultChannel\" to a unique, random string known only to your devicesCoordinate with IT
\nUDP 9876 โ Service discoveryTCP 9877 โ Print/scan data transferTCP 9878 โ Server monitoring statusUse Only on Trusted Networks
\nShaPrint is packaged as a fully self-contained Standalone Setup. You do not need to install the .NET Runtime manually.
\nShaPrint_Setup_vX.Y.Z.exe from the GitHub Releases page.\n\n[!IMPORTANT]
\n
Native Driver Requirement: To guarantee print fidelity, you must install the official printer driver on the Client PC. For example, if the Server is hosting an Epson L3210, you must install the Epson L3210 driver on the Client PC beforehand.
Ctrl + P, select ShaPrint - [Printer Name], and Print!To compile the source code and generate the installer yourself, ensure you have the .NET 8 SDK and Inno Setup 6 installed.
\nOpen a terminal in the root directory and run the following commands to publish the binaries:
\n# Publish the main WPF Application\ndotnet publish ShaPrint.WpfApp/ShaPrint.WpfApp.csproj -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -p:IncludeNativeLibrariesForSelfExtract=true\n\n# Publish the Background Updater\ndotnet publish ShaPrint.Updater/ShaPrint.Updater.csproj -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -p:IncludeNativeLibrariesForSelfExtract=true\nUsing PowerShell, compile the .iss script:
& 'C:\\Program Files (x86)\\Inno Setup 6\\ISCC.exe' installer.iss\nYour compiled installer (ShaPrint_Setup_v1.0.x.exe) will be generated inside the Output\\ directory.
9876/UDP and 9877/TCP are open on the Server's Windows Firewall. If you are on a different subnet, auto-discovery will not workโuse the \"Specific Server IP\" feature.Muslimtify keeps you consistent with your daily prayers by delivering accurate prayer times and timely desktop notifications. Designed for Linux and Windows, it automatically calculates prayer schedules and reminds you 30, 15, and 5 minutes before the Adhan โ or at your own custom intervals โ and when it's time to pray. All calculations run locally, requiring no internet connection or external services.
\nMuslimtify supports 21 international calculation methods including MWL, ISNA, Umm al-Qura (Makkah), Egyptian General Authority, Kemenag (Indonesia), JAKIM (Malaysia), Diyanet (Turkey), and more. The default method is Kemenag. With persistent configuration and minimal setup, Muslimtify integrates seamlessly into your daily routine without interrupting your workflow.
\n\n\n[!Note]\nPrayer time calculations are powered by libmuslim, a portable library extracted from this project to enable a more flexible and reusable ecosystem for Muslim developers.
\n
| Linux | \nWindows | \n
|---|---|
\n\nRoadmap
\n\n
\n- \n
Merge commandlocation autoandmethod autointo (only)config autoand optimize auto detection per (249) country- \n
Refactor from timer-driven into a portable long-running loop- \n
Add custom adzan sound notifications- \n
Re-design command-line (BREAKING CHANGES)- Add read lat/long from user GPS
\n- Add GUI (see branch gui to see a progress)
\n- Distribute to Flatpak
\n- MacOS support (if devices is available)
\n- Wearable Device support
\n- Embedded Device Support
\n
\n\n[!Important]\nThis project is available for Linux and Windows users, but not yet for Mac users because we need a Mac device to make Muslimtify run on macOS. We are looking for brothers and sisters who have a Mac and experience in low-level C programming to contribute to the project and help bring Muslimtify to macOS. Alternatively, you can support us via GitHub Sponsors in the sponsor section.
\n
Every release ships ready-to-run binaries for Linux and Windows on the\nReleases page.
\nLinux (x86_64 or aarch64) โ the binaries are dynamically linked, so\ninstall the runtime libraries first, then extract and install:
# Ubuntu/Debian\nsudo apt install libnotify4 libcurl4\n# Fedora/RHEL\nsudo dnf install libnotify libcurl\n# Arch\nsudo pacman -S libnotify curl\n\ntar xzf muslimtify-<version>-linux-<arch>.tar.gz\nsudo cp -r muslimtify-<version>-linux-<arch>/{bin,lib,share} /usr/local/\nmuslimtify daemon install\nWindows (x64 or arm64) โ download and run the matching installer:
muslimtify-<version>-setup-x64.exe # Intel/AMD\nmuslimtify-<version>-setup-arm64.exe # ARM\nVerify any download against the published checksums:
\nsha256sum -c SHA256SUMS\nyay -S muslimtify\nsudo dnf copr enable rizukirr/muslimtify\nsudo dnf install muslimtify\nsudo add-apt-repository ppa:rizukirr/muslimtify\nsudo apt update\nsudo apt install muslimtify\nInstall dependencies:
\n# Ubuntu/Debian\nsudo apt install git build-essential cmake pkg-config libnotify-dev libcurl4-openssl-dev\n\n# Fedora/RHEL\nsudo dnf install git gcc cmake pkgconfig libnotify-devel libcurl-devel\n\n# Arch Linux\nsudo pacman -S git base-devel cmake pkgconfig libnotify curl\nClone, install, and enable background checks:
\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\nsudo ./install.sh\nmuslimtify daemon install\nwinget install muslimtify\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\n.\\install.ps1\nmuslimtify daemon install\nTo remove the Windows install later, run .\\uninstall.ps1.
If you prefer building manually first:
\ncmake -S . -B build\ncmake --build build --config Release\ncmake --install build --config Release\nmuslimtify daemon install\nRun muslimtify daemon status to check if Muslimtify is registered with systemd. If no status is found, run muslimtify daemon install to register the service and ensure it runs as expected.
Muslimtify automatically selects the standard prayer time calculation method based on your country and location. Run muslimtify to verify that your configuration is correct. If the automatic selection does not meet your needs, you can set it manually using muslimtify method <key-method>. A full list of available methods is documented here.
Muslimtify can be configured with CLI commands or by editing config.json\nmanually.
Config paths:
\n~/.config/muslimtify/config.json~/.cache/muslimtify%APPDATA%\\muslimtify\\config.json%LOCALAPPDATA%\\muslimtifyCommon setup commands:
\nmuslimtify location set --auto # detect location from IP\nmuslimtify location set --auto --city=Mansoura # auto-detect but use your own city label\nmuslimtify method --auto # select method from the detected country\nmuslimtify location set --lat=-6.175 --long=106.82 # set location manually (uses system timezone)\nmuslimtify location set --timezone=Asia/Jakarta # override timezone\nmuslimtify location set --city=Jakarta # add a city label\nmuslimtify location set --refresh-interval=21600 # re-check location every 6h (0=off, min 3600)\nmuslimtify method --list # list all available calculation methods\nmuslimtify method mwl # set calculation method\nmuslimtify madzhab hanafi # set madzhab (shafi/hanafi)\nmuslimtify notification --reminder --all 30 15 5 # set every prayer's reminders (minutes before adhan)\nmuslimtify notification --reminder fajr 30 15 5 # set reminders for a single prayer\nmuslimtify notification # show current notification settings\nmuslimtify location # show current location\nResetting the configuration is done by deleting config.json. Muslimtify falls\nback to built-in defaults when the file is missing, and rewrites it the next\ntime you change a setting. Validation runs automatically every time the config\nis loaded.
Muslimtify supports the following calculation methods:
\n| Key | \nMethod | \nRegion | \n
|---|---|---|
mwl | \nMuslim World League | \nEurope, Far East | \n
makkah | \nUmm al-Qura, Makkah | \nArabian Peninsula | \n
isna | \nISNA | \nNorth America | \n
egypt | \nEgyptian General Authority | \nAfrica, Middle East | \n
karachi | \nUniv. Islamic Sciences, Karachi | \nPakistan, India, Bangladesh | \n
turkey | \nDiyanet, Turkey | \nTurkey | \n
singapore | \nMUIS, Singapore | \nSingapore | \n
jakim | \nJAKIM, Malaysia | \nMalaysia | \n
kemenag | \nKEMENAG, Indonesia | \nIndonesia (default) | \n
france | \nUOIF, France | \nFrance | \n
russia | \nSpiritual Admin., Russia | \nRussia | \n
dubai | \nGAIAE, Dubai | \nUAE | \n
qatar | \nMin. of Awqaf, Qatar | \nQatar | \n
kuwait | \nMin. of Awqaf, Kuwait | \nKuwait | \n
jordan | \nMin. of Awqaf, Jordan | \nJordan | \n
gulf | \nGulf Region | \nGulf states | \n
tunisia | \nMin. of Religious Affairs | \nTunisia | \n
algeria | \nMin. of Religious Affairs | \nAlgeria | \n
morocco | \nMin. of Habous, Morocco | \nMorocco | \n
portugal | \nComunidade Islamica de Lisboa | \nPortugal | \n
moonsighting | \nMoonsighting Committee | \nWorldwide | \n
You can also use a custom method by setting \"method\": \"custom\" in config.json with your own fajr_angle and isha_angle values.
Manual JSON editing is useful when you want precise control over enabled\nprayers, reminder offsets, notification settings, or location data.
\n\nDefault config.json{\n \"location\": {\n \"latitude\": 0.0,\n \"longitude\": 0.0,\n \"timezone\": \"UTC\",\n \"timezone_offset\": 0.0,\n \"auto_detect\": true,\n \"city\": \"\",\n \"country\": \"\"\n },\n \"prayers\": {\n \"fajr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"sunrise\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuha\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuhr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"asr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"maghrib\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"isha\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n }\n },\n \"notification\": {\n \"timeout\": 5000,\n \"urgency\": \"critical\",\n \"sound\": \"adhan\",\n \"sound_alarm\": \"alarm\",\n \"sound_reminder\": \"reminder\",\n \"icon\": \"muslimtify\"\n },\n \"calculation\": {\n \"method\": \"kemenag\",\n \"madhab\": \"shafi\"\n }\n}\nmuslimtify daemon status to confirm the background service is running.notify-send \"Test\" \"Hello\".muslimtify location set --auto again.muslimtify location set --lat=<latitude> --long=<longitude>.\nIf the host machine is in a different region than the coordinates, override\nthe timezone with --timezone=<iana>, e.g.\nmuslimtify location set --lat=-6.21 --long=106.84 --timezone=Asia/Jakarta.ipinfo.io if auto detection keeps failing.Contributions are welcome. See CONTRIBUTING.md for workflow,\nstyle, and testing guidance.
\nMuslimtify is released under the MIT License. See the repository license files\nfor details.
\nA lightweight C header-only library for calculating Islamic prayer times. Supports 21 international calculation methods including MWL, ISNA, Umm al-Qura (Makkah), Egyptian General Authority, Kemenag (Indonesia), JAKIM (Malaysia), Diyanet (Turkey), and more. The default method is Kemenag.
\n| Key | \nMethod | \nRegion | \n
|---|---|---|
mwl | \nMuslim World League | \nEurope, Far East | \n
makkah | \nUmm al-Qura, Makkah | \nArabian Peninsula | \n
isna | \nISNA | \nNorth America | \n
egypt | \nEgyptian General Authority | \nAfrica, Middle East | \n
karachi | \nUniv. Islamic Sciences, Karachi | \nPakistan, India, Bangladesh | \n
turkey | \nDiyanet, Turkey | \nTurkey | \n
singapore | \nMUIS, Singapore | \nSingapore | \n
jakim | \nJAKIM, Malaysia | \nMalaysia | \n
kemenag | \nKEMENAG, Indonesia | \nIndonesia (default) | \n
france | \nUOIF, France | \nFrance | \n
russia | \nSpiritual Admin., Russia | \nRussia | \n
dubai | \nGAIAE, Dubai | \nUAE | \n
qatar | \nMin. of Awqaf, Qatar | \nQatar | \n
kuwait | \nMin. of Awqaf, Kuwait | \nKuwait | \n
jordan | \nMin. of Awqaf, Jordan | \nJordan | \n
gulf | \nGulf Region | \nGulf states | \n
tunisia | \nMin. of Religious Affairs | \nTunisia | \n
algeria | \nMin. of Religious Affairs | \nAlgeria | \n
morocco | \nMin. of Habous, Morocco | \nMorocco | \n
portugal | \nComunidade Islamica de Lisboa | \nPortugal | \n
moonsighting | \nMoonsighting Committee | \nWorldwide | \n
You can also use a custom method by passing CALC_CUSTOM with your own angles.
Each method defines a set of parameters:
\nFor detailed mathematical formulas and worked examples, see docs/KEMENAG_METHOD.md.
This is a single-header library, so you can simply include prayertimes.h in your project.
# Compile the CLI tool\ngcc -O3 -o libmuslim main.c -lm\n\n# Run example (Bekasi, November 21, 2025)\n./libmuslim 2025 11 21 -6.2851291 106.9814968 7.0\n#include \"prayertimes.h\"\n\n// Use the default method (Kemenag)\nconst MethodParams *params = method_params_get(CALC_KEMENAG);\n\nstruct PrayerTimes times = calculate_prayer_times(\n 2025, // year\n 11, // month\n 21, // day\n -6.2851291, // latitude (negative = South)\n 106.9814968, // longitude (positive = East)\n 7.0, // timezone offset (WIB = UTC+7)\n params // calculation method\n);\n\nchar buffer[16];\nformat_time_hm(times.fajr, buffer, sizeof(buffer));\nprintf(\"Fajr: %s\\n\", buffer);\n// Use MWL method\nconst MethodParams *mwl = method_params_get(CALC_MWL);\nstruct PrayerTimes times = calculate_prayer_times(2025, 11, 21, 51.5074, -0.1278, 0.0, mwl);\n\n// Look up method by string key\nCalcMethod method = method_from_string(\"isna\");\nconst MethodParams *params = method_params_get(method);\n\n\nNOTE:
\nprayertimes.hdoes not handle Daylight Saving Time. The\ntimezoneargument is a fixed numeric UTC offset in hours, and the library\nuses it exactly as given โ it has no notion of dates, zones, or DST rules.\nThis is deliberate: DST is a political rule, not an astronomical one, and\nkeeping it out leavesprayertimes.ha pure, dependency-free (only<math.h>)\nsingle header. For a DST-active date you must pass the DST-adjusted offset\n(e.g.1.0for London in summer,0.0in winter).
If you want libmuslim to compute the correct offset for you, use the optional\ncompanion header timezone.h. It resolves an IANA zone name and\ndate to a UTC offset with DST applied, using the host operating system's\ntimezone database:
#define MUSLIM_TIMEZONE_IMPLEMENTATION // in exactly ONE translation unit\n#include \"timezone.h\"\n#include \"prayertimes.h\"\n\nchar zone[64];\nget_system_timezone(zone, sizeof(zone)); // e.g. \"Europe/London\"\ndouble tz = parse_timezone_offset(zone, time(NULL)); // DST already applied\n\nconst MethodParams *mwl = method_params_get(CALC_MWL);\nstruct PrayerTimes times = calculate_prayer_times(2026, 7, 15, 51.5074, -0.1278, tz, mwl);\nUnlike prayertimes.h, timezone.h touches the OS (POSIX tzset/tm_gmtoff\nor the Win32 timezone APIs), so it is optional โ include it only if you\nwant this resolution done for you. On a platform without a timezone database,\nkeep supplying the offset yourself.
./libmuslim <year> <month> <day> <latitude> <longitude> <timezone>\nExample output:
\nFajr = 04:05\nSunrise = 05:22\nDhuha = 05:50\nDhuhr = 11:41\nAsr = 15:04\nMaghrib = 17:54\nIsha = 19:07\nThe calculations have been verified against official data sources and match within ยฑ1-2 minute accuracy depending on the method. See the worked examples in docs/KEMENAG_METHOD.md for detailed verification.
Copyright 2025 Rizki Rakasiwi.\n\nLicensed under the Apache License, Version 2.0 (the \"License\");\nyou may not use this file except in compliance with the License.\nYou may obtain a copy of the License at\n\n http://www.apache.org/licenses/LICENSE-2.0\n\nUnless required by applicable law or agreed to in writing, software\ndistributed under the License is distributed on an \"AS IS\" BASIS,\nWITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\nSee the License for the specific language governing permissions and\nlimitations under the License.\nContributions are welcome! Please ensure any changes to calculation methods are verified against official data sources.
\n" + "readmeHtml": "A lightweight C header-only library for calculating Islamic prayer times. Supports 21 international calculation methods including MWL, ISNA, Umm al-Qura (Makkah), Egyptian General Authority, Kemenag (Indonesia), JAKIM (Malaysia), Diyanet (Turkey), and more. The default method is Kemenag.
\n| Key | \nMethod | \nRegion | \n
|---|---|---|
mwl | \nMuslim World League | \nEurope, Far East | \n
makkah | \nUmm al-Qura, Makkah | \nArabian Peninsula | \n
isna | \nISNA | \nNorth America | \n
egypt | \nEgyptian General Authority | \nAfrica, Middle East | \n
karachi | \nUniv. Islamic Sciences, Karachi | \nPakistan, India, Bangladesh | \n
turkey | \nDiyanet, Turkey | \nTurkey | \n
singapore | \nMUIS, Singapore | \nSingapore | \n
jakim | \nJAKIM, Malaysia | \nMalaysia | \n
kemenag | \nKEMENAG, Indonesia | \nIndonesia (default) | \n
france | \nUOIF, France | \nFrance | \n
russia | \nSpiritual Admin., Russia | \nRussia | \n
dubai | \nGAIAE, Dubai | \nUAE | \n
qatar | \nMin. of Awqaf, Qatar | \nQatar | \n
kuwait | \nMin. of Awqaf, Kuwait | \nKuwait | \n
jordan | \nMin. of Awqaf, Jordan | \nJordan | \n
gulf | \nGulf Region | \nGulf states | \n
tunisia | \nMin. of Religious Affairs | \nTunisia | \n
algeria | \nMin. of Religious Affairs | \nAlgeria | \n
morocco | \nMin. of Habous, Morocco | \nMorocco | \n
portugal | \nComunidade Islamica de Lisboa | \nPortugal | \n
moonsighting | \nMoonsighting Committee | \nWorldwide | \n
You can also use a custom method by passing CALC_CUSTOM with your own angles.
Each method defines a set of parameters:
\nFor detailed mathematical formulas and worked examples, see docs/KEMENAG_METHOD.md.
This is a single-header library, so you can simply include prayertimes.h in your project.
# Compile the CLI tool\ngcc -O3 -o libmuslim main.c -lm\n\n# Run example (Bekasi, November 21, 2025)\n./libmuslim 2025 11 21 -6.2851291 106.9814968 7.0\n#include \"prayertimes.h\"\n\n// Use the default method (Kemenag)\nconst MethodParams *params = method_params_get(CALC_KEMENAG);\n\nstruct PrayerTimes times = calculate_prayer_times(\n 2025, // year\n 11, // month\n 21, // day\n -6.2851291, // latitude (negative = South)\n 106.9814968, // longitude (positive = East)\n 7.0, // timezone offset (WIB = UTC+7)\n params // calculation method\n);\n\nchar buffer[16];\nformat_time_hm(times.fajr, buffer, sizeof(buffer));\nprintf(\"Fajr: %s\\n\", buffer);\n// Use MWL method\nconst MethodParams *mwl = method_params_get(CALC_MWL);\nstruct PrayerTimes times = calculate_prayer_times(2025, 11, 21, 51.5074, -0.1278, 0.0, mwl);\n\n// Look up method by string key\nCalcMethod method = method_from_string(\"isna\");\nconst MethodParams *params = method_params_get(method);\nmt_days_from_civil converts a civil (proleptic Gregorian) date to a day number counted from 1970-01-01, and mt_civil_from_days converts it back. They let you walk a range of dates without touching struct tm or mktime, so there are no DST or local-time hazards in the loop itself.
// Print Fajr for every day in July 2026\nlong start = mt_days_from_civil(2026, 7, 1);\nlong end = mt_days_from_civil(2026, 7, 31);\n\nfor (long serial = start; serial <= end; serial++) {\n int y, m, d;\n mt_civil_from_days(serial, &y, &m, &d);\n\n struct PrayerTimes t = calculate_prayer_times(y, m, d, -6.2851291, 106.9814968, 7.0, params);\n\n char buffer[16];\n format_time_hm(t.fajr, buffer, sizeof(buffer));\n printf(\"%04d-%02d-%02d Fajr: %s\\n\", y, m, d, buffer);\n}\nBoth are static inline, so they carry no link-time cost and need no PRAYERTIMES_IMPLEMENTATION definition. Day numbers are signed, and dates before 1970 are negative.
\n\nNOTE:
\nprayertimes.hdoes not handle Daylight Saving Time. The\ntimezoneargument is a fixed numeric UTC offset in hours, and the library\nuses it exactly as given โ it has no notion of dates, zones, or DST rules.\nThis is deliberate: DST is a political rule, not an astronomical one, and\nkeeping it out leavesprayertimes.ha pure, dependency-free (only<math.h>)\nsingle header. For a DST-active date you must pass the DST-adjusted offset\n(e.g.1.0for London in summer,0.0in winter).
If you want libmuslim to compute the correct offset for you, use the optional\ncompanion header timezone.h. It resolves an IANA zone name and\ndate to a UTC offset with DST applied, using the host operating system's\ntimezone database:
#define MUSLIM_TIMEZONE_IMPLEMENTATION // in exactly ONE translation unit\n#include \"timezone.h\"\n#include \"prayertimes.h\"\n\nchar zone[64];\nget_system_timezone(zone, sizeof(zone)); // e.g. \"Europe/London\"\ndouble tz = parse_timezone_offset(zone, time(NULL)); // DST already applied\n\nconst MethodParams *mwl = method_params_get(CALC_MWL);\nstruct PrayerTimes times = calculate_prayer_times(2026, 7, 15, 51.5074, -0.1278, tz, mwl);\nUnlike prayertimes.h, timezone.h touches the OS (POSIX tzset/tm_gmtoff\nor the Win32 timezone APIs), so it is optional โ include it only if you\nwant this resolution done for you. On a platform without a timezone database,\nkeep supplying the offset yourself.
./libmuslim <year> <month> <day> <latitude> <longitude> <timezone>\nExample output:
\nFajr = 04:05\nSunrise = 05:22\nDhuha = 05:50\nDhuhr = 11:41\nAsr = 15:04\nMaghrib = 17:54\nIsha = 19:07\nThe calculations have been verified against official data sources and match within ยฑ1-2 minute accuracy depending on the method. See the worked examples in docs/KEMENAG_METHOD.md for detailed verification.
Copyright 2025 Rizki Rakasiwi.\n\nLicensed under the Apache License, Version 2.0 (the \"License\");\nyou may not use this file except in compliance with the License.\nYou may obtain a copy of the License at\n\n http://www.apache.org/licenses/LICENSE-2.0\n\nUnless required by applicable law or agreed to in writing, software\ndistributed under the License is distributed on an \"AS IS\" BASIS,\nWITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\nSee the License for the specific language governing permissions and\nlimitations under the License.\nContributions are welcome! Please ensure any changes to calculation methods are verified against official data sources.
\n" }, { "fullName": "AdityaZxxx/sheltermark", From a692be868c80559f30f96b7a003487d0d9d9d179 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Wed, 22 Jul 2026 09:43:12 +0000 Subject: [PATCH 16/25] Sync content data --- src/data/projects.json | 46 +++++++++++++++++++++--------------------- 1 file changed, 23 insertions(+), 23 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index 71f6703..16c3c12 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -129,7 +129,7 @@ "opensid", "sistem-informasi-desa" ], - "updatedAt": "2026-07-21T17:03:15Z", + "updatedAt": "2026-07-22T07:59:43Z", "pushedAt": "2026-07-18T13:09:40Z", "latestRelease": { "name": "Rilis v2607.0.0", @@ -142,7 +142,7 @@ "createdAt": "2016-05-21T10:55:38Z", "openIssues": 353, "openPullRequests": 7, - "subscribers": 110, + "subscribers": 111, "communityHealth": 50, "readmeHtml": "
OpenSID adalah Sistem Informasi Desa (SID) yang dikembangkan secara terbuka dan kolaboratif oleh komunitas yang peduli dengan SID.
\nSID diharapkan dapat membantu pemerintah desa dalam beberapa hal berikut:
\n\n\nOpenSID bertujuan agar sebanyak mungkin desa di Indonesia dapat menerapkan sistem informasi untuk memajukan desa masing-masing..
\n
Strategi pengembangan OpenSID adalah untuk:
\nOpenSID dikelola di GitHub untuk:
\nPanduan pemasangan dan penggunaan OpenSID tersedia di Panduan OpenSID.
\nSistem Informasi Desa (SID) pertama kali dikembangkan oleh Combine Resource Institution sejak tahun 2009. Hak cipta awal dimiliki oleh Combine Resource Institution (http://lumbungkomunitas.net/).
\nSistem ini dikelola berdasarkan lisensi GNU General Public License Versi 3 (http://www.gnu.org/licenses/gpl.html).
\nVersi GitHub ini dikembangkan sejak Mei 2016, gratis dan bebas dimanfaatkan serta dikembangkan oleh semua desa. Hak Cipta OpenSID kini dipegang oleh Perkumpulan Desa Digital Terbuka (https://opendesa.id), sebuah lembaga hukum yang dibentuk khusus untuk mengelola OpenSID.
\nDemo Versi Publik OpenSID dapat dicoba di: https://demosid.opendesa.id
Modul admin: https://demosid.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Demo Versi Premium OpenSID dapat dicoba di: https://berputar.opendesa.id
Modul admin: https://berputar.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Bergabunglah dengan Forum Pengguna dan Pegiat OpenSID di Facebook atau di Telegram.
Forum ini bersifat informal, sebagai wadah berbagi informasi dan saling membantu dalam menggunakan dan mengembangkan OpenSID.
Laporkan masalah, usulan, atau permintaan pengembangan OpenSID melalui issue GitHub.
Kontribusi dari komunitas SID sangat dihargai, baik untuk dokumentasi di Wiki OpenSID maupun untuk source code di repo utama.
\n
Peduli OpenSID dan misi membangun desa? Dukung OpenSID di sini.
\n\nAtau donasi langsung melalui rekening bank. Info lengkap di sini.
\n
Apakah desa, lembaga, atau perusahaan Anda mendapat manfaat dari OpenSID?
Bantu kami mengembangkan OpenSID dengan menjadi sponsor.
Logo sponsor Anda akan tampil di sini dengan tautan ke situs Anda.
Berikut adalah para kontributor luar biasa yang telah membantu mengembangkan OpenSID:
\nA modern, project-aware terminal manager built with Tauri
\nTermul treats workspaces as first-class citizens, allowing you to organize terminals by project with persistent sessions, snapshots, and a clean tabbed interface.
\n
Getting Started ยท Features ยท Documentation ยท Contributing ยท Report Bug ยท Request Feature
\n| Feature | \nDescription | \n
|---|---|
| Project-Based Workspaces | \nOrganize terminals by project with dedicated workspace directories, separate state, and per-project configuration | \n
| Pane-Based Split Layout | \nSplit your workspace into resizable panes and arrange terminals, editors, and browser tabs side by side | \n
| Tabbed Interface | \nWindows Terminal-style tab bar with drag-and-drop reordering, rename, and context menu | \n
| Multiple Shell Support | \nAuto-detects PowerShell, CMD, Git Bash, WSL, fish, zsh, and more; switch shells per tab | \n
| Feature | \nDescription | \n
|---|---|
| Code Editor | \nBuilt-in code editor with syntax highlighting, file buffers, dirty-state tracking, and save/reload | \n
| Markdown Editor | \nRich markdown editing powered by BlockNote with live preview, table of contents, and heading navigation | \n
| Mermaid Diagrams | \nRender Mermaid diagrams inline within your markdown documents | \n
| File Explorer | \nFull file tree with create, rename, delete, clipboard operations, drag-and-drop, and context menus | \n
| File Watching | \nLive file watching for real-time updates as files change on disk | \n
| Feature | \nDescription | \n
|---|---|
| Embedded Browser Tabs | \nBrowse the web directly inside your workspace using child webview tabs โ no app switching | \n
| Annotation Workflow | \nCapture browser states, annotate with severity and intent labels, review, and export | \n
| Annotation Export | \nPackage annotations with metadata into structured export formats | \n
| Feature | \nDescription | \n
|---|---|
| Command Palette | \nGlobal command launcher (Ctrl+K / Ctrl+Shift+P) for project switching, workspace actions, and more | \n
| Command History | \nPer-project and aggregate command history viewer with search | \n
| Keyboard Shortcuts | \nFully customizable shortcut bindings for every action | \n
| Git Integration | \nStatus bar shows current branch, working directory, git status, and exit code | \n
| Custom Title Bar | \nDesktop-native title bar with window controls, sidebar toggles, and settings navigation | \n
| Feature | \nDescription | \n
|---|---|
| Auto-Updater | \nBuilt-in update infrastructure with signed artifacts โ get notified and update without leaving the app | \n
| State Management | \nZustand-powered reactive stores for projects, terminals, workspace layout, editor buffers, browser sessions, and settings | \n
| Configurable Settings | \nTerminal and UI preferences, color picker, theme customization, and shell configuration | \n
| Cross-Platform | \nWorks on Windows, macOS, and Linux with native platform packaging | \n
| Error Boundaries | \nGraceful error handling with runtime error boundaries and user-friendly fallback UI | \n
| Domain | \nKey Components | \nZustand Store | \n
|---|---|---|
| Workspace | \nWorkspaceLayout, PaneRenderer, PaneContent, WorkspaceTabBar | \nworkspace-store | \n
| Terminal | \nConnectedTerminal, XTerminal, TerminalSearchBar, ActivityIndicator | \nterminal-store | \n
| Editor | \nEditorPanel, CodeEditor, MarkdownEditor, EditorToolbar, MermaidBlock | \neditor-store | \n
| Browser | \nBrowserPanel, BrowserControls, AnnotationPanel, AnnotationExportModal | \nbrowser-session-store, annotation-store | \n
| File Explorer | \nFileExplorer, FileTreeNode, FileTreeContextMenu | \nโ | \n
| Snapshots | \nCreateSnapshotModal, RestoreSnapshotModal, DeleteSnapshotModal | \nsnapshot-store | \n
| Projects | \nProjectSidebar, NewProjectModal | \nproject-store | \n
| Settings | \nShortcutRecorder, ColorPickerPopover, ContextBarSettingsPopover | \napp-settings-store, context-bar-settings-store | \n
| Updates | \nUpdateAvailableToast, UpdateReadyModal | \nupdater-store | \n
| Shared | \nCommandPalette, ContextMenu, ConfirmDialog, ShellSelector, ErrorBoundary | \nโ | \n
brew tap gnoviawan/termul\nbrew install --cask termul\ncurl -fsSL https://raw.githubusercontent.com/gnoviawan/termul/main/scripts/install.sh | bash\nWindows users should install the .exe or .msi from GitHub Releases. Manual DMG downloads in a browser may still hit Gatekeeper, so macOS users should prefer Homebrew or curl.
| Dependency | \nVersion | \nNotes | \n
|---|---|---|
| Bun | \n1.3+ | \nJavaScript runtime and package manager | \n
| Rust | \nLatest stable | \nRequired for Tauri builds | \n
xcode-select --install\nsudo apt update\nsudo apt install libwebkit2gtk-4.1-dev \\\n build-essential curl wget file \\\n libxdo-dev libssl-dev \\\n libayatana-appindicator3-dev \\\n librsvg2-dev patchelf\nsudo dnf install webkit2gtk4.1-devel \\\n gcc gcc-c++ libopenssl-devel \\\n appindicator-devel librsvg2-devel \\\n patchelf\ncurl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh\nrustc --version && cargo --version\n# Clone the repository\ngit clone https://github.com/gnoviawan/termul.git\ncd termul\n\n# Install dependencies\nbun install\n\n# Launch in development mode\nbun run dev\nThis repository also includes a standalone Vite landing page under landing/.
# Install landing page dependencies (from landing/)\ncd landing && bun install\n\n# Start the landing page dev server\nbun run landing:dev\n\n# Lint the landing page\nbun run landing:lint\n\n# Build the landing page for production\nbun run landing:build\n# Build for your current platform\nbun run build\n\n# Platform-specific builds\nbun run build:tauri:win # Windows (x64)\nbun run build:tauri:mac-arm # macOS (Apple Silicon)\nbun run build:tauri:mac-x64 # macOS (Intel)\nbun run build:tauri:linux # Linux (x64)\n\n# Debug build (faster compilation, larger binary)\nbun run build:tauri:debug\nBuild output: src-tauri/target/release/bundle/
| Action | \nHow | \n
|---|---|
| New terminal | \nClick + next to tabs | \n
| Select specific shell | \nClick the dropdown arrow | \n
| Reorder tabs | \nDrag and drop | \n
| Rename tab | \nDouble-click the tab | \n
| Context menu | \nRight-click (rename, close, kill process) | \n
| Action | \nDefault Shortcut | \n
|---|---|
| New Terminal | \nCtrl+T | \n
| Next Tab | \nCtrl+PageDown | \n
| Previous Tab | \nCtrl+PageUp | \n
| Command Palette | \nCtrl+K / Ctrl+Shift+P | \n
\n\nShortcuts are customizable in Settings. On Tauri/WebView2, browser-reserved shortcuts such as
\nCtrl+Tabare not used as defaults because they are not reliably interceptable.
| Layer | \nTechnology | \n
|---|---|
| Desktop Runtime | \nTauri 2.0 | \n
| Backend | \nRust | \n
| UI Framework | \nReact 18 | \n
| Type System | \nTypeScript | \n
| Build Tool | \nVite | \n
| Styling | \nTailwind CSS + shadcn/ui | \n
| State Management | \nZustand | \n
| Terminal Emulation | \ntauri-pty + xterm.js | \n
| Animations | \nFramer Motion | \n
| Plugin | \nPurpose | \n
|---|---|
@tauri-apps/plugin-fs | \nFilesystem access | \n
@tauri-apps/plugin-store | \nConfiguration persistence | \n
@tauri-apps/plugin-os | \nOS information | \n
@tauri-apps/plugin-dialog | \nNative dialogs | \n
@tauri-apps/plugin-clipboard-manager | \nClipboard operations | \n
@tauri-apps/plugin-updater | \nAutomatic updates | \n
@tauri-apps/plugin-process | \nProcess management | \n
src/\nโโโ renderer/ # React frontend\nโ โโโ components/ # UI components\nโ โโโ hooks/ # Custom React hooks\nโ โโโ lib/ # Runtime adapters & desktop integration\nโ โโโ pages/ # Page components\nโ โโโ stores/ # Zustand stores\nโโโ shared/ # Shared types (main/renderer)\nsrc-tauri/ # Rust backend, config & bundling\ndocs/electron-old/ # Archived Electron docs & migration history\nThe renderer uses an adapter/service layer to keep desktop integrations isolated from UI code:
\nsrc/renderer/lib/\nโโโ tauri-*.ts # Tauri-native integrations\nโโโ *.ts # Runtime-safe facades & helpers\nโโโ __tests__/ # Regression & parity coverage\nbun run dev # Development mode with hot reload\nbun run test # Run tests\nbun run test:watch # Tests in watch mode\nbun run typecheck # Type checking\nbun run lint # Linting\nbun run tauri <command> # Direct Tauri CLI access\nssh-profiles.json.StrictHostKeyChecking=accept-new; do not override UserKnownHostsFile to /dev/null/NUL because that disables persistent host-key verification.ssh2 channel_direct_tcpip over the active SSH session; remote/reverse forwarding is not supported by the MVP command path yet.ssh2 intentionally stays on system OpenSSL for now. Enabling vendored-openssl forces a local OpenSSL source build that can fail in Windows/MSYS environments without a complete Perl module setup.Contributions are welcome! Please read the Contributing Guide for details on our code of conduct and the process for submitting pull requests.
\nThis project is licensed under the MIT License โ see the LICENSE file for details.
\nBuilt with โค๏ธ by gnoviawan
\nMuslimtify keeps you consistent with your daily prayers by delivering accurate prayer times and timely desktop notifications. Designed for Linux and Windows, it automatically calculates prayer schedules and reminds you 30, 15, and 5 minutes before the Adhan โ or at your own custom intervals โ and when it's time to pray. All calculations run locally, requiring no internet connection or external services.
\nMuslimtify supports 21 international calculation methods including MWL, ISNA, Umm al-Qura (Makkah), Egyptian General Authority, Kemenag (Indonesia), JAKIM (Malaysia), Diyanet (Turkey), and more. The default method is Kemenag. With persistent configuration and minimal setup, Muslimtify integrates seamlessly into your daily routine without interrupting your workflow.
\n\n\n[!Note]\nPrayer time calculations are powered by libmuslim, a portable library extracted from this project to enable a more flexible and reusable ecosystem for Muslim developers.
\n
| Linux | \nWindows | \n
|---|---|
\n\nRoadmap
\n\n
\n- \n
Merge commandlocation autoandmethod autointo (only)config autoand optimize auto detection per (249) country- \n
Refactor from timer-driven into a portable long-running loop- \n
Add custom adzan sound notifications- \n
Re-design command-line (BREAKING CHANGES)- Add read lat/long from user GPS
\n- Add GUI (see branch gui to see a progress)
\n- Distribute to Flatpak
\n- MacOS support (if devices is available)
\n- Wearable Device support
\n- Embedded Device Support
\n
\n\n[!Important]\nThis project is available for Linux and Windows users, but not yet for Mac users because we need a Mac device to make Muslimtify run on macOS. We are looking for brothers and sisters who have a Mac and experience in low-level C programming to contribute to the project and help bring Muslimtify to macOS. Alternatively, you can support us via GitHub Sponsors in the sponsor section.
\n
Every release ships ready-to-run binaries for Linux and Windows on the\nReleases page.
\nLinux (x86_64 or aarch64) โ the binaries are dynamically linked, so\ninstall the runtime libraries first, then extract and install:
# Ubuntu/Debian\nsudo apt install libnotify4 libcurl4\n# Fedora/RHEL\nsudo dnf install libnotify libcurl\n# Arch\nsudo pacman -S libnotify curl\n\ntar xzf muslimtify-<version>-linux-<arch>.tar.gz\nsudo cp -r muslimtify-<version>-linux-<arch>/{bin,lib,share} /usr/local/\nmuslimtify daemon install\nWindows (x64 or arm64) โ download and run the matching installer:
muslimtify-<version>-setup-x64.exe # Intel/AMD\nmuslimtify-<version>-setup-arm64.exe # ARM\nVerify any download against the published checksums:
\nsha256sum -c SHA256SUMS\nyay -S muslimtify\nsudo dnf copr enable rizukirr/muslimtify\nsudo dnf install muslimtify\nsudo add-apt-repository ppa:rizukirr/muslimtify\nsudo apt update\nsudo apt install muslimtify\nInstall dependencies:
\n# Ubuntu/Debian\nsudo apt install git build-essential cmake pkg-config libnotify-dev libcurl4-openssl-dev\n\n# Fedora/RHEL\nsudo dnf install git gcc cmake pkgconfig libnotify-devel libcurl-devel\n\n# Arch Linux\nsudo pacman -S git base-devel cmake pkgconfig libnotify curl\nClone, install, and enable background checks:
\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\nsudo ./install.sh\nmuslimtify daemon install\nwinget install muslimtify\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\n.\\install.ps1\nmuslimtify daemon install\nTo remove the Windows install later, run .\\uninstall.ps1.
If you prefer building manually first:
\ncmake -S . -B build\ncmake --build build --config Release\ncmake --install build --config Release\nmuslimtify daemon install\nRun muslimtify daemon status to check if Muslimtify is registered with systemd. If no status is found, run muslimtify daemon install to register the service and ensure it runs as expected.
Muslimtify automatically selects the standard prayer time calculation method based on your country and location. Run muslimtify to verify that your configuration is correct. If the automatic selection does not meet your needs, you can set it manually using muslimtify method <key-method>. A full list of available methods is documented here.
Muslimtify can be configured with CLI commands or by editing config.json\nmanually.
Config paths:
\n~/.config/muslimtify/config.json~/.cache/muslimtify%APPDATA%\\muslimtify\\config.json%LOCALAPPDATA%\\muslimtifyCommon setup commands:
\nmuslimtify location set --auto # detect location from IP\nmuslimtify location set --auto --city=Mansoura # auto-detect but use your own city label\nmuslimtify method --auto # select method from the detected country\nmuslimtify location set --lat=-6.175 --long=106.82 # set location manually (uses system timezone)\nmuslimtify location set --timezone=Asia/Jakarta # override timezone\nmuslimtify location set --city=Jakarta # add a city label\nmuslimtify location set --refresh-interval=21600 # re-check location every 6h (0=off, min 3600)\nmuslimtify method --list # list all available calculation methods\nmuslimtify method mwl # set calculation method\nmuslimtify madzhab hanafi # set madzhab (shafi/hanafi)\nmuslimtify notification --reminder --all 30 15 5 # set every prayer's reminders (minutes before adhan)\nmuslimtify notification --reminder fajr 30 15 5 # set reminders for a single prayer\nmuslimtify notification # show current notification settings\nmuslimtify location # show current location\nResetting the configuration is done by deleting config.json. Muslimtify falls\nback to built-in defaults when the file is missing, and rewrites it the next\ntime you change a setting. Validation runs automatically every time the config\nis loaded.
Muslimtify supports the following calculation methods:
\n| Key | \nMethod | \nRegion | \n
|---|---|---|
mwl | \nMuslim World League | \nEurope, Far East | \n
makkah | \nUmm al-Qura, Makkah | \nArabian Peninsula | \n
isna | \nISNA | \nNorth America | \n
egypt | \nEgyptian General Authority | \nAfrica, Middle East | \n
karachi | \nUniv. Islamic Sciences, Karachi | \nPakistan, India, Bangladesh | \n
turkey | \nDiyanet, Turkey | \nTurkey | \n
singapore | \nMUIS, Singapore | \nSingapore | \n
jakim | \nJAKIM, Malaysia | \nMalaysia | \n
kemenag | \nKEMENAG, Indonesia | \nIndonesia (default) | \n
france | \nUOIF, France | \nFrance | \n
russia | \nSpiritual Admin., Russia | \nRussia | \n
dubai | \nGAIAE, Dubai | \nUAE | \n
qatar | \nMin. of Awqaf, Qatar | \nQatar | \n
kuwait | \nMin. of Awqaf, Kuwait | \nKuwait | \n
jordan | \nMin. of Awqaf, Jordan | \nJordan | \n
gulf | \nGulf Region | \nGulf states | \n
tunisia | \nMin. of Religious Affairs | \nTunisia | \n
algeria | \nMin. of Religious Affairs | \nAlgeria | \n
morocco | \nMin. of Habous, Morocco | \nMorocco | \n
portugal | \nComunidade Islamica de Lisboa | \nPortugal | \n
moonsighting | \nMoonsighting Committee | \nWorldwide | \n
You can also use a custom method by setting \"method\": \"custom\" in config.json with your own fajr_angle and isha_angle values.
Manual JSON editing is useful when you want precise control over enabled\nprayers, reminder offsets, notification settings, or location data.
\n\nDefault config.json{\n \"location\": {\n \"latitude\": 0.0,\n \"longitude\": 0.0,\n \"timezone\": \"UTC\",\n \"timezone_offset\": 0.0,\n \"auto_detect\": true,\n \"city\": \"\",\n \"country\": \"\"\n },\n \"prayers\": {\n \"fajr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"sunrise\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuha\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuhr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"asr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"maghrib\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"isha\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n }\n },\n \"notification\": {\n \"timeout\": 5000,\n \"urgency\": \"critical\",\n \"sound\": \"adhan\",\n \"sound_alarm\": \"alarm\",\n \"sound_reminder\": \"reminder\",\n \"icon\": \"muslimtify\"\n },\n \"calculation\": {\n \"method\": \"kemenag\",\n \"madhab\": \"shafi\"\n }\n}\nmuslimtify daemon status to confirm the background service is running.notify-send \"Test\" \"Hello\".muslimtify location set --auto again.muslimtify location set --lat=<latitude> --long=<longitude>.\nIf the host machine is in a different region than the coordinates, override\nthe timezone with --timezone=<iana>, e.g.\nmuslimtify location set --lat=-6.21 --long=106.84 --timezone=Asia/Jakarta.ipinfo.io if auto detection keeps failing.Contributions are welcome. See CONTRIBUTING.md for workflow,\nstyle, and testing guidance.
\nMuslimtify is released under the MIT License. See the repository license files\nfor details.
\nMuslimtify keeps you consistent with your daily prayers by delivering accurate prayer times and timely desktop notifications. Designed for Linux and Windows, it automatically calculates prayer schedules and reminds you 30, 15, and 5 minutes before the Adhan, or at your own custom intervals, and when it's time to pray. Every prayer time is calculated locally on your machine, with no accounts and no tracking. The only thing that touches the network is location detection via ipinfo.io, and even that is optional: set your coordinates manually, or read them from a GPS receiver with location gps on, and Muslimtify makes no network request at all.
Muslimtify supports 21 international calculation methods including MWL, ISNA, Umm al-Qura (Makkah), Egyptian General Authority, Kemenag (Indonesia), JAKIM (Malaysia), Diyanet (Turkey), and more. The default method is Kemenag. With persistent configuration and minimal setup, Muslimtify integrates seamlessly into your daily routine without interrupting your workflow.
\n\n\n[!Note]\nPrayer time calculations are powered by libmuslim, a portable library extracted from this project to enable a more flexible and reusable ecosystem for Muslim developers.
\n
| Linux | \nWindows | \n
|---|---|
\n\nRoadmap
\n\n
\n- \n
Merge commandlocation autoandmethod autointo (only)config autoand optimize auto detection per (249) country- \n
Refactor from timer-driven into a portable long-running loop- \n
Add custom adzan sound notifications- \n
Re-design command-line (BREAKING CHANGES)- \n
Add read lat/long from user GPS- Add GUI (see branch gui to see a progress)
\n- Distribute to Flatpak
\n- MacOS support (if devices is available)
\n- Wearable Device support
\n- Embedded Device Support
\n
\n\n[!Important]\nThis project is available for Linux and Windows users, but not yet for Mac users because we need a Mac device to make Muslimtify run on macOS. We are looking for brothers and sisters who have a Mac and experience in low-level C programming to contribute to the project and help bring Muslimtify to macOS. Alternatively, you can support us via GitHub Sponsors in the sponsor section.
\n
Every release ships ready-to-run binaries for Linux and Windows on the\nReleases page.
\nLinux (x86_64 or aarch64): the binaries are dynamically linked, so\ninstall the runtime libraries first, then extract and install:
# Ubuntu/Debian\nsudo apt install libnotify4 libcurl4\n# Fedora/RHEL\nsudo dnf install libnotify libcurl\n# Arch\nsudo pacman -S libnotify curl\n\ntar xzf muslimtify-<version>-linux-<arch>.tar.gz\nsudo cp -r muslimtify-<version>-linux-<arch>/{bin,lib,share} /usr/local/\nmuslimtify daemon install\nWindows (x64 or arm64): download and run the matching installer:
muslimtify-<version>-setup-x64.exe # Intel/AMD\nmuslimtify-<version>-setup-arm64.exe # ARM\nVerify any download against the published checksums:
\nsha256sum -c SHA256SUMS\nyay -S muslimtify\nsudo dnf copr enable rizukirr/muslimtify\nsudo dnf install muslimtify\nsudo add-apt-repository ppa:rizukirr/muslimtify\nsudo apt update\nsudo apt install muslimtify\nInstall dependencies:
\n# Ubuntu/Debian\nsudo apt install git build-essential cmake pkg-config libnotify-dev libcurl4-openssl-dev\n\n# Fedora/RHEL\nsudo dnf install git gcc cmake pkgconfig libnotify-devel libcurl-devel\n\n# Arch Linux\nsudo pacman -S git base-devel cmake pkgconfig libnotify curl\nGPS is optional and needs nothing at build time. Install gpsd only if you want\nMuslimtify to read coordinates from a local receiver:
# Ubuntu/Debian\nsudo apt install gpsd\n# Fedora/RHEL\nsudo dnf install gpsd\n# Arch Linux\nsudo pacman -S gpsd\nClone, install, and enable background checks:
\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\nsudo ./install.sh\nmuslimtify daemon install\ninstall.sh compiles as the user who invoked sudo rather than as root, and\nrefuses to build from a source tree that is group- or world-writable, since\nanything planted there would otherwise run with root privileges. If it reports\nunsafe permissions, fix the listed paths so each is owned by root or by you and\nis not writable by others, then re-run.
winget install muslimtify\nBuilding on Windows requires MSVC. The build stops with an explicit message if\nanother compiler is used.
\ngit clone https://github.com/rizukirr/muslimtify.git\ncd muslimtify\n.\\install.ps1\nmuslimtify daemon install\nTo remove the Windows install later, run .\\uninstall.ps1.
If you prefer building manually first:
\ncmake -S . -B build\ncmake --build build --config Release\ncmake --install build --config Release\nmuslimtify daemon install\nRun muslimtify daemon status to check if Muslimtify is registered with systemd. If no status is found, run muslimtify daemon install to register the service and ensure it runs as expected.
Muslimtify automatically selects the standard prayer time calculation method based on your country and location. Run muslimtify to verify that your configuration is correct. If the automatic selection does not meet your needs, you can set it manually using muslimtify method <key-method>. A full list of available methods is documented here.
Muslimtify can be configured with CLI commands or by editing config.json\nmanually.
Config paths:
\n~/.config/muslimtify/config.json~/.cache/muslimtify%APPDATA%\\muslimtify\\config.json%LOCALAPPDATA%\\muslimtifyCommon setup commands:
\nmuslimtify location set --auto # detect location from IP\nmuslimtify location set --auto --city=Mansoura # auto-detect but use your own city label\nmuslimtify method --auto # select method from the detected country\nmuslimtify location set --lat=-6.175 --long=106.82 # set location manually (uses system timezone)\nmuslimtify location set --timezone=Asia/Jakarta # override timezone\nmuslimtify location set --city=Jakarta # add a city label\nmuslimtify location set --refresh-interval=21600 # re-check location every 6h (0=off, min 3600)\nmuslimtify location gps on # read coordinates from a local GPS receiver\nmuslimtify location gps off # go back to ipinfo network geolocation\nmuslimtify location gps # show whether GPS is enabled\nmuslimtify method --list # list all available calculation methods\nmuslimtify method mwl # set calculation method\nmuslimtify madzhab hanafi # set madzhab (shafi/hanafi)\nmuslimtify notification --reminder --all 30 15 5 # set every prayer's reminders (minutes before adhan)\nmuslimtify notification --reminder fajr 30 15 5 # set reminders for a single prayer\nmuslimtify notification # show current notification settings\nmuslimtify location # show current location\nGPS is off by default and maps to a single use_gps key in the location block\nof config.json. On Linux the coordinates come from a running gpsd, read over\na local socket on 127.0.0.1:2947, with no libgps build dependency. On Windows\nthey come from the WinRT Geolocator, which needs location access enabled in\nSettings. location gps on probes the receiver first and refuses to enable if\nnone is reachable, so a missing daemon fails immediately rather than degrading\nsilently later. Whenever GPS has no fix, ipinfo.io is used instead. GPS\nsupplies coordinates only, so the timezone is still taken from the host system.
The timezone itself is validated when you set it. A name the system cannot\nresolve is rejected outright rather than saved and silently treated as UTC, and\nthe offset used to compute prayer times is derived from the IANA name for the\ndate being calculated, so daylight saving is handled automatically.
\nResetting the configuration is done by deleting config.json. Muslimtify falls\nback to built-in defaults when the file is missing, and rewrites it the next\ntime you change a setting. Validation runs automatically every time the config\nis loaded.
Muslimtify supports the following calculation methods:
\n| Key | \nMethod | \nRegion | \n
|---|---|---|
mwl | \nMuslim World League | \nEurope, Far East | \n
makkah | \nUmm al-Qura, Makkah | \nArabian Peninsula | \n
isna | \nISNA | \nNorth America | \n
egypt | \nEgyptian General Authority | \nAfrica, Middle East | \n
karachi | \nUniv. Islamic Sciences, Karachi | \nPakistan, India, Bangladesh | \n
turkey | \nDiyanet, Turkey | \nTurkey | \n
singapore | \nMUIS, Singapore | \nSingapore | \n
jakim | \nJAKIM, Malaysia | \nMalaysia | \n
kemenag | \nKEMENAG, Indonesia | \nIndonesia (default) | \n
france | \nUOIF, France | \nFrance | \n
russia | \nSpiritual Admin., Russia | \nRussia | \n
dubai | \nGAIAE, Dubai | \nUAE | \n
qatar | \nMin. of Awqaf, Qatar | \nQatar | \n
kuwait | \nMin. of Awqaf, Kuwait | \nKuwait | \n
jordan | \nMin. of Awqaf, Jordan | \nJordan | \n
gulf | \nGulf Region | \nGulf states | \n
tunisia | \nMin. of Religious Affairs | \nTunisia | \n
algeria | \nMin. of Religious Affairs | \nAlgeria | \n
morocco | \nMin. of Habous, Morocco | \nMorocco | \n
portugal | \nComunidade Islamica de Lisboa | \nPortugal | \n
moonsighting | \nMoonsighting Committee | \nWorldwide | \n
You can also use a custom method by setting \"method\": \"custom\" in config.json with your own fajr_angle and isha_angle values.
Manual JSON editing is useful when you want precise control over enabled\nprayers, reminder offsets, notification settings, or location data.
\n\nDefault config.json{\n \"location\": {\n \"latitude\": 0.0,\n \"longitude\": 0.0,\n \"timezone\": \"UTC\",\n \"timezone_offset\": 0.0,\n \"auto_detect\": true,\n \"city\": \"\",\n \"country\": \"\"\n },\n \"prayers\": {\n \"fajr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"sunrise\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuha\": {\n \"enabled\": false,\n \"adhan\": \"\",\n \"adhan_enabled\": false,\n \"reminders\": [],\n \"offset\": 0\n },\n \"dhuhr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"asr\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"maghrib\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n },\n \"isha\": {\n \"enabled\": true,\n \"adhan\": \"\",\n \"adhan_enabled\": true,\n \"reminders\": [30, 15, 5],\n \"offset\": 0\n }\n },\n \"notification\": {\n \"timeout\": 5000,\n \"urgency\": \"critical\",\n \"sound\": \"adhan\",\n \"sound_alarm\": \"alarm\",\n \"sound_reminder\": \"reminder\",\n \"icon\": \"muslimtify\"\n },\n \"calculation\": {\n \"method\": \"kemenag\",\n \"madhab\": \"shafi\"\n }\n}\nmuslimtify daemon status to confirm the background service is running.notify-send \"Test\" \"Hello\".muslimtify location set --auto again.muslimtify location set --lat=<latitude> --long=<longitude>.\nIf the host machine is in a different region than the coordinates, override\nthe timezone with --timezone=<iana>, e.g.\nmuslimtify location set --lat=-6.21 --long=106.84 --timezone=Asia/Jakarta.ipinfo.io if auto detection keeps failing.muslimtify location gps on probes the receiver before saving, so it refuses\nrather than enabling something that cannot work. The message names the missing\npiece:
cannot reach gpsd: install and start gpsd. Muslimtify reads it over a local\nsocket on 127.0.0.1:2947.no GPS device detected: gpsd is running but sees no hardware. Connect the\nreceiver and confirm gpsd picked it up.location access is turned off: on Windows, enable Settings > Privacy &\nsecurity > Location, then try again.Being told GPS is enabled with no fix yet is not an error. The setting is saved\nand ipinfo.io is used until the receiver locks on. If the daemon or device\nlater disappears, Muslimtify warns once and turns GPS off so it stops retrying\non every cycle. A denied permission does not turn it off, because granting\naccess in Settings is enough for the next attempt to succeed.
The name must resolve on this system, so check the spelling against the IANA\ndatabase, for example Asia/Jakarta or Europe/London. Zones that legitimately\nsit at UTC+0, such as Africa/Abidjan, are accepted.
This is almost always daylight saving, which is handled automatically only when\na valid IANA zone is saved. Run muslimtify location and check the timezone\nfield. The gmt field shows the offset in effect today rather than the one\nrecorded when you last set your location. If the zone is wrong or empty, set it\nwith muslimtify location set --timezone=<iana>.
Triggers missed within the previous 15 minutes still fire when the daemon\ncatches up. Anything older is dropped without firing, so resuming from a long\nsuspend does not replay a stack of stale Adhans.
\nContributions are welcome. See CONTRIBUTING.md for workflow,\nstyle, and testing guidance.
\nMuslimtify is released under the MIT License. See the repository license files\nfor details.
\n![\"Discord\"](\"https://img.shields.io/badge/Join-Discord-5865F2?logo=discord&logoColor=white\"){kind=icon}
![\"GitHub](\"https://img.shields.io/github/stars/jipraks/yt-short-clipper?style=social\")
![\"License\"](\"https://img.shields.io/badge/license-MIT-blue.svg\"){kind=icon}
\n
๐ฌ Automated YouTube to Short-Form Content Pipeline
\nTransform long-form YouTube videos (podcasts, interviews, vlogs) into engaging short-form content for TikTok, Instagram Reels, and YouTube Shorts โ powered by AI.
\nDownload the desktop app for your platform:
\n| Platform | \nDownload | \nNotes | \n
|---|---|---|
| Windows | \nLatest Release (.exe) | \nWindows 10+ | \n
| macOS | \nLatest Release (.dmg) | \nmacOS Catalina+, Apple Silicon & Intel | \n
Then follow the complete setup guide:
\nWhat you'll learn:
\nIf you want to contribute or run from source:
\n![\"Platform\"](\"https://img.shields.io/badge/platform-Windows%20%7C%20macOS-lightgrey\"){kind=icon}
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ\nโ YT-Short-Clipper โ\nโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค\nโ โ\nโ โโโโโโโโโโโโ โโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโโ โ\nโ โ YouTube โโโโโถโ Downloader โโโโโถโ Subtitle โ โ\nโ โ URL โ โ (yt-dlp) โ โ Parser โ โ\nโ โโโโโโโโโโโโ โโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโโ โ\nโ โ โ\nโ โผ โ\nโ โโโโโโโโโโโโโโโโโโโ โ\nโ โ Highlight Finder โ โ\nโ โ (GPT-4) โ โ\nโ โโโโโโโโโโโโโโโโโโโ โ\nโ โ โ\nโ โผ โ\nโ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ\nโ โ Video Processing โ โ\nโ โ โโโโโโโโโโโโโโ โโโโโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโโโ โ โ\nโ โ โ Clipper โโโถโ Portrait โโโถโ Hook Generator โ โ โ\nโ โ โ (FFmpeg) โ โ Converter โ โ (TTS + Overlay) โ โ โ\nโ โ โโโโโโโโโโโโโโ โ OpenCV / โ โโโโโโโโโโโโโโโโโโโโโโ โ โ\nโ โ โ MediaPipe โ โ โ โ\nโ โ โโโโโโโโโโโโโโ โผ โ โ\nโ โ โโโโโโโโโโโโโโโโโโ โ โ\nโ โ โCaption Generatorโ โ โ\nโ โ โ (Whisper) โ โ โ\nโ โ โโโโโโโโโโโโโโโโโโ โ โ\nโ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ\nโ โ โ\nโ โผ โ\nโ โโโโโโโโโโโโโโโโโโโ โ\nโ โ Output Clips โ โ\nโ โ + Metadata โ โ\nโ โโโโโโโโโโโโโโโโโโโ โ\nโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ\n| Dependency | \nVersion | \nPurpose | \n
|---|---|---|
| Python | \n3.10+ | \nRuntime | \n
| FFmpeg | \n4.4+ | \nVideo processing | \n
| yt-dlp | \nLatest | \nYouTube downloading | \n
| Deno | \n2.x | \nRequired by yt-dlp for some extractors | \n
See requirements.txt for the full list. Key dependencies:
\ncustomtkinter>=5.2.0\nopenai>=1.0.0\nopencv-python>=4.8.0\nnumpy>=1.24.0\nPillow>=10.0.0\nmediapipe>=0.10.0\nrequests>=2.31.0\nyt-dlp>=2026.3.17\ngoogle-generativeai>=0.7.0\ngoogle-api-python-client>=2.100.0\ngoogle-auth-oauthlib>=1.1.0\n\n\nNote: The app uses OpenAI Whisper API instead of local Whisper model.
\n
The app supports 10+ AI providers including:
\nSee GUIDE.md or PANDUAN.md for detailed API setup instructions.
\n\n\nNote: This section is for developers who want to run the app from source code. If you're a regular user, please follow the User Guide or Panduan Indonesia instead.
\n
git clone https://github.com/jipraks/yt-short-clipper.git\ncd yt-short-clipper\nWindows (using Chocolatey):
\nchoco install ffmpeg yt-dlp\nmacOS (using Homebrew):
\nbrew install ffmpeg yt-dlp\nUbuntu/Debian:
\nsudo apt update\nsudo apt install ffmpeg\npip install yt-dlp\npip install -r requirements.txt\npython app.py\nThe app will create a config.json file on first run where you can save your AI API keys and other settings.
yt-short-clipper/\nโโโ app.py # Main GUI application (entry point)\nโโโ clipper_core.py # Core processing logic (download, AI, video)\nโโโ version.py # Version info and update URL\nโโโ youtube_uploader.py # YouTube upload functionality\nโโโ tiktok_uploader.py # TikTok upload functionality\nโโโ requirements.txt # Python dependencies\nโโโ build.spec # PyInstaller build config (Windows)\nโโโ build_macos.spec # PyInstaller build config (macOS)\nโโโ build_web.spec # PyInstaller build config (Web version)\nโโโ components/ # Reusable UI widgets\nโ โโโ ai_provider_card.py # AI provider configuration card\nโ โโโ page_layout.py # Page layout components\nโ โโโ progress_step.py # Progress step indicator\nโโโ config/ # Configuration management\nโ โโโ ai_provider_config.py # AI provider definitions\nโ โโโ config_manager.py # Config file read/write\nโโโ dialogs/ # Modal dialogs\nโ โโโ model_selector.py # AI model search/select dialog\nโ โโโ repliz_upload.py # Repliz upload dialog\nโ โโโ terms_of_service.py # ToS dialog\nโ โโโ tiktok_upload.py # TikTok upload dialog\nโ โโโ youtube_upload.py # YouTube upload dialog\nโโโ pages/ # GUI pages\nโ โโโ browse_page.py # Browse output clips\nโ โโโ clipping_page.py # Clipping progress\nโ โโโ contact_page.py # Contact/feedback\nโ โโโ highlight_selection_page.py # Select highlights to process\nโ โโโ processing_page.py # Processing progress\nโ โโโ results_page.py # Results display\nโ โโโ session_browser_page.py # Browse previous sessions\nโ โโโ settings_page.py # Settings hub\nโ โโโ status_pages.py # API & Library status pages\nโ โโโ settings/ # Settings sub-pages\nโ โโโ ai_api_settings.py # AI API configuration\nโ โโโ ai_providers/ # Per-provider settings\nโ โโโ output_settings.py # Output directory settings\nโ โโโ performance_settings.py # GPU & performance\nโ โโโ watermark_settings.py # Watermark configuration\nโ โโโ ...\nโโโ utils/ # Utility modules\nโ โโโ dependency_manager.py # Auto-download FFmpeg, Deno\nโ โโโ gpu_detector.py # GPU detection & encoder selection\nโ โโโ helpers.py # Path helpers, platform detection\nโ โโโ logger.py # Logging utilities\nโโโ assets/ # App icons and images\nโ โโโ icon.png # App icon (PNG)\nโ โโโ icon.ico # App icon (Windows)\nโ โโโ icon.icns # App icon (macOS)\nโโโ web/ # Web UI (experimental)\n โโโ index.html\n โโโ app.js\n โโโ css/\n โโโ components/\noutput/\nโโโ 20240115-143001/ # Session folder (timestamp-based)\n โโโ master.mp4 # Final clip\n โโโ data.json # Metadata\nEach clip folder contains a data.json file with metadata:
{\n \"title\": \"๐ฅ Momen Kocak Saat Pembully Datang Minta Maaf\",\n \"hook_text\": \"Mantan pembully TIARA datang ke rumah minta endorse salad buah\",\n \"start_time\": \"00:15:23,000\",\n \"end_time\": \"00:17:05,000\",\n \"duration_seconds\": 102.0,\n \"has_hook\": true,\n \"has_captions\": true,\n \"youtube_title\": \"๐ฅ Momen Kocak Saat Pembully Datang Minta Maaf\",\n \"youtube_description\": \"Siapa sangka mantan pembully malah datang minta endorse! ๐ #podcast #viral #fyp\",\n \"youtube_tags\": [\"shorts\", \"viral\", \"podcast\"]\n}\nAll settings can be configured through the GUI Settings page (โ๏ธ button in the app).
\nFor complete setup instructions with screenshots, see:
\n\n| Parameter | \nDefault | \nDescription | \n
|---|---|---|
num_clips | \n5 | \nNumber of clips to generate | \n
min_duration | \n60s | \nMinimum clip duration | \n
max_duration | \n120s | \nMaximum clip duration | \n
target_duration | \n90s | \nIdeal clip duration | \n
temperature | \n1.0 | \nAI creativity (0.0-2.0) | \n
| Parameter | \nDefault | \nDescription | \n
|---|---|---|
output_resolution | \n1080x1920 | \nOutput video resolution | \n
min_frames_before_switch | \n210 | \nFrames before speaker switch (~7s at 30fps) | \n
switch_threshold | \n3.0 | \nMovement multiplier to trigger switch | \n
| Parameter | \nDefault | \nDescription | \n
|---|---|---|
language | \nid | \nTranscription language | \n
chunk_size | \n4 | \nWords per caption line | \n
| Parameter | \nDefault | \nDescription | \n
|---|---|---|
tts_voice | \nnova | \nOpenAI TTS voice (nova/shimmer/alloy) | \n
tts_speed | \n1.0 | \nSpeech speed | \n
max_words | \n15 | \nMaximum words in hook text | \n
tts_model | \ntts-1 | \nTTS model (tts-1 or tts-1-hd) | \n
The captions use CapCut-style formatting:
\nFont: Arial Black (platform-dependent fallback)\nSize: 65px\nColor: White (#FFFFFF)\nHighlight: Yellow (#00FFFF)\nOutline: 4px Black\nShadow: 2px\nPosition: Lower third (400px from bottom)\nEstimated OpenAI API costs per video (5 clips):
\n| Feature | \nModel | \nEst. Cost | \n
|---|---|---|
| Highlight Detection | \nGPT-4.1 | \n~$0.05-0.15 | \n
| TTS Voiceover | \nTTS-1 | \n~$0.01/clip | \n
| Captions | \nWhisper API | \n~$0.01/clip | \n
Total estimate: ~$0.10-0.25 per video (5 clips)
\nThe desktop app shows real-time token usage and cost estimation during processing.
\npip install -r requirements.txt\npip install pyinstaller\n\npyinstaller build.spec\n# Output: dist/YTShortClipper.exe\nRequires Python 3.10+ and create-dmg (brew install create-dmg).
pip install -r requirements.txt\npip install pyinstaller\n\n# Build .app bundle\npython -m PyInstaller build_macos.spec --clean --noconfirm\n\n# Create DMG (optional)\ncreate-dmg \\\n --volname \"YTShortClipper\" \\\n --volicon \"assets/icon.icns\" \\\n --window-size 600 400 \\\n --icon \"YTShortClipper.app\" 150 185 \\\n --app-drop-link 450 185 \\\n \"dist/YTShortClipper.dmg\" \\\n \"dist/YTShortClipper.app\"\nmacOS notes:
\n~/Library/Application Support/YTShortClipper/ (persists across app updates)Contributions are welcome! We greatly appreciate contributions from anyone.
\n# 1. Fork this repo (click the Fork button on GitHub)\n\n# 2. Clone your fork\ngit clone https://github.com/YOUR-USERNAME/yt-short-clipper.git\ncd yt-short-clipper\n\n# 3. Add upstream remote\ngit remote add upstream https://github.com/jipraks/yt-short-clipper.git\n\n# 4. Create a new branch\ngit checkout -b feature/your-new-feature\n\n# 5. Make changes, then commit\ngit add .\ngit commit -m \"feat: description of changes\"\n\n# 6. Push to your fork\ngit push origin feature/your-new-feature\n\n# 7. Create a Pull Request on GitHub\n| Type | \nDescription | \n
|---|---|
| ๐ Bug Report | \nReport bugs in the Issues tab | \n
| ๐ก Feature Request | \nRequest new features in Issues | \n
| ๐ Documentation | \nImprove docs, fix typos, add examples | \n
| ๐ง Code | \nFix bugs, add features, improve performance | \n
๐ Complete guide available in CONTRIBUTING.md - includes Git tutorial for beginners!
\nThis project is licensed under the MIT License - see the LICENSE file for details.
\nMade with โ by Aji Prakoso for content creators
\n| \n | \n |
|---|---|
| ๐ | \nn8n & Automation eCourse | \n
| ๐ธ | \n@jipraks on Instagram | \n
| ๐ฌ | \nAji Prakoso on YouTube | \n
| ๐ | \nAbout Aji Prakoso | \n
\n Noise-canceling context and long-term memory for your AI agent โ lossy, but always reversible, and it never fabricates a result. Stop paying Claude to read 10,000 lines of terminal noise.\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Always reversible ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Every other compressor asks you to trust that what it cut didn't matter. OMNI doesn't ask โ it guarantees, and each guarantee is backed by code you can read:
\n| Guarantee | \nHow | \nProof | \n
|---|---|---|
| Get the original back, byte-for-byte | \neverything cut is archived in a local SQLite RewindStore (SHA-256 โ content); the agent gets a hash and calls omni_retrieve | \nHow it works | \n
| Never fabricates a result | \na distiller that parsed no signal returns the raw output, never a green no errors / passed string | \n#143 | \n
| Failures are never masked | \na command that exits non-zero passes through verbatim | \n#120 | \n
| Structured data is never touched | \nJSON / YAML / NDJSON / CSV pass through byte-for-byte | \npipeline::format | \n
| Numbers are measured, not aspirational | \n1,810 real traces replayed on the release binary โ and 63.6% of calls net zero, which we publish too | \nBenchmarks | \n
That is the one thing a bigger compression number can't buy: you can always recover the original, and it will never lie to your agent.
\nProblem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
\n ![\"KeiRouter](\"https://raw.githubusercontent.com/mydisha/keirouter/main/keirouter-logo.png\"){kind=logo}
\n
\n Your friendly, blazing-fast, self-hostable AI gateway.\n
\n ![\"CI\"](\"https://github.com/mydisha/keirouter/actions/workflows/ci.yml/badge.svg\"){kind=icon}
\n ![\"Go](\"https://img.shields.io/badge/Go-1.24+-00ADD8?logo=go&logoColor=white\"){kind=icon}
![\"Docker](\"https://img.shields.io/badge/Docker-GHCR-2496ED?logo=docker&logoColor=white\"){kind=icon}
![\"License:](\"https://img.shields.io/badge/License-MIT-yellow.svg\"){kind=icon}
\n
\n ![\"KeiRouter](\"https://raw.githubusercontent.com/mydisha/keirouter/main/assets/keirouter-social-banner.png\")
\n
You use AI coding tools โ Claude Code, Cursor, Cline, or honestly anything that talks to OpenAI or Anthropic. And you know the headaches: a drawer full of API keys, rate limits that hit at the worst time, and a token bill that keeps creeping up.
\nKeiRouter is the smart middleman that makes all of that someone else's problem. Point your tools at one local endpoint and let it handle the boring stuff โ routing each request to the right model, failing over the moment a provider taps out, caching repeat questions, and squeezing oversized prompts down before they ever cost you a token.
\nOh, and it's written in Go. So it sips memory (~20 MB), boots instantly, ships as a single binary, and comes with a dashboard that's actually nice to look at. โจ
\n\n\nHeads up: KeiRouter is under active development. Peek at Architecture to see what's wired up today.
\n
Routing & reliability
\nCost control
\nSafety & governance
\nOperations & UX
\n\n ![\"Manage](\"https://raw.githubusercontent.com/mydisha/keirouter/main/assets/keirouter-providers-banner.png\")
\n
\n ![\"Intelligent](\"https://raw.githubusercontent.com/mydisha/keirouter/main/assets/keirouter-usage-banner.png\")
\n
Grab whichever path matches what's already on your machine โ no need to install things you won't use:
\n| Method | \nYou'll need | \nGreat for | \n
|---|---|---|
| Homebrew | \nmacOS or Linux with Homebrew | \nThe fastest way to a prebuilt binary | \n
| Windows | \nWindows 10/11 with PowerShell | \nPrebuilt binary, no Go/Node | \n
| From source | \nGo 1.24+ and Node.js 20+ | \nLocal hacking / latest main | \n
| Docker | \nJust Docker | \nClean, isolated runs | \n
| Docker Compose | \nDocker + Docker Compose | \nVPS / production / Coolify | \n
brew tap mydisha/keirouter https://github.com/mydisha/keirouter\nbrew install keirouter\n\nkeirouter -bootstrap # mint your first API key (printed once โ don't blink)\nkeirouter start # fire up the server on :20180\nNeeds Go 1.24+ and Node.js 20+. No cloning, no .env, no config wrangling:
curl -fsSL https://raw.githubusercontent.com/mydisha/keirouter/main/scripts/quickstart.sh | bash\nIt clones the repo to ~/keirouter, installs everything, and starts the backend on :20180 and the dashboard on :5180.
\n\n\nOption C โ Docker ยท no Go/Node, no problemAlready cloned it? Just
\nmake setupfrom the project root and you're off.
curl -fsSL https://raw.githubusercontent.com/mydisha/keirouter/main/scripts/install.sh | bash -s -- --docker\ngit clone https://github.com/mydisha/keirouter.git\ncd keirouter\ncp .env.example .env # set KEIROUTER_MASTER_KEY before going to prod\ndocker compose up -d --build\nVPS, PostgreSQL, and Coolify notes live in deploy/README.md.
\n\nOption E โ Windows ยท prebuilt, no Go/Node neededOpen PowerShell and run:
\nirm https://raw.githubusercontent.com/mydisha/keirouter/main/scripts/install.ps1 | iex\nIt downloads the latest prebuilt binary, drops it (plus the dashboard) into %LOCALAPPDATA%\\KeiRouter, and adds it to your PATH. Open a new terminal afterwards, then:
keirouter -bootstrap # mint your first API key (printed once)\nkeirouter start # start the server on :20180\n\n\nPin a version or change the location with
\n$env:KEIROUTER_VERSION/$env:KEIROUTER_DIRbefore running the one-liner.
| How you installed | \nOpen this | \nPassword | \n
|---|---|---|
From source (quickstart / make setup) | \nhttp://localhost:5180 | \nkeirouter | \n
| Homebrew / Docker / production | \nhttp://localhost:20180 | \nkeirouter | \n
It'll nudge you to change that password the second you log in (please do ๐). Allergic to UIs? Mint a key straight from the terminal:
\nkeirouter -bootstrap # prints a kr_ key, once\nIn your AI tool of choice (Cursor, Claude Code, Cline, you name it), set:
\nhttp://localhost:20180/v1kr_ key from the dashboard or -bootstrapopenai/gpt-4o, or the name of a fallback chain you cooked upThat's it. You're routing.
\nEvery request runs through a deterministic token-saving pipeline before it gets translated to the provider's format โ so the savings work the same no matter which provider you land on. There are five savers, each independently toggleable from Settings โ Token Saving, and a dashboard that shows you exactly how much you clawed back.
\n| Saver | \nSide | \nThe pitch | \n
|---|---|---|
| RTK / Slimmer | \nInput | \nShrinks chunky tool output (diffs, greps, listings, build logs) locally before it ever leaves your machine. | \n
| Headroom | \nInput | \nRoutes request messages through an external Headroom proxy for deeper compression. Fail-open โ if the proxy sneezes, your request sails through untouched. Also sniffs out \"phantom savings\" so only real wins get counted. | \n
| Terse | \nOutput | \nDrops a concise-output directive so the model skips the small talk and gives you the goods. | \n
| Caveman | \nOutput | \nTerse's stronger cousin (Wenyan / ๆ่จๆ levels included) โ trims output tokens by 65โ75%. | \n
| Ponytail | \nOutput | \nInjects a \"lazy senior dev\" system prompt (lite / full / ultra) that nudges the model toward the smallest possible change. Stacks on top of Terse or Caveman. | \n
\n\nTerse and Caveman both inject a system directive, so they're mutually exclusive โ pick one. Ponytail happily layers on top of either.
\n
Pipeline order: normalizer โ RTK โ Headroom โ Terse / Caveman โ Ponytail โ provider translation.
Headroom is its own open-source compression proxy. KeiRouter just calls its /v1/compress endpoint, so you spin it up locally first. One gotcha worth shouting about: the headroom CLI lives in the Python package โ the npm package is a library only, so npm install -g headroom-ai will leave you staring at command not found. Don't say we didn't warn you. ๐
# The clean way: pipx isolates the CLI and sorts out your PATH (needs Python 3.10+)\npipx install \"headroom-ai[all]\"\npipx ensurepath # puts headroom on your PATH โ then restart your shell\n\nheadroom proxy --port 8787 # start the proxy\nheadroom doctor # make sure it's actually happy\nOn Ubuntu and pip is fighting you (PEP 668 blocks global installs)? Go --user and make sure ~/.local/bin is on your PATH:
pip install --user \"headroom-ai[all]\"\nexport PATH=\"$HOME/.local/bin:$PATH\" # drop this in ~/.zshrc or ~/.bashrc\nThen over in Settings โ Token Saving โ Headroom: flip it on, set the Proxy URL to http://localhost:8787, and hit Test connection to confirm the handshake. Green check? You're golden.
Why bet on one model when you can have a backup plan? Build a chain in the dashboard. Say you name one coding:
openai/gpt-4o โ your first pickdeepseek/deepseek-chat โ steps in if the first one rate-limits or face-plantsThen just set your app's model to chain:coding (or plain coding) and let KeiRouter sweat the failover.
Chat completions are just the start. KeiRouter proxies the whole buffet:
\n| Capability | \nEndpoint | \n
|---|---|
| Image generation | \n/v1/images/generations | \n
| Speech-to-text | \n/v1/audio/transcriptions | \n
| Text-to-speech | \n/v1/audio/speech | \n
| Embeddings | \n/v1/embeddings | \n
| Web search | \n/v1/search | \n
| Web fetch | \n/v1/web/fetch | \n
Copy-pasting API keys gets old fast. Connect providers straight from the Connections page with OAuth โ sign in once, and KeiRouter quietly refreshes your tokens in the background.
\n| Provider | \nFlow | \n
|---|---|
| Claude | \nAnthropic OAuth | \n
| GitHub Copilot | \nGitHub device flow | \n
| Gemini CLI | \nGoogle device flow | \n
| KiloCode | \nCustom device-auth | \n
| Qoder | \nPKCE device-token flow | \n
| CodeBuddy (Tencent) | \nBrowser-poll flow | \n
| Cursor | \nToken import flow | \n
Plans are reusable budget policies you can stamp onto any API key โ write the rules once, apply them everywhere. Each plan covers:
\n0 = unlimited)daily, weekly, monthly, or totalclaude-*, gpt-4* (empty = everything's fair game)Every tenant gets a default plan out of the box. Manage them on the Plans page and assign them in Keys settings.
\nKeep your gateway (and your upstream quotas) from getting hammered with per-key RPM, TPM, and concurrency caps. For a single-instance setup, the in-memory limiter is all you need:
\nlimits:\n enabled: true\n backend: memory\n default_rpm: 600\n default_tpm: 200000\n default_concurrency: 50\n window: 1m\n cleanup_interval: 1m\nThose defaults only apply to keys without a plan. The moment a key has one, the plan's rpm_limit, tpm_limit, and concurrency_limit take over (0 = unlimited).
A built-in content-safety layer runs detectors against every request and response. Policies stack global โ provider โ model โ chain โ API key and merge at request time, so the most specific rule wins.
\n| Detector | \nCatches | \nDefault engine | \nOptional engine | \n
|---|---|---|---|
| PII | \nEmail, phone, credit card, IBAN, IP, URL, NIK / NPWP / Indonesian passport | \nNative Go (Presidio-compatible) | \nMicrosoft Presidio HTTP sidecar | \n
| Prompt Injection | \nIgnore-previous, role override, DAN, prompt-leak, safety bypass | \nNative regex catalog | \nโ | \n
| Topics | \nAllow-list / block-list of topics | \nKeyword + n-gram | \nEmbedding similarity | \n
| Toxicity | \nProfanity, hate, harassment, violence, sexual (id + en) | \nNative catalog | \nOpenAI Moderation API | \n
| Bias (outbound) | \nPolitical, gender, ethnic, religious bias in responses | \nNative bilingual lexicon | \nโ | \n
log_only, warn, mask (rewrite it), or block (refuse it). Strictest action across detectors wins.redact, replace, mask, hash, anonymize, or block.keirouter_guardrail_decisions_total and keirouter_guardrail_eval_seconds at /metrics.allow_external_engines flag forces every detector offline; KEIROUTER_GUARDRAILS__AUDIT_RETENTION_DAYS controls retention; the test endpoint is capped at 10 req/min.Starter templates ship in the dashboard's \"From template\" picker (Indonesia PII ยท Strict safety ยท Compliance audit ยท Public chatbot ยท Alerts-only), and you can export/import policies as a JSON bundle.
\nWant NER-based PII detection (PERSON, LOCATION, full multilingual)? Spin up the optional Presidio sidecar:
docker compose -f compose.yaml -f compose.postgres.yaml -f compose.presidio.yaml up -d\nThen flip any PII policy's engine to presidio in the dashboard.
Rebrand the admin dashboard and the public Usage Portal from Settings โ Branding:
\nsage-terra, ocean, midnight, and friendsThe CLI Tools page spits out ready-to-paste config snippets for the usual suspects:
\n\n\nClaude Code ยท Cursor ยท Cline ยท GitHub Copilot ยท DeepSeek ยท KiloCode ยท OpenCode ยท OpenClaw ยท Hermes ยท JCode ยท Droid ยท CodeBuddy
\n
Copy, paste into your tool's config, done.
\nA dedicated, no-admin-required view at /portal where teammates keep an eye on their own usage โ quota and spend, token usage over time, compression savings, and plan limits. All it asks for is the API key. No keys to the kingdom required.
๐ง LLM / Chat
\n| Category | \nProviders | \n
|---|---|
| Major Cloud | \nOpenAI, Anthropic, Google Gemini, Vertex AI, Azure OpenAI, AWS (Kiro) | \n
| Free / Free Tier | \nOpenRouter (27+ free models), NVIDIA NIM, Ollama (Cloud & Local), Cloudflare Workers AI, BytePlus ModelArk | \n
| China / Asia | \nDeepSeek, Qwen (Alibaba), GLM, Kimi (Moonshot), MiniMax, Volcengine Ark, Xiaomi MiMo, SiliconFlow, iFlow | \n
| OAuth / IDE | \nClaude Code, GitHub Copilot, Cursor IDE, Cline, Kilo Code, OpenAI Codex, CodeBuddy (Tencent), Kimi Coding | \n
| Performance | \nGroq, Cerebras, SambaNova, DeepInfra | \n
| Specialized | \nxAI (Grok), Mistral, Perplexity, Cohere, AI21 Labs, Reka AI | \n
| Aggregators | \nTogether AI, Fireworks AI, Nebius AI, OpenCode, AIML API, Vercel AI Gateway | \n
| Emerging | \nBlackbox AI, Chutes AI, Hyperbolic, Lepton AI, Kluster AI, MorphLLM, LongCat, Puter AI, GLHF, SumoPod, Scaleway, NLP Cloud, and many more | \n
| Custom | \nAny OpenAI- or Anthropic-compatible endpoint (self-hosted, proxy, etc.) | \n
๐จ Media & Search
\n| Type | \nProviders | \n
|---|---|
| Image Generation | \nOpenAI DALLยทE, Gemini Imagen, Cloudflare, Fal.ai, Stability AI, Black Forest Labs, Recraft, Topaz, Runway ML, NanoBanana, HuggingFace, SD WebUI, ComfyUI | \n
| Text-to-Speech | \nOpenAI TTS, NVIDIA NIM, ElevenLabs, Deepgram, Cartesia, PlayHT, AWS Polly, Google TTS, Edge TTS, Inworld, Coqui, Tortoise | \n
| Speech-to-Text | \nOpenAI Whisper, Groq Whisper, Deepgram, AssemblyAI, Gemini STT, HuggingFace | \n
| Embeddings | \nOpenAI, Gemini, Mistral, Together AI, Fireworks AI, Nebius, Voyage AI, Jina AI, OpenRouter | \n
| Web Search | \nTavily, Exa, Serper, Brave Search, SearXNG, Perplexity, xAI, Google PSE, Linkup, SearchAPI, You.com, OpenAI | \n
| Web Fetch | \nTavily, Exa, Firecrawl, Jina Reader | \n
Out of the box, KeiRouter runs on an embedded SQLite database โ zero config, zero fuss. Running it for a team? Switch to PostgreSQL: copy config.example.yaml and run with -config, or use environment variables like KEIROUTER_SERVER__PORT=8080. Docker/Coolify examples are in deploy/README.md.
Curious what happens after you hit send? Here's the life of a request:
\n/api/*) only listens to localhost by default. Exposing it? Put it behind a reverse proxy and set a stable master_key.Found a security issue? Please follow SECURITY.md instead of opening a public issue. ๐
\nmake setup # first time: installs deps + starts backend (:20180) and dashboard (:5180)\nmake dev # after that: just start the servers\nmake test # run the backend test suite\nmake build # build the backend binary + frontend assets\nPRs and ideas are always welcome โ start with CONTRIBUTING.md.
\nMIT โ see LICENSE. Go build something cool. ๐ ๏ธ
\n" @@ -829,8 +829,8 @@ "static-site", "sveltekit" ], - "updatedAt": "2026-07-22T04:44:08Z", - "pushedAt": "2026-07-22T09:18:51Z", + "updatedAt": "2026-07-22T12:03:53Z", + "pushedAt": "2026-07-22T12:01:43Z", "latestRelease": { "name": "Bansos v0.0.13", "tagName": "v0.0.13", @@ -841,10 +841,10 @@ "licenseSpdx": "MIT", "createdAt": "2026-06-11T13:28:54Z", "openIssues": 3, - "openPullRequests": 2, + "openPullRequests": 1, "subscribers": 0, "communityHealth": 100, - "readmeHtml": "\n
๐ฎ๐ฉ Indonesia (Default) ยท ๐ English
\nBantuan sosial untuk developer jelata
bansos.dev adalah open-source katalog info bagi-bagi berkah, promo gratisan, dan diskonan tools coding paling legit khusus untuk developer jelata di Indonesia. Dibuat biar portofolio kita-kita tetep menyala walau dompet lagi sekarat. Nyari domain gratis, hosting free-tier, cloud credits, API credits, database gratisan, atau startup credits? Di sini tempat ngumpulnya! 100% Gratisan, No Clickbait, No Ribet. fr fr ๐
\nSitus ini dibangun sebagai static SvelteKit site yang super SEO-friendly, data-driven, aman di mode terang/gelap, dan gampang banget buat dikontribusikan lewat email atau merge request.
\nbansos developer, promo developer Indonesia, domain gratis, cloud credits gratis, API credits, hosting free tier, startup credits, developer tools gratis, open source Indonesia, SvelteKit static site.
src/lib/data/bansos.json.Situs ini di-deploy dan di-hosting menggunakan Cloudflare Pages dengan adapter @sveltejs/adapter-cloudflare. Setiap kali ada merge request atau push ke branch main, Cloudflare secara otomatis memicu build dan mendistribusikan situs statis super cepat beserta seluruh dynamic OG image yang sudah di-prerender.
npm install\nnpm run dev\nnpm run build\nValidasi lokal:
\nnpm run check\nnpm run lint\nsrc/lib/data/bansos.json # data utama listing bansos\nsrc/lib/data/bansos.ts # helper selector, sorting, dan contributor stats\nsrc/lib/components/ # komponen UI reusable\nsrc/routes/list/ # halaman list dan detail bansos\nsrc/routes/contribute/ # panduan kontribusi publik\nscripts/add-bansos.mjs # script lokal tambah data\npackages/bansosdev-cli/ # CLI bansosdev (disabled untuk submit publik)\nUntuk saat ini, submit publik yang aktif adalah via email dan Git clone. Jalur form, npx CLI, dan bot dinonaktifkan sementara karena spam.
\n\n\n[!TIP]\nSoon: Submisi via Discord & Telegram Bot!\nKami sedang membangun integrasi bot agar kamu bisa mengirimkan bansos baru secara otomatis langsung dari server Discord atau channel Telegram.\nSembari menunggu, yuk gabung ke komunitas kami:
\n\n
\n- Discord Server untuk ngobrol, diskusi, dan submit via chat (coming soon).
\n- Telegram Channel untuk dapetin update instan promo developer terbaru langsung di HP-mu.
\n
Opsi ini sangat cocok buat kamu yang ingin berbagi info dengan cepat tanpa perlu menyentuh terminal.
\nSubmit publik via npx bansosdev add sedang dinonaktifkan sementara karena spam. Dokumentasi CLI tetap disimpan untuk maintainer dan pengujian lokal, tetapi jangan dipakai untuk submit publik saat ini.
npx bansosdev add\nCLI akan menuntunmu mengisi field demi field untuk menyiapkan data lokal.
\nKamu juga bisa mengirimkan data langsung menggunakan argumen CLI:
\nnpx bansosdev add \\\n --id contoh-bansos \\\n --title \"Contoh Bansos Developer\" \\\n --provider \"Example Provider\" \\\n --description \"Deskripsi singkat bansos.\" \\\n --benefits \"Benefit satu|Benefit dua\" \\\n --validity-type fixed \\\n --validity-date 2026-06-30 \\\n --validity-desc \"Berlaku khusus pelajar\" \\\n --published-at 2026-06-13 \\\n --requirements \"Buat akun|Klaim program\" \\\n --cta-link \"https://example.com\" \\\n --contributor-name \"Nama Kamu\" \\\n --contributor-url \"https://example.com\" \\\n --tags \"Cloud,Gratisan\"\n--validity-type wajib: pilih fixed, uncertain, atau forever.--validity-date wajib jika --validity-type fixed, memakai format YYYY-MM-DD (Berfungsi sebagai Tanggal Berakhir).--validity-desc opsional untuk catatan masa berlaku, kuota, atau syarat khusus.--published-at opsional untuk tanggal mulai berlaku (start date) dalam format YYYY-MM-DD. Default adalah hari ini.--source opsional untuk sumber verifikasi; bisa berupa URL atau teks biasa.\n\nCatatan Otomatisasi:
\n\n
\n- Parameter
\nproviderakan diekstrak secara otomatis dari domaincta-link.- Parameter
\nstatusakan dihitung otomatis (active,upcoming, atauexpired) berdasarkan tanggalpublished-atdanvalidity-date.
npx bansosdev add ... --mode json\nOpsi ini bagi kamu yang ingin menguji kode secara lokal atau memodifikasi file secara langsung.
\nClone repositori ini ke komputermu:
\ngit clone https://gitlab.com/wauputr4/bansos.git\ncd bansos\nnpm install\nTambahkan data secara lokal menggunakan helper script:
\nnpm run add:bansos -- \\\n --id contoh-bansos \\\n --title \"Contoh Bansos Developer\" \\\n --provider \"Example Provider\" \\\n --description \"Deskripsi singkat bansos.\" \\\n --benefits \"Benefit satu|Benefit dua\" \\\n --validity-type fixed \\\n --validity-date 2026-06-30 \\\n --requirements \"Buat akun|Klaim program\" \\\n --cta-link \"https://example.com\" \\\n --contributor-name \"Nama Kamu\" \\\n --contributor-url \"https://example.com\" \\\n --tags \"Cloud,Gratisan\"\nScript ini akan memvalidasi data dan menyimpannya di file data terstruktur src/lib/data/bansos.json.
Argumen --benefits dan --requirements dipisahkan dengan |.\nArgumen --tags dipisahkan dengan koma.
Buat branch baru, tambahkan commit, push ke fork, dan kirim merge request ke repositori utama.
\nMode direct untuk submit otomatis sedang dinonaktifkan. Untuk perubahan maintainer, gunakan Git clone, commit manual, dan merge request ke main.
npx bansosdev add ... --mode json\nPerintah di atas hanya untuk mengecek payload JSON secara lokal.
\nDetail lengkap CLI lihat docs/bansosdev-cli.md.
\nListing yang baik sebaiknya menyertakan:
\nCloud, Domain, AI Credits, Startup, atau No Credit Card.Ikuti Code of Conduct.
\nProyek bansos.dev dibangun secara gratis oleh komunitas. Jika proyek ini membantumu menghemat budget developer-mu, silakan kirim dukungan via email ke me@wau.my.id.
\n\n[!NOTE]\nSoon: Kami berencana menghadirkan fitur di mana donatur/pengunjung bisa mengirimkan dukungan (donasi) langsung ke masing-masing kontributor yang mendaftarkan/menulis listing bansos tersebut.
\n
MIT. Lihat LICENSE.
\nbansos.dev adalah platform komunitas open-source yang bertujuan membantu sesama developer Indonesia menemukan program bantuan sosial yang sah dan legal dari provider resmi. Kami tidak terafiliasi dengan provider mana pun.
\nKami dengan tegas melarang:
\nSemua informasi yang ditampilkan bersifat referensi. Selalu verifikasi langsung ke situs resmi provider sebelum melakukan klaim. Kami tidak bertanggung jawab atas perubahan kebijakan sepihak dari provider, interpretasi manfaat yang keliru, ataupun penyalahgunaan informasi oleh pihak tidak bertanggung jawab.
\nDengan menggunakan bansos.dev, Anda menyetujui bahwa platform ini hanyalah katalog komunitas dan segala klaim, transaksi, atau interaksi dengan provider sepenuhnya merupakan tanggung jawab pribadi masing-masing pengguna.
\n" + "readmeHtml": "\n
๐ฎ๐ฉ Indonesia (Default) ยท ๐ English
\nBantuan sosial untuk developer jelata
bansos.dev adalah open-source katalog info bagi-bagi berkah, promo gratisan, dan diskonan tools coding paling legit khusus untuk developer jelata di Indonesia. Dibuat biar portofolio kita-kita tetep menyala walau dompet lagi sekarat. Nyari domain gratis, hosting free-tier, cloud credits, API credits, database gratisan, atau startup credits? Di sini tempat ngumpulnya! 100% Gratisan, No Clickbait, No Ribet. fr fr ๐
\nSitus ini dibangun sebagai static SvelteKit site yang super SEO-friendly, data-driven, aman di mode terang/gelap, dan gampang banget buat dikontribusikan lewat email atau pull request.
\nbansos developer, promo developer Indonesia, domain gratis, cloud credits gratis, API credits, hosting free tier, startup credits, developer tools gratis, open source Indonesia, SvelteKit static site.
src/lib/data/bansos/.Situs ini di-deploy dan di-hosting menggunakan Cloudflare Pages dengan adapter @sveltejs/adapter-cloudflare. Setiap kali ada pull request atau push ke branch main, Cloudflare secara otomatis memicu build dan mendistribusikan situs statis super cepat beserta seluruh dynamic OG image yang sudah di-prerender.
npm install\nnpm run dev\nnpm run build\nValidasi lokal:
\nnpm run check\nnpm run lint\nsrc/lib/data/bansos/<slug>/ # index.json + README tiap listing\nsrc/lib/data/bansos/contributors/ # profil kontributor\nsrc/lib/data/bansos.ts # loader, selector, sorting, dan stats\nsrc/lib/components/ # komponen UI reusable\nsrc/routes/list/ # halaman list dan detail bansos\nsrc/routes/contribute/ # panduan kontribusi publik\nscripts/add-bansos.mjs # script lokal tambah data\npackages/bansosdev-cli/ # CLI lama (submit publik dinonaktifkan)\nSetiap listing wajib memiliki contributorSlug yang terhubung dua arah dengan manifest profil.\nAvatar diambil otomatis dari GitHub bila tersedia; profil tanpa GitHub tetap tampil memakai dua\ninisial. Semua tautan nama kontributor di situs mengarah ke profil internal bansos.dev.\nKonten profil dapat disesuaikan melalui src/lib/data/bansos/contributors/<slug>/README.md.\nProfil publik memakai URL canonical https://bansos.dev/<slug>/; validator mencegah slug contributor\nbentrok dengan route situs atau shortlink bansos.
Untuk saat ini, submit publik yang aktif adalah via AI Agent, email, dan Git clone. Jalur form, CLI publik, dan bot dinonaktifkan sementara karena spam.
\n\n\n[!TIP]\nSoon: Submisi via Discord & Telegram Bot!\nKami sedang membangun integrasi bot agar kamu bisa mengirimkan bansos baru secara otomatis langsung dari server Discord atau channel Telegram.\nSembari menunggu, yuk gabung ke komunitas kami:
\n\n
\n- Discord Server untuk ngobrol, diskusi, dan submit via chat (coming soon).
\n- Telegram Channel untuk dapetin update instan promo developer terbaru langsung di HP-mu.
\n
Install skill resmi bansos.dev agar agent memahami struktur listing, profil contributor, validasi, dan alur PR terbaru:
\nnpx skills add wauputr4/skill-bansos --skill '*' --agent '*'\nSetelah terpasang, berikan link sumber dan minta agent memakai $bansos-add-entry. Review data dan diff sebelum membuka pull request.
Opsi ini sangat cocok buat kamu yang ingin berbagi info dengan cepat tanpa perlu menyentuh terminal.
\nOpsi ini bagi kamu yang ingin menguji kode secara lokal atau memodifikasi file secara langsung.
\nFork dan clone repositori ini ke komputermu memakai GitHub CLI:
\ngh repo fork wauputr4/bansos --clone\ncd bansos\nnpm install\nTambahkan data secara lokal menggunakan helper script:
\nnpm run add:bansos -- \\\n --id contoh-bansos \\\n --title \"Contoh Bansos Developer\" \\\n --provider \"Example Provider\" \\\n --description \"Deskripsi singkat bansos.\" \\\n --benefits \"Benefit satu|Benefit dua\" \\\n --validity-type fixed \\\n --validity-date 2026-06-30 \\\n --requirements \"Buat akun|Klaim program\" \\\n --cta-link \"https://example.com\" \\\n --contributor-slug username-kamu \\\n --contributor-name \"Nama Kamu\" \\\n --contributor-url \"https://example.com\" \\\n --tags \"Cloud,Gratisan\"\nScript ini akan memvalidasi data lalu membuat src/lib/data/bansos/<slug>/index.json dan README listing.
Argumen --benefits dan --requirements dipisahkan dengan |.\nArgumen --tags dipisahkan dengan koma.\n--contributor-slug wajib pada setiap submit. --contributor-name wajib dan\n--contributor-url opsional hanya jika profil contributor tersebut belum ada.
Buat branch baru, tambahkan commit, push ke fork, dan kirim pull request ke repositori utama.
\nListing yang baik sebaiknya menyertakan:
\nCloud, Domain, AI Credits, Startup, atau No Credit Card.Ikuti Code of Conduct.
\nProyek bansos.dev dibangun secara gratis oleh komunitas. Jika proyek ini membantumu menghemat budget developer-mu, silakan kirim dukungan via email ke me@wau.my.id.
\n\n[!NOTE]\nSoon: Kami berencana menghadirkan fitur di mana donatur/pengunjung bisa mengirimkan dukungan (donasi) langsung ke masing-masing kontributor yang mendaftarkan/menulis listing bansos tersebut.
\n
MIT. Lihat LICENSE.
\nbansos.dev adalah platform komunitas open-source yang bertujuan membantu sesama developer Indonesia menemukan program bantuan sosial yang sah dan legal dari provider resmi. Kami tidak terafiliasi dengan provider mana pun.
\nKami dengan tegas melarang:
\nSemua informasi yang ditampilkan bersifat referensi. Selalu verifikasi langsung ke situs resmi provider sebelum melakukan klaim. Kami tidak bertanggung jawab atas perubahan kebijakan sepihak dari provider, interpretasi manfaat yang keliru, ataupun penyalahgunaan informasi oleh pihak tidak bertanggung jawab.
\nDengan menggunakan bansos.dev, Anda menyetujui bahwa platform ini hanyalah katalog komunitas dan segala klaim, transaksi, atau interaksi dengan provider sepenuhnya merupakan tanggung jawab pribadi masing-masing pengguna.
\n" }, { "fullName": "giosakti/duragent", @@ -983,8 +983,8 @@ "stars": 45, "forks": 10, "topics": [], - "updatedAt": "2026-07-21T04:20:00Z", - "pushedAt": "2026-07-21T04:20:02Z", + "updatedAt": "2026-07-22T09:59:47Z", + "pushedAt": "2026-07-22T09:59:29Z", "latestRelease": { "name": "v0.15.1", "tagName": "v0.15.1", @@ -2266,13 +2266,13 @@ "static-site", "sveltekit" ], - "updatedAt": "2026-06-25T07:10:51Z", - "pushedAt": "2026-06-14T21:13:08Z", + "updatedAt": "2026-07-22T11:40:45Z", + "pushedAt": "2026-07-22T11:40:45Z", "latestRelease": { - "name": "skill-bansos v0.1.1", - "tagName": "v0.1.1", - "url": "https://github.com/wauputr4/skill-bansos/releases/tag/v0.1.1", - "publishedAt": "2026-06-13T11:36:57Z" + "name": "skill-bansos v0.2.0", + "tagName": "v0.2.0", + "url": "https://github.com/wauputr4/skill-bansos/releases/tag/v0.2.0", + "publishedAt": "2026-07-22T11:40:46Z" }, "archived": false, "licenseSpdx": "MIT", @@ -2281,7 +2281,7 @@ "openPullRequests": 0, "subscribers": 0, "communityHealth": 100, - "readmeHtml": "Open Agent Skills for AI agents working on bansos.dev and wauputr4/bansos.
These skills follow the skills.sh / npx skills format: each skill lives in skills/<skill-name>/ with a required SKILL.md, optional references/, and optional agents/openai.yaml UI metadata.
bansos-add-entry: add or review bansos.dev listings using source-backed data and the npx bansosdev add workflow.bansos-develop-feature: develop UI, SEO, data, and feature changes for the static SvelteKit bansos.dev site.Install all skills to every supported agent:
\nnpx skills add wauputr4/skill-bansos --skill '*' --agent '*'\nInstall only the submit-entry skill:
\nnpx skills add wauputr4/skill-bansos --skill bansos-add-entry\nUse each skill directly by path:
\nUse $bansos-add-entry to add a new bansos.dev listing from this source: ...\nUse $bansos-develop-feature to improve the mobile contribution page UI.\nValidate before publishing or opening a pull request:
\npython3 /path/to/skill-creator/scripts/quick_validate.py skills/bansos-add-entry\npython3 /path/to/skill-creator/scripts/quick_validate.py skills/bansos-develop-feature\nMIT. See LICENSE.
\n" + "readmeHtml": "Open Agent Skills for AI agents working on bansos.dev and wauputr4/bansos.
These skills follow the skills.sh / npx skills format: each skill lives in skills/<skill-name>/ with a required SKILL.md, optional references/, and optional agents/openai.yaml UI metadata.
bansos-add-entry: add or review folder-based bansos.dev listings with source verification and contributor attribution.bansos-develop-feature: develop UI, SEO, data, and feature changes for the SvelteKit and Cloudflare Pages site.Install all skills to every supported agent:
\nnpx skills add wauputr4/skill-bansos --skill '*' --agent '*'\nInstall only the submit-entry skill:
\nnpx skills add wauputr4/skill-bansos --skill bansos-add-entry\nUse each skill directly by path:
\nUse $bansos-add-entry to add a new bansos.dev listing from this source: ...\nUse $bansos-develop-feature to improve the mobile contribution page UI.\nValidate before publishing or opening a pull request:
\npython3 /path/to/skill-creator/scripts/quick_validate.py skills/bansos-add-entry\npython3 /path/to/skill-creator/scripts/quick_validate.py skills/bansos-develop-feature\nRelease tags are created from merged main. The current folder-based workflow update is planned for v0.2.0; see CONTRIBUTING.md.
MIT. See LICENSE.
\n" }, { "fullName": "wauputr4/agent-proxmox", From 3226f576bb9f334e497549c00362e665cc97c361 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Wed, 22 Jul 2026 14:29:16 +0000 Subject: [PATCH 18/25] Sync content data --- src/data/projects.json | 44 +++++++++++++++++++++--------------------- src/data/revival.json | 4 ++-- 2 files changed, 24 insertions(+), 24 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index 7a2428a..2b6342a 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -10,7 +10,7 @@ "homepage": "https://uaparser.dev/", "language": "JavaScript", "stars": 10168, - "forks": 1220, + "forks": 1219, "topics": [ "analytics", "bot-detection", @@ -123,13 +123,13 @@ "url": "https://github.com/OpenSID/OpenSID", "homepage": "", "language": "PHP", - "stars": 1212, - "forks": 1167, + "stars": 1213, + "forks": 1168, "topics": [ "opensid", "sistem-informasi-desa" ], - "updatedAt": "2026-07-22T07:59:43Z", + "updatedAt": "2026-07-22T13:40:31Z", "pushedAt": "2026-07-18T13:09:40Z", "latestRelease": { "name": "Rilis v2607.0.0", @@ -142,7 +142,7 @@ "createdAt": "2016-05-21T10:55:38Z", "openIssues": 353, "openPullRequests": 7, - "subscribers": 111, + "subscribers": 110, "communityHealth": 50, "readmeHtml": "
OpenSID adalah Sistem Informasi Desa (SID) yang dikembangkan secara terbuka dan kolaboratif oleh komunitas yang peduli dengan SID.
\nSID diharapkan dapat membantu pemerintah desa dalam beberapa hal berikut:
\n\n\nOpenSID bertujuan agar sebanyak mungkin desa di Indonesia dapat menerapkan sistem informasi untuk memajukan desa masing-masing..
\n
Strategi pengembangan OpenSID adalah untuk:
\nOpenSID dikelola di GitHub untuk:
\nPanduan pemasangan dan penggunaan OpenSID tersedia di Panduan OpenSID.
\nSistem Informasi Desa (SID) pertama kali dikembangkan oleh Combine Resource Institution sejak tahun 2009. Hak cipta awal dimiliki oleh Combine Resource Institution (http://lumbungkomunitas.net/).
\nSistem ini dikelola berdasarkan lisensi GNU General Public License Versi 3 (http://www.gnu.org/licenses/gpl.html).
\nVersi GitHub ini dikembangkan sejak Mei 2016, gratis dan bebas dimanfaatkan serta dikembangkan oleh semua desa. Hak Cipta OpenSID kini dipegang oleh Perkumpulan Desa Digital Terbuka (https://opendesa.id), sebuah lembaga hukum yang dibentuk khusus untuk mengelola OpenSID.
\nDemo Versi Publik OpenSID dapat dicoba di: https://demosid.opendesa.id
Modul admin: https://demosid.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Demo Versi Premium OpenSID dapat dicoba di: https://berputar.opendesa.id
Modul admin: https://berputar.opendesa.id/index.php/siteman
Username: admin, Password: sid304
Bergabunglah dengan Forum Pengguna dan Pegiat OpenSID di Facebook atau di Telegram.
Forum ini bersifat informal, sebagai wadah berbagi informasi dan saling membantu dalam menggunakan dan mengembangkan OpenSID.
Laporkan masalah, usulan, atau permintaan pengembangan OpenSID melalui issue GitHub.
Kontribusi dari komunitas SID sangat dihargai, baik untuk dokumentasi di Wiki OpenSID maupun untuk source code di repo utama.
\n
Peduli OpenSID dan misi membangun desa? Dukung OpenSID di sini.
\n\nAtau donasi langsung melalui rekening bank. Info lengkap di sini.
\n
Apakah desa, lembaga, atau perusahaan Anda mendapat manfaat dari OpenSID?
Bantu kami mengembangkan OpenSID dengan menjadi sponsor.
Logo sponsor Anda akan tampil di sini dengan tautan ke situs Anda.
Berikut adalah para kontributor luar biasa yang telah membantu mengembangkan OpenSID:
\n\n Noise-canceling context and long-term memory for your AI agent โ lossy, but always reversible, and it never fabricates a result. Stop paying Claude to read 10,000 lines of terminal noise.\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Always reversible ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Every other compressor asks you to trust that what it cut didn't matter. OMNI doesn't ask โ it guarantees, and each guarantee is backed by code you can read:
\n| Guarantee | \nHow | \nProof | \n
|---|---|---|
| Get the original back, byte-for-byte | \neverything cut is archived in a local SQLite RewindStore (SHA-256 โ content); the agent gets a hash and calls omni_retrieve | \nHow it works | \n
| Never fabricates a result | \na distiller that parsed no signal returns the raw output, never a green no errors / passed string | \n#143 | \n
| Failures are never masked | \na command that exits non-zero passes through verbatim | \n#120 | \n
| Structured data is never touched | \nJSON / YAML / NDJSON / CSV pass through byte-for-byte | \npipeline::format | \n
| Numbers are measured, not aspirational | \n1,810 real traces replayed on the release binary โ and 63.6% of calls net zero, which we publish too | \nBenchmarks | \n
That is the one thing a bigger compression number can't buy: you can always recover the original, and it will never lie to your agent.
\nProblem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
Superagent is an interactive, terminal-based AI coding assistant designed to facilitate the cycle of development, testing, debugging, and application optimization directly from your workspace.
\nIt features a cyberpunk-styled terminal user interface built with terminal UI components, automatic tracking of model context token limits, a robust security permission layer, a 3-tier multi-agent orchestration system (Master Agent โ Superagent โ Subagent), and persistent integration with local terminal shells.
\n![\"Superagent](\"https://raw.githubusercontent.com/RudyCity/superagent/main/assets/Video_SuperAgent.gif\")
In modern software development, developers frequently switch context between writing code, running terminal commands, inspecting system logs, searching documentation, and interacting with Large Language Models (LLMs).
\nSuperagent bridges this gap by providing an integrated terminal environment that understands your project's context automatically using a project specification file (agents.md), automates execution of independent tasks through secondary agents (subagents), and tracks LLM context window limits in real-time. Security is a primary design goal: every file modification, tool invocation, and shell command execution requires explicit user authorization.
Unlike standard headless execution bots or basic shell wrappers, Superagent is designed from the ground up as a fully interactive developer workspace companion:
\n/compact now to force compaction on demand, /pin to protect important messages from being compacted, and /compaction-history to audit all compaction events./knowledge to browse and search pinned knowledge across ALL sessions and projects. AI agents can access this via search_pinned_knowledge and load_pinned_session tools, enabling them to learn from previous sessions' decisions and context./checkpoint). If an experimental approach fails, you can revert back instantly to a previous checkpoint, restoring the entire session timeline.superagent --multi. Note: Multi-agent mode is currently experimental./terminal, Superagent spawns a real, popped-up host emulator terminal window. This is perfect for running interactive servers, watch scripts, and commands that require manual inputs..env or log files cluttering your project codebase. All API keys, environment settings, and session logs are kept safe and clean in your user's global directory (~/.superagent-r/).~/.superagent-r/workspace-caches/ to bypass redundant scans. It dynamically monitors and updates the cache when workspace changes occur in the agent loop. Additionally, git worktree directories created for Superagents (~/.superagent-r/worktrees/<name>) are automatically configured as trusted (safe.directory) in git to prevent \"dubious ownership\" warnings and errors.ask_question tool at every decision point โ choosing implementations, resolving ambiguity, or selecting approaches โ ensuring the AI never guesses or assumes on the user's behalf./terminal init wizard.D:\\images\\screenshot.png) directly into the prompt to auto-attach it. Press Ctrl+V to automatically capture image binary data from your system clipboard (cross-platform support for Windows, macOS, and Linux). Attached images are rendered in a sleek visual queue above the prompt and transmitted as high-fidelity multimodal inputs to vision-capable models (e.g. Claude 3.5 Sonnet, GPT-4o), with automatic token tracking.internal-hooks/<name>/ with a hook.json schema definition and an index.js entrypoint. Hooks are auto-discovered on startup and registered as first-class agent tools. Use /ih init <name> to scaffold the project, /ih dev <name> to run and test it locally, and /ih active to pick which hooks are active via an interactive multi-select checkbox dialog. Active selections are persisted per-project in ~/.superagent-r/model-config.json.Superagent is built on modern Node.js technologies for high performance and modular architecture:
\nai, @ai-sdk/openai, @ai-sdk/anthropic) for structured and streaming interactions.superagent/\nโโโ src/\nโ โโโ cli.tsx # Main entrypoint; routes --multi flag to masterAgent\nโ โโโ app.tsx # React UI wrapper and command handling logic\nโ โโโ core/\nโ โ โโโ agent.ts # Core cognitive loop and instruction runner\nโ โ โโโ masterAgent.ts # Master Agent orchestrator (3-tier entry point)\nโ โ โโโ config.ts # Environment variable and global config management\nโ โ โโโ checkpoints.ts # Conversation state checkpoint save/load logic\nโ โ โโโ slash-commands.ts # Interactive command definitions\nโ โ โโโ tools/\nโ โ โโโ types.ts # Shared types: AgentTier, SubagentInstance, ToolSet\nโ โ โโโ toolsets.ts # ToolSet definitions per tier (master/super/sub)\nโ โ โโโ prompts.ts # System prompts per tier with dynamic context\nโ โ โโโ state.ts # Shared subagent registry and event emitters\nโ โ โโโ shellTools.ts # Command execution and background task control\nโ โ โโโ systemTools.ts # File operations, directory creation, port checks\nโ โ โโโ subagentTools.ts # Subagent instantiation (superagent tier)\nโ โ โโโ superagentTools.ts # Superagent orchestration tools (master tier)\nโ โ โโโ dynamicHooks.ts # Internal hook discovery, loading, and active state\nโ โ โโโ networkTools.ts # Web content fetch and browser integration\nโ โโโ components/ # React Ink components (visual stats, wizards)\nโโโ bin/ # Portable Python and setup scripts\nโโโ tests/ # Unit test suites using Vitest\nโโโ package.json # Project manifest and scripts\nA rich terminal interface showing live statistics on active prompt sizes, completion token counts, token cost summaries, active models, remaining context windows, and real-time model generation speed (tokens per second).
\nAllows developers to save the current state of a coding conversation and restore it at any point using /checkpoint save <name> and /checkpoint restore <id>. This allows you to safely experiment with different implementations. Checkpoints can be browsed, restored, or deleted via an interactive wizard (launched by /checkpoint, /checkpoint list, or Ctrl+P in multi-agent mode). Use the --resume or -r flag to continue where you left off. Multi-agent sessions are fully serialized, ensuring smooth restore and resume of running tasks and interactive prompts. Auto-checkpointing creates snapshots automatically on every user message and before destructive tool operations, with a cooldown to prevent excessive saves โ ensuring you always have a safe rollback point.
--multi) (Experimental)\n\n[!WARNING]\nMulti-agent mode (
\n--multi) is currently experimental and not recommended for production environments.
Launch with superagent --multi to activate the full 3-tier hierarchy:
superagent --multi\n โ\n Master Agent (orchestrator tier)\n Tools: invoke_superagent, await_superagents, merge_superagents, manage_superagents, define_superagent, send_message_to_superagent, manage_subagents, git_worktree\n Spawns Superagents with git worktree isolation\n โ\n Superagent (per-feature coordinator/lead)\n Tools: shell + file tools, invoke_subagent, manage_subagents, git_worktree\n Isolated in its own git worktree\n Can spawn Subagents for atomic ops\n โ\n Subagent (atomic operation tier)\n Tools: file tools only (read/write/search)\n Ephemeral, single-purpose execution\nTier Responsibilities:
\nAdvanced Orchestration Features:
\nโโโ and โโโ connectors) mapping active relationships between the Master Agent, spawned Superagents, and running subagents/tasks.ViolationRecord event is emitted with severity levels (warning or critical). These violations automatically propagate up the agent hierarchy (Subagent โ Superagent โ Master Agent), enabling parent agents to track, log, and take corrective action โ including halting or redirecting the offending agent.ask_question tool at every decision point. Agents must present multiple-choice options to the user rather than guessing, assuming, or making unilateral decisions about implementation approaches, file selections, or design trade-offs.Standard subagent roles (single-agent mode):
\n/terminal)Runs development servers, local builds, or test watchers in popped-up, visible OS terminal windows (Windows cmd, macOS Terminal, Linux x-terminal). It includes an AI-assisted preset initializer (/terminal init) to auto-configure workspace command presets.
For complex changes, the Master Agent enforces structured planning and execution boundaries:
\nconstraints (files or logic NOT to modify) and acceptanceCriteria (a checklist of specific test cases to satisfy).implementation_plan.md and requires explicit user review and approval before any execution begins, guaranteeing complete oversight.The merge system uses a safe-by-default strategy that prevents file corruption:
\n<<<<<<< in files)mode: 'patch'): Lightweight mode that skips worktree creation and operates directly in the parent's working directory. Ideal for small, targeted fixes (1-2 lines). Much faster than spawning a full Superagent. Includes safety warnings if the parent worktree has uncommitted changes.baseBranch: 'feat/...'): When a Superagent needs to build on top of another feature branch instead of the current HEAD, specify baseBranch to create the worktree from that branch. Useful for dependent features or building on top of in-progress work.Example:
\ninvoke_superagent({\n role: 'fix-html-corrupt',\n task: 'Fix duplicate closing tags in Toolbar component',\n branch: 'fix/toolbar-html',\n baseBranch: 'feat/separate-compressor-menu', // Build on top of this branch\n mode: 'patch' // Quick fix, no worktree needed\n})\nAll agent operations, including single-agent and 3-tier multi-agent processes, are dynamically logged to a central log file in the user's home directory (~/.superagent-r/superagent.log). The log maintains tier-aware indentation to cleanly trace parallel execution branches.
Beyond manual /checkpoint commands, Superagent creates checkpoints automatically:
In multi-agent mode, the permission layer emits structured ViolationRecord events whenever a child agent attempts a blocked operation. These violations include:
\"warning\" for soft blocks (e.g., scope creep attempts) and \"critical\" for hard policy violations (e.g., unauthorized file modifications).Superagent features several robust subsystems that ensure stability, execution safety, and a seamless developer workflow:
\nandroidSetup.ts)Superagent proactively audits and prepares your local machine's developer environment:
\nripgrep (rg for high-speed workspace indexing) or curl (on Windows) is missing on your host machine, Superagent automatically downloads, extracts, and places the binaries locally in ~/.superagent-r/bin/.install.cmd for Windows) and Shell (install.sh for macOS/Linux) scripts.masterAgent.ts, superagentTools.ts, subagentTools.ts)For parallel feature development, Superagent implements a 3-tier hierarchy:
\n~/.superagent-r/worktrees/<name>), preventing file conflicts between concurrent agents.node_modules into the worktree using platform-specific symlinks (or directory junctions on Windows) instead of re-installing dependencies.define_superagent with custom system prompts, enabling domain-specific behaviors.send_message_to_superagent, permitting real-time collaboration. Multi-agent sessions are fully serialized, allowing dynamic resumption and routing of paused instances.manage_superagents with kill or kill_all actions) robustly cleans up and removes their corresponding Git worktrees, avoiding disk clutter.agentLocalStorage to track and isolate concurrent task logging paths, preventing environment variable clashes during parallel execution.~/.superagent-r/superagent.log.permissions.ts)A dedicated validation layer inspects all terminal execution commands before they are executed. It immediately blocks destructive command invocations, including:
\nrm -rf /, rmdir /s /q C:\\, etc.)Format-Volume, Initialize-Disk, mkfs)shutdown, reboot, Stop-Computer)curl/wget | sh, Invoke-Expression/iex)schedule)Superagent implements a background scheduler supporting:
\nwait: true) showing a real-time countdown indicator directly on the stdout terminal.5m or 1h).Built into the core agent loop (agent.ts), the auto-checkpoint system:
AUTO_CHECKPOINT_COOLDOWN_MS) prevents excessive checkpoint creation during rapid interactions.checkpoint_auto) appears in the terminal UI.Model configuration (model-config.json) uses atomic write operations to prevent file corruption. If the process is interrupted (e.g., Ctrl+C), the config file remains intact โ writes are first written to a temporary file and then atomically renamed, ensuring zero risk of partial/corrupt state.
\n\n[!WARNING]\nThe Chrome Extension integration and local server are currently experimental features.
\n
Superagent features a built-in REST API and Server-Sent Events (SSE) server (server.ts) that enables two-way integration with the browser via a Chrome Extension SidePanel:
superagent --server, starting an HTTP server on port 7888 (or custom port). The CLI automatically trust-checks the workspace directory initialized by the browser client.Clone and navigate into the repository:
\ngit clone <repository-url>\ncd superagent\nInstall dependencies:\nUsing npm:
\nnpm install\nOr using Bun:
\nbun install\nMake Superagent Executable Globally:\nTo install the superagent command globally on your system so you can invoke it from any directory, build the project and register it:
Using npm:
\nnpm run build\nnpm link\nOr using Bun:
\nbun run build\nbun link\nThis compiles the TypeScript files to JavaScript and registers a global symlink pointing to your local repository build. Now, you can start the assistant from any directory simply by typing:
\nsuperagent\n(To uninstall the global symlink, run npm unlink inside this directory).
If you want to use the local development version of Superagent inside another project using Bun:
\nIn the superagent repository root directory, register the package:
bun link\nIn your target project's root directory, link the registered package:
\nbun link superagent\nStart the assistant in your target project using bunx with the --bun flag (to run it fully under the Bun runtime instead of Node.js):
bunx --bun superagent\nAlternatively, you can add a script in your target project's package.json:
\"scripts\": {\n \"superagent\": \"superagent\"\n}\nAnd run it using:
\nbun run superagent\nConfigure Global API Credentials:\nSuperagent stores all configuration โ provider credentials, model settings, rate limits, and system settings โ in a centralized JSON config file at ~/.superagent-r/model-config.json. The easiest way to configure everything is through the interactive slash commands:
superagent\n# Then inside the terminal UI:\n/login # Add API keys and configure providers\n/model # Set active AI models per tier\n/settings # Configure rate limits, concurrency, streaming, etc.\nAlternatively, you can create a .env file in ~/.superagent-r/ for optional runtime overrides:
# Global model override (format: \"provider:model\" or just \"model\")\n# MODEL=openai:gpt-4o\n\n# Enable multi-agent mode via flag (or set SUPERAGENT_MULTI=true)\n# SUPERAGENT_MULTI=false\nSuperagent natively supports configuring multiple API providers concurrently. All provider profiles, API keys, and model settings are stored in ~/.superagent-r/model-config.json and managed through slash commands:
/login โ Add or update provider profiles with API keys./model โ Set active models per agent tier (Master, Superagent, Subagent)./settings โ Configure rate limits, concurrency, streaming, context window, and max iterations.Custom providers (e.g., self-hosted Claude API proxies, Ollama, vLLM) are supported via /login custom <base_url> <key>. Anthropic-compatible endpoints are automatically detected and run with the Anthropic driver.
To dynamically switch your active API provider or model at runtime, use the /login or /model slash commands. Changes take effect immediately without restarting the assistant.
\n\n[!WARNING]\nThe Chrome Extension is currently experimental and may contain bugs or incomplete features.
\n
Superagent includes a developer Chrome Extension that provides a cyberpunk-themed sidepanel interface to interact with your agent workspace directly inside the browser.
\nchrome://extensions/.chrome-extension folder located at the root of your cloned Superagent repository.superagent --server 7888\nRun the following scripts during development:
\nStart Development Mode:\nUsing npm:
\nnpm run dev\nOr using Bun:
\nbun run dev\nStart Multi-Agent Mode (3-tier orchestration):\nUsing npm:
\nnpm run dev -- --multi\nOr using Bun:
\nbun run dev --multi\n(Or globally: superagent --multi)
Start Chrome Extension API Server:\nUsing npm:
\nnpm run dev -- --server [port]\nOr using Bun:
\nbun run dev --server [port]\n(Or globally: superagent --server [port])
Resume Last Session:\nUsing npm:
\nnpm run dev -- --resume\nOr using Bun:
\nbun run dev --resume\nCompile TypeScript:\nUsing npm:
\nnpm run build\nOr using Bun:
\nbun run build\nRun Production Build:\nUsing npm:
\nnpm start\nOr using Bun:
\nbun start\nRun Unit Tests:\nUsing npm:
\nnpm test\nOr using Bun:
\nbun test\nSuperagent supports a wide range of slash commands within the terminal chat to manage session state, configure the assistant, and run commands.
\n/new: Starts a fresh conversation session. Wipes the chat history, resets agent states, and deletes temporary checkpoints./resume: Opens an interactive visual wizard listing previous session histories, allowing you to select and resume any past conversation./clear: Wipes the visual logs and terminal chat screen while maintaining the current conversation history./compact: Shows the current ContextManager status including compaction count, total tokens saved, current state, and last compaction strategy used./compact now: Forces manual compaction on demand. Displays tokens before/after, tokens saved, and the strategy used (truncation, summarization, or semantic)./pin: Pin important messages to prevent them from being removed during compaction. Pinned messages store full content, agent tags, and sync to the global knowledge store. Subcommands: /pin list (view pinned messages with metadata), /pin last (pin the last user message), /pin unpin <id> (remove a pin), /pin view <index> (view full pinned content), /pin tag <index> <label> (tag a pinned message), /pin list-messages (show all messages with indexes)./knowledge (alias: /k): Browse and search the global pinned knowledge store โ important messages pinned across ALL sessions and projects. Subcommands: /knowledge (list all entries), /knowledge <query> (search), /knowledge projects (list projects with pins)./search-history <query> (alias: /sh): Search conversation history. Add --all flag to search across ALL sessions and projects (e.g., /search-history auth login --all). Add --debug flag to display step-by-step logs of the AI semantic matching and summary generation process (e.g., /search-history auth login --debug)./compaction-history (alias: /ch): View the full audit trail of compaction events with timestamps, strategies used, tokens saved, and messages preserved./quit or /exit: Safely exits the application./checkpoint (or /checkpoint <name>): Saves a snapshot of your current conversation history, active model state, and planning states./checkpoint list (or /checkpoint with no args): Opens an interactive wizard listing all saved checkpoints with relative timestamps, message counts, and Git commit tags. From the wizard, you can restore or delete any checkpoint./checkpoint restore <id>: Restores a checkpoint by its ID. If no ID is provided, opens a pre-filtered restore wizard. Automatically terminates running subagents/tasks and reverts the agent's internal state to the checkpoint./checkpoint delete <id>: Deletes a specific checkpoint by its ID. If no ID is provided, opens a pre-filtered delete wizard for interactive selection.Ctrl+P to open the interactive checkpoint browser wizard./goal <description>: Activates Goal Mode. The assistant enters a persistent, autonomous loop (up to 200 iterations) to accomplish the goal (e.g., /goal write a full suite of unit tests for auth.ts)./init: Runs a system audit. Checks OS info, Node.js version, Git repository status, active model configuration, and auto-generates the agents.md specification file./agents: Lists all active subagents and details about the preconfigured types (researcher, coder, reviewer)./processes (or /procs): Displays active background processes managed by the agent, along with a visual progress bar and a checklist parsed from the current task.md./terminal <command>: Spawns a visible, popped-up terminal window executing the specified command./terminal preset <name> (or /terminal <preset_name>): Executes a command preset defined in your terminal-presets.json or .superagent-r/terminal-presets.json./terminal init: Launches an interactive, AI-guided wizard that scans your workspace files (like package.json, Cargo.toml, etc.), suggests relevant run commands, and writes them to .superagent-r/terminal-presets.json./skills: Displays a visual wizard containing all currently installed automation templates and guidelines./install <owner/repo>: Installs new automated developer skills directly from remote repositories via npx skills add./ih init <name> (alias: /internal-hooks init <name>): Scaffolds a new internal hook project workspace under internal-hooks/<name>/, creating hook.json (tool schema), package.json, index.js (entrypoint), and test-payload.json (dev test fixture). The new hook is automatically activated and hot-reloaded into the agent's toolset./ih dev <name>: Runs the hook's dev script (from package.json) or falls back to the command field in hook.json, piping test-payload.json as stdin. Ideal for rapid local iteration./ih active: Opens an interactive multi-select checkbox dialog listing all discovered hooks. Space to check/uncheck, Enter to confirm. The selected set is saved per-project in ~/.superagent-r/model-config.json and takes effect immediately./login: Opens a visual wizard to add API credentials, switch active providers, or list configured providers. You can also log in directly via /login <key> or /login custom <base_url> <key>./model <name>: Switches the active Large Language Model (e.g., /model openai/gpt-4o or /model google/gemini-2.5-flash). Running without arguments prints the active model name.Developed and maintained by:
\nFor guidelines on how to contribute to features and bug fixes, please see CONTRIBUTING.md.
\nThis project is licensed under the MIT License - see the LICENSE file for details.
\nCopyright (c) 2026 Rudy H. hrudy715@gmail.com
\n" + "readmeHtml": "Superagent is an interactive, terminal-based AI coding assistant designed to facilitate the cycle of development, testing, debugging, and application optimization directly from your workspace.
\nIt features a cyberpunk-styled terminal user interface built with terminal UI components, automatic tracking of model context token limits, a robust security permission layer, a 3-tier multi-agent orchestration system (Master Agent โ Superagent โ Subagent), and persistent integration with local terminal shells.
\n
In modern software development, developers frequently switch context between writing code, running terminal commands, inspecting system logs, searching documentation, and interacting with Large Language Models (LLMs).
\nSuperagent bridges this gap by providing an integrated terminal environment that understands your project's context automatically using a project specification file (agents.md), automates execution of independent tasks through secondary agents (subagents), and tracks LLM context window limits in real-time. Security is a primary design goal: every file modification, tool invocation, and shell command execution requires explicit user authorization.
Unlike standard headless execution bots or basic shell wrappers, Superagent is designed from the ground up as a fully interactive developer workspace companion:
\n/compact now to force compaction on demand, /pin to protect important messages from being compacted, and /compaction-history to audit all compaction events./knowledge to browse and search pinned knowledge across ALL sessions and projects. AI agents can access this via search_pinned_knowledge and load_pinned_session tools, enabling them to learn from previous sessions' decisions and context./checkpoint). If an experimental approach fails, you can revert back instantly to a previous checkpoint, restoring the entire session timeline.superagent --multi. Note: Multi-agent mode is currently experimental./terminal, Superagent spawns a real, popped-up host emulator terminal window. This is perfect for running interactive servers, watch scripts, and commands that require manual inputs..env or log files cluttering your project codebase. All API keys, environment settings, and session logs are kept safe and clean in your user's global directory (~/.superagent-r/).~/.superagent-r/workspace-caches/ to bypass redundant scans. It dynamically monitors and updates the cache when workspace changes occur in the agent loop. Additionally, git worktree directories created for Superagents (~/.superagent-r/worktrees/<name>) are automatically configured as trusted (safe.directory) in git to prevent \"dubious ownership\" warnings and errors.ask_question tool at every decision point โ choosing implementations, resolving ambiguity, or selecting approaches โ ensuring the AI never guesses or assumes on the user's behalf./terminal init wizard.D:\\images\\screenshot.png) directly into the prompt to auto-attach it. Press Ctrl+V to automatically capture image binary data from your system clipboard (cross-platform support for Windows, macOS, and Linux). Attached images are rendered in a sleek visual queue above the prompt and transmitted as high-fidelity multimodal inputs to vision-capable models (e.g. Claude 3.5 Sonnet, GPT-4o), with automatic token tracking.internal-hooks/<name>/ with a hook.json schema definition and an index.js entrypoint. Hooks are auto-discovered on startup and registered as first-class agent tools. Use /ih init <name> to scaffold the project, /ih dev <name> to run and test it locally, and /ih active to pick which hooks are active via an interactive multi-select checkbox dialog. Active selections are persisted per-project in ~/.superagent-r/model-config.json.Superagent is built on modern Node.js technologies for high performance and modular architecture:
\nai, @ai-sdk/openai, @ai-sdk/anthropic) for structured and streaming interactions.superagent/\nโโโ src/\nโ โโโ cli.tsx # Main entrypoint; routes --multi flag to masterAgent\nโ โโโ app.tsx # React UI wrapper and command handling logic\nโ โโโ core/\nโ โ โโโ agent.ts # Core cognitive loop and instruction runner\nโ โ โโโ masterAgent.ts # Master Agent orchestrator (3-tier entry point)\nโ โ โโโ config.ts # Environment variable and global config management\nโ โ โโโ checkpoints.ts # Conversation state checkpoint save/load logic\nโ โ โโโ slash-commands.ts # Interactive command definitions\nโ โ โโโ tools/\nโ โ โโโ types.ts # Shared types: AgentTier, SubagentInstance, ToolSet\nโ โ โโโ toolsets.ts # ToolSet definitions per tier (master/super/sub)\nโ โ โโโ prompts.ts # System prompts per tier with dynamic context\nโ โ โโโ state.ts # Shared subagent registry and event emitters\nโ โ โโโ shellTools.ts # Command execution and background task control\nโ โ โโโ systemTools.ts # File operations, directory creation, port checks\nโ โ โโโ subagentTools.ts # Subagent instantiation (superagent tier)\nโ โ โโโ superagentTools.ts # Superagent orchestration tools (master tier)\nโ โ โโโ dynamicHooks.ts # Internal hook discovery, loading, and active state\nโ โ โโโ academicSearchTools.ts # Academic journal search engine API integrations\nโ โ โโโ networkTools.ts # Web content fetch and browser integration\nโ โโโ components/ # React Ink components (visual stats, wizards)\nโโโ bin/ # Portable Python and setup scripts\nโโโ tests/ # Unit test suites using Vitest\nโโโ package.json # Project manifest and scripts\nA rich terminal interface showing live statistics on active prompt sizes, completion token counts, token cost summaries, active models, remaining context windows, and real-time model generation speed (tokens per second).
\nAllows developers to save the current state of a coding conversation and restore it at any point using /checkpoint save <name> and /checkpoint restore <id>. This allows you to safely experiment with different implementations. Checkpoints can be browsed, restored, or deleted via an interactive wizard (launched by /checkpoint, /checkpoint list, or Ctrl+P in multi-agent mode). Use the --resume or -r flag to continue where you left off. Multi-agent sessions are fully serialized, ensuring smooth restore and resume of running tasks and interactive prompts. Auto-checkpointing creates snapshots automatically on every user message and before destructive tool operations, with a cooldown to prevent excessive saves โ ensuring you always have a safe rollback point.
--multi) (Experimental)\n\n[!WARNING]\nMulti-agent mode (
\n--multi) is currently experimental and not recommended for production environments.
Launch with superagent --multi to activate the full 3-tier hierarchy:
superagent --multi\n โ\n Master Agent (orchestrator tier)\n Tools: invoke_superagent, await_superagents, merge_superagents, manage_superagents, define_superagent, send_message_to_superagent, manage_subagents, git_worktree\n Spawns Superagents with git worktree isolation\n โ\n Superagent (per-feature coordinator/lead)\n Tools: shell + file tools, invoke_subagent, manage_subagents, git_worktree\n Isolated in its own git worktree\n Can spawn Subagents for atomic ops\n โ\n Subagent (atomic operation tier)\n Tools: file tools only (read/write/search)\n Ephemeral, single-purpose execution\nTier Responsibilities:
\nAdvanced Orchestration Features:
\nโโโ and โโโ connectors) mapping active relationships between the Master Agent, spawned Superagents, and running subagents/tasks.ViolationRecord event is emitted with severity levels (warning or critical). These violations automatically propagate up the agent hierarchy (Subagent โ Superagent โ Master Agent), enabling parent agents to track, log, and take corrective action โ including halting or redirecting the offending agent.ask_question tool at every decision point. Agents must present multiple-choice options to the user rather than guessing, assuming, or making unilateral decisions about implementation approaches, file selections, or design trade-offs.Standard subagent roles (single-agent mode):
\n/terminal)Runs development servers, local builds, or test watchers in popped-up, visible OS terminal windows (Windows cmd, macOS Terminal, Linux x-terminal). It includes an AI-assisted preset initializer (/terminal init) to auto-configure workspace command presets.
For complex changes, the Master Agent enforces structured planning and execution boundaries:
\nconstraints (files or logic NOT to modify) and acceptanceCriteria (a checklist of specific test cases to satisfy).implementation_plan.md and requires explicit user review and approval before any execution begins, guaranteeing complete oversight.The merge system uses a safe-by-default strategy that prevents file corruption:
\n<<<<<<< in files)mode: 'patch'): Lightweight mode that skips worktree creation and operates directly in the parent's working directory. Ideal for small, targeted fixes (1-2 lines). Much faster than spawning a full Superagent. Includes safety warnings if the parent worktree has uncommitted changes.baseBranch: 'feat/...'): When a Superagent needs to build on top of another feature branch instead of the current HEAD, specify baseBranch to create the worktree from that branch. Useful for dependent features or building on top of in-progress work.Example:
\ninvoke_superagent({\n role: 'fix-html-corrupt',\n task: 'Fix duplicate closing tags in Toolbar component',\n branch: 'fix/toolbar-html',\n baseBranch: 'feat/separate-compressor-menu', // Build on top of this branch\n mode: 'patch' // Quick fix, no worktree needed\n})\nAll agent operations, including single-agent and 3-tier multi-agent processes, are dynamically logged to a central log file in the user's home directory (~/.superagent-r/superagent.log). The log maintains tier-aware indentation to cleanly trace parallel execution branches.
Beyond manual /checkpoint commands, Superagent creates checkpoints automatically:
In multi-agent mode, the permission layer emits structured ViolationRecord events whenever a child agent attempts a blocked operation. These violations include:
\"warning\" for soft blocks (e.g., scope creep attempts) and \"critical\" for hard policy violations (e.g., unauthorized file modifications).Superagent features several robust subsystems that ensure stability, execution safety, and a seamless developer workflow:
\nandroidSetup.ts)Superagent proactively audits and prepares your local machine's developer environment:
\nripgrep (rg for high-speed workspace indexing) or curl (on Windows) is missing on your host machine, Superagent automatically downloads, extracts, and places the binaries locally in ~/.superagent-r/bin/.install.cmd for Windows) and Shell (install.sh for macOS/Linux) scripts.masterAgent.ts, superagentTools.ts, subagentTools.ts)For parallel feature development, Superagent implements a 3-tier hierarchy:
\n~/.superagent-r/worktrees/<name>), preventing file conflicts between concurrent agents.node_modules into the worktree using platform-specific symlinks (or directory junctions on Windows) instead of re-installing dependencies.define_superagent with custom system prompts, enabling domain-specific behaviors.send_message_to_superagent, permitting real-time collaboration. Multi-agent sessions are fully serialized, allowing dynamic resumption and routing of paused instances.manage_superagents with kill or kill_all actions) robustly cleans up and removes their corresponding Git worktrees, avoiding disk clutter.agentLocalStorage to track and isolate concurrent task logging paths, preventing environment variable clashes during parallel execution.~/.superagent-r/superagent.log.permissions.ts)A dedicated validation layer inspects all terminal execution commands before they are executed. It immediately blocks destructive command invocations, including:
\nrm -rf /, rmdir /s /q C:\\, etc.)Format-Volume, Initialize-Disk, mkfs)shutdown, reboot, Stop-Computer)curl/wget | sh, Invoke-Expression/iex)schedule)Superagent implements a background scheduler supporting:
\nwait: true) showing a real-time countdown indicator directly on the stdout terminal.5m or 1h).Built into the core agent loop (agent.ts), the auto-checkpoint system:
AUTO_CHECKPOINT_COOLDOWN_MS) prevents excessive checkpoint creation during rapid interactions.checkpoint_auto) appears in the terminal UI.Model configuration (model-config.json) uses atomic write operations to prevent file corruption. If the process is interrupted (e.g., Ctrl+C), the config file remains intact โ writes are first written to a temporary file and then atomically renamed, ensuring zero risk of partial/corrupt state.
\n\n[!WARNING]\nThe Chrome Extension integration and local server are currently experimental features.
\n
Superagent features a built-in REST API and Server-Sent Events (SSE) server (server.ts) that enables two-way integration with the browser via a Chrome Extension SidePanel:
superagent --server, starting an HTTP server on port 7888 (or custom port). The CLI automatically trust-checks the workspace directory initialized by the browser client.Clone and navigate into the repository:
\ngit clone <repository-url>\ncd superagent\nInstall dependencies:\nUsing npm:
\nnpm install\nOr using Bun:
\nbun install\nMake Superagent Executable Globally:\nTo install the superagent command globally on your system so you can invoke it from any directory, build the project and register it:
Using npm:
\nnpm run build\nnpm link\nOr using Bun:
\nbun run build\nbun link\nThis compiles the TypeScript files to JavaScript and registers a global symlink pointing to your local repository build. Now, you can start the assistant from any directory simply by typing:
\nsuperagent\n(To uninstall the global symlink, run npm unlink inside this directory).
If you want to use the local development version of Superagent inside another project using Bun:
\nIn the superagent repository root directory, register the package:
bun link\nIn your target project's root directory, link the registered package:
\nbun link superagent\nStart the assistant in your target project using bunx with the --bun flag (to run it fully under the Bun runtime instead of Node.js):
bunx --bun superagent\nAlternatively, you can add a script in your target project's package.json:
\"scripts\": {\n \"superagent\": \"superagent\"\n}\nAnd run it using:
\nbun run superagent\nConfigure Global API Credentials:\nSuperagent stores all configuration โ provider credentials, model settings, rate limits, and system settings โ in a centralized JSON config file at ~/.superagent-r/model-config.json. The easiest way to configure everything is through the interactive slash commands:
superagent\n# Then inside the terminal UI:\n/login # Add API keys and configure providers\n/model # Set active AI models per tier\n/settings # Configure rate limits, concurrency, streaming, etc.\nAlternatively, you can create a .env file in ~/.superagent-r/ for optional runtime overrides:
# Global model override (format: \"provider:model\" or just \"model\")\n# MODEL=openai:gpt-4o\n\n# Enable multi-agent mode via flag (or set SUPERAGENT_MULTI=true)\n# SUPERAGENT_MULTI=false\nSuperagent natively supports configuring multiple API providers concurrently. All provider profiles, API keys, and model settings are stored in ~/.superagent-r/model-config.json and managed through slash commands:
/login โ Add or update provider profiles with API keys./model โ Set active models per agent tier (Master, Superagent, Subagent)./settings โ Configure rate limits, concurrency, streaming, context window, and max iterations.Custom providers (e.g., self-hosted Claude API proxies, Ollama, vLLM) are supported via /login custom <base_url> <key>. Anthropic-compatible endpoints are automatically detected and run with the Anthropic driver.
To dynamically switch your active API provider or model at runtime, use the /login or /model slash commands. Changes take effect immediately without restarting the assistant.
\n\n[!WARNING]\nThe Chrome Extension is currently experimental and may contain bugs or incomplete features.
\n
Superagent includes a developer Chrome Extension that provides a cyberpunk-themed sidepanel interface to interact with your agent workspace directly inside the browser.
\nchrome://extensions/.chrome-extension folder located at the root of your cloned Superagent repository.superagent --server 7888\nRun the following scripts during development:
\nStart Development Mode:\nUsing npm:
\nnpm run dev\nOr using Bun:
\nbun run dev\nStart Multi-Agent Mode (3-tier orchestration):\nUsing npm:
\nnpm run dev -- --multi\nOr using Bun:
\nbun run dev --multi\n(Or globally: superagent --multi)
Start Chrome Extension API Server:\nUsing npm:
\nnpm run dev -- --server [port]\nOr using Bun:
\nbun run dev --server [port]\n(Or globally: superagent --server [port])
Resume Last Session:\nUsing npm:
\nnpm run dev -- --resume\nOr using Bun:
\nbun run dev --resume\nCompile TypeScript:\nUsing npm:
\nnpm run build\nOr using Bun:
\nbun run build\nRun Production Build:\nUsing npm:
\nnpm start\nOr using Bun:
\nbun start\nRun Unit Tests:\nUsing npm:
\nnpm test\nOr using Bun:
\nbun test\nSuperagent supports a wide range of slash commands within the terminal chat to manage session state, configure the assistant, and run commands.
\n/new: Starts a fresh conversation session. Wipes the chat history, resets agent states, and deletes temporary checkpoints./resume: Opens an interactive visual wizard listing previous session histories, allowing you to select and resume any past conversation./clear: Wipes the visual logs and terminal chat screen while maintaining the current conversation history./compact: Shows the current ContextManager status including compaction count, total tokens saved, current state, and last compaction strategy used./compact now: Forces manual compaction on demand. Displays tokens before/after, tokens saved, and the strategy used (truncation, summarization, or semantic)./pin: Pin important messages to prevent them from being removed during compaction. Pinned messages store full content, agent tags, and sync to the global knowledge store. Subcommands: /pin list (view pinned messages with metadata), /pin last (pin the last user message), /pin unpin <id> (remove a pin), /pin view <index> (view full pinned content), /pin tag <index> <label> (tag a pinned message), /pin list-messages (show all messages with indexes)./knowledge (alias: /k): Browse and search the global pinned knowledge store โ important messages pinned across ALL sessions and projects. Subcommands: /knowledge (list all entries), /knowledge <query> (search), /knowledge projects (list projects with pins)./search-history <query> (alias: /sh): Search conversation history. Add --all flag to search across ALL sessions and projects (e.g., /search-history auth login --all). Add --debug flag to display step-by-step logs of the AI semantic matching and summary generation process (e.g., /search-history auth login --debug)./compaction-history (alias: /ch): View the full audit trail of compaction events with timestamps, strategies used, tokens saved, and messages preserved./quit or /exit: Safely exits the application./checkpoint (or /checkpoint <name>): Saves a snapshot of your current conversation history, active model state, and planning states./checkpoint list (or /checkpoint with no args): Opens an interactive wizard listing all saved checkpoints with relative timestamps, message counts, and Git commit tags. From the wizard, you can restore or delete any checkpoint./checkpoint restore <id>: Restores a checkpoint by its ID. If no ID is provided, opens a pre-filtered restore wizard. Automatically terminates running subagents/tasks and reverts the agent's internal state to the checkpoint./checkpoint delete <id>: Deletes a specific checkpoint by its ID. If no ID is provided, opens a pre-filtered delete wizard for interactive selection.Ctrl+P to open the interactive checkpoint browser wizard./goal <description>: Activates Goal Mode. The assistant enters a persistent, autonomous loop (up to 200 iterations) to accomplish the goal (e.g., /goal write a full suite of unit tests for auth.ts)./init: Runs a system audit. Checks OS info, Node.js version, Git repository status, active model configuration, and auto-generates the agents.md specification file./agents: Lists all active subagents and details about the preconfigured types (researcher, coder, reviewer)./processes (or /procs): Displays active background processes managed by the agent, along with a visual progress bar and a checklist parsed from the current task.md./terminal <command>: Spawns a visible, popped-up terminal window executing the specified command./terminal preset <name> (or /terminal <preset_name>): Executes a command preset defined in your terminal-presets.json or .superagent-r/terminal-presets.json./terminal init: Launches an interactive, AI-guided wizard that scans your workspace files (like package.json, Cargo.toml, etc.), suggests relevant run commands, and writes them to .superagent-r/terminal-presets.json./skills: Displays a visual wizard containing all currently installed automation templates and guidelines./install <owner/repo>: Installs new automated developer skills directly from remote repositories via npx skills add./ih init <name> (alias: /internal-hooks init <name>): Scaffolds a new internal hook project workspace under internal-hooks/<name>/, creating hook.json (tool schema), package.json, index.js (entrypoint), and test-payload.json (dev test fixture). The new hook is automatically activated and hot-reloaded into the agent's toolset./ih dev <name>: Runs the hook's dev script (from package.json) or falls back to the command field in hook.json, piping test-payload.json as stdin. Ideal for rapid local iteration./ih active: Opens an interactive multi-select checkbox dialog listing all discovered hooks. Space to check/uncheck, Enter to confirm. The selected set is saved per-project in ~/.superagent-r/model-config.json and takes effect immediately./login: Opens a visual wizard to add API credentials, switch active providers, or list configured providers. You can also log in directly via /login <key> or /login custom <base_url> <key>./model <name>: Switches the active Large Language Model (e.g., /model openai/gpt-4o or /model google/gemini-2.5-flash). Running without arguments prints the active model name.Developed and maintained by:
\nFor guidelines on how to contribute to features and bug fixes, please see CONTRIBUTING.md.
\nThis project is licensed under the MIT License - see the LICENSE file for details.
\nCopyright (c) 2026 Rudy H. hrudy715@gmail.com
\n" }, { "fullName": "muslimtify-org/muslimtify", diff --git a/src/data/revival.json b/src/data/revival.json index 65ddeef..1138a76 100644 --- a/src/data/revival.json +++ b/src/data/revival.json @@ -81,13 +81,13 @@ "description": "", "url": "https://github.com/MBenedictt/JudolSlayerProject", "homepage": "", - "stars": 102, + "stars": 101, "forks": 26, "language": "JavaScript", "topics": [], "license": null, "createdAt": "2025-04-04T15:50:35Z", - "updatedAt": "2026-07-17T04:19:14Z", + "updatedAt": "2026-07-22T14:01:50Z", "pushedAt": "2025-04-17T01:39:08Z", "archived": false, "disabled": false From ae867da0be027cd8c68cd45d59550f517297699c Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Wed, 22 Jul 2026 16:53:07 +0000 Subject: [PATCH 19/25] Sync content data --- src/data/projects.json | 78 +++++++++++++++++++++--------------------- src/data/revival.json | 4 +-- 2 files changed, 41 insertions(+), 41 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index 2b6342a..baf92fe 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -140,7 +140,7 @@ "archived": false, "licenseSpdx": "", "createdAt": "2016-05-21T10:55:38Z", - "openIssues": 353, + "openIssues": 355, "openPullRequests": 7, "subscribers": 110, "communityHealth": 50, @@ -313,8 +313,8 @@ "token-reduction", "token-savings" ], - "updatedAt": "2026-07-22T14:01:45Z", - "pushedAt": "2026-07-22T13:58:48Z", + "updatedAt": "2026-07-22T16:36:35Z", + "pushedAt": "2026-07-22T16:35:08Z", "latestRelease": { "name": "v0.6.3", "tagName": "v0.6.3", @@ -324,11 +324,11 @@ "archived": false, "licenseSpdx": "MIT", "createdAt": "2026-03-15T03:38:04Z", - "openIssues": 20, + "openIssues": 21, "openPullRequests": 1, "subscribers": 4, "communityHealth": 71, - "readmeHtml": "\n Noise-canceling context and long-term memory for your AI agent โ lossy, but always reversible, and it never fabricates a result. Stop paying Claude to read 10,000 lines of terminal noise.\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Always reversible ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Every other compressor asks you to trust that what it cut didn't matter. OMNI doesn't ask โ it guarantees, and each guarantee is backed by code you can read:
\n| Guarantee | \nHow | \nProof | \n
|---|---|---|
| Get the original back, byte-for-byte | \neverything cut is archived in a local SQLite RewindStore (SHA-256 โ content); the agent gets a hash and calls omni_retrieve | \nHow it works | \n
| Never fabricates a result | \na distiller that parsed no signal returns the raw output, never a green no errors / passed string | \n#143 | \n
| Failures are never masked | \na command that exits non-zero passes through verbatim | \n#120 | \n
| Structured data is never touched | \nJSON / YAML / NDJSON / CSV pass through byte-for-byte | \npipeline::format | \n
| Numbers are measured, not aspirational | \n1,810 real traces replayed on the release binary โ and 63.6% of calls net zero, which we publish too | \nBenchmarks | \n
That is the one thing a bigger compression number can't buy: you can always recover the original, and it will never lie to your agent.
\nProblem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. omni recall surfaces the exact solution in under 10ms.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust for imperceptible latency.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
No. OMNI is written in Rust and executes the distillation pipeline in under 10ms.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
\n Noise-canceling context and long-term memory for your AI agent โ lossy, but always reversible, and it never fabricates a result. Stop paying Claude to read 10,000 lines of terminal noise.\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Always reversible ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Every other compressor asks you to trust that what it cut didn't matter. OMNI doesn't ask โ it guarantees, and each guarantee is backed by code you can read:
\n| Guarantee | \nHow | \nProof | \n
|---|---|---|
| Get the original back, byte-for-byte | \neverything cut is archived in a local SQLite RewindStore (SHA-256 โ content); the agent gets a hash and calls omni_retrieve | \nHow it works | \n
| Never fabricates a result | \na distiller that parsed no signal returns the raw output, never a green no errors / passed string | \n#143 | \n
| Failures are never masked | \na command that exits non-zero passes through verbatim | \n#120 | \n
| Structured data is never touched | \nJSON / YAML / NDJSON / CSV pass through byte-for-byte | \npipeline::format | \n
| Numbers are measured, not aspirational | \n1,810 real traces replayed on the release binary โ and 63.6% of calls net zero, which we publish too | \nBenchmarks | \n
That is the one thing a bigger compression number can't buy: you can always recover the original, and it will never lie to your agent.
\nProblem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. The agent surfaces it through the omni_recall MCP tool before it repeats the mistake.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust, though the end-to-end cost is not zero.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
Yes, measurably, and the cost grows with your history. The distillation pipeline itself runs in single-digit milliseconds, but every hooked command also writes to the local RewindStore: a 496-byte git status takes ~82 ms against a fresh database and ~308 ms against a 97 MB one, and a 16.5 KB cargo test takes ~276 ms. Budget for it. OMNI_PASSTHROUGH=1 skips the pipeline entirely when you need the raw output back.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
\n \n
Give your AI a memory that never leaves your machine.
\n\n Your AI forgets everything between sessions. Uteke fixes that โ one binary, fully offline, ~45ms recall.\n
\n \n \n \n \n \n
\n ๐ฌ๐ง English ยท ๐ฎ๐ฉ Bahasa Indonesia\n
# Install (macOS, Linux, Windows)\ncurl -sSL codecora.dev/install | sh\n\n# Store a memory\nuteke remember \"Deploy v2.1 to staging at 3pm\"\n\n# Search it back โ by meaning, not just keywords\nuteke recall \"when do we deploy?\"\nThat's it. No API keys. No Docker. No Python. No cloud.
\nFirst run downloads the embedding model (~188MB, one-time) and you're running.
\nWant richer memories? Add metadata:
\nuteke remember \"Deploy v2.1 to staging\" \\\n --tags deploy,staging \\\n --entity staging-server \\\n --category infrastructure\n| Method | \nCommand | \n
|---|---|
| Homebrew | \nbrew install codecoradev/tap/uteke | \n
| Cargo | \ncargo install uteke-cli | \n
| Docker | \ndocker run -d -p 127.0.0.1:8767:8767 -v uteke-data:/data ghcr.io/codecoradev/uteke:latest | \n
| Binary | \nGitHub Releases (macOS, Linux, Windows) | \n
๐ Full install guide ยท Docker docs
\n\n\nAre you an AI agent reading this? Your user just shared this repo link with you.\nRun
\nuteke onboardto interactively configure uteke for your environment โ it will\ndetect your install, ask which agent you are, toggle features, write config, and\nshow you every command available. If non-interactive:uteke onboard --yes --agent <your-agent>.
New to uteke? One command takes you from zero to productive:
\nuteke onboard\nThe wizard will:
\n~/.uteke/uteke.toml with your selectionsuteke init)Non-interactive (CI, scripts, AI agents):
\nuteke onboard --yes --agent hermes --namespace default\n๐ Full onboarding docs ยท CLI reference
\nYou just spent 2 hours explaining your codebase to ChatGPT. Next session? Blank slate. Again.
\nEvery AI tool forgets. Context windows fill up, sessions end, and your AI starts over every single time. Uteke gives it persistent memory โ and keeps it on your machine.
\n| \n | Uteke | \nMnemosyne | \nMem0 | \nAgentMemory | \nLetta | \nZep | \nEngram | \n
|---|---|---|---|---|---|---|---|
| Language | \nRust (single binary) | \nPython (pip) | \nPython | \nTypeScript | \nPython | \nPython | \nGo (single binary) | \n
| Setup | \nOne binary (curl | sh) | \npip install + venv | \npip + Docker + Qdrant | \nnpm + Docker (iii-engine) | \npip + Docker + Postgres | \npip + Docker + Neo4j | \nOne binary | \n
| API keys | \nโ None | \nโ ๏ธ For remote embeddings | \nโ OpenAI/LLM | \nโ LLM key | \nโ LLM key | \nโ LLM key | \nโ None | \n
| Works offline | \nโ Fully | \nโ ๏ธ Optional | \nโ Cloud embedding | \nโ Needs LLM | \nโ Needs LLM | \nโ Needs LLM + vector DB | \nโ Fully | \n
| Search | \nHybrid (Vector + FTS5 + RRF) | \nsqlite-vec + FTS5 | \nVector + Graph | \nVector + Graph | \nVector | \nTemporal Graph | \nFTS5 only | \n
| Recall speed | \n~45ms | \n~50ms+ | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \nNetwork round-trip | \n~Fast (local) | \n
| Multi-agent | \nโ Rooms (built-in collaboration) | \nโ ๏ธ Shared API | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Time-travel | \nโ Native point-in-time | \nโ ๏ธ Temporal triples | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| MCP server | \nโ JSON-RPC + HTTP | \nโ stdio + SSE | \nโ | \nโ | \nโ | \nโ | \nโ | \n
| Your data | \nโ Never leaves machine | \nโ Local-first | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ ๏ธ Sent to LLM cloud | \nโ Local | \n
| License | \nApache 2.0 | \nMIT | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \nApache 2.0 | \n
\n\nUteke vs Mnemosyne: Both are local-first with semantic + FTS5 search. Mnemosyne is the closest competitor (~1.5K stars, Python). Uteke wins on single binary (no Python runtime), rooms, time-travel queries, and zero runtime dependencies.
\n
\n\nUteke vs Engram: Both are single-binary, offline, no-API-key tools. But Engram is FTS5-only (keyword search). Uteke adds vector semantic search + RRF fusion + rooms + time-travel + graph relationships + smart decay + document engine + batch import. Same simplicity thesis, 10ร the features.
\n
\n\nUteke vs AgentMemory/Mem0/Letta/Zep: Those are powerful โ but all require cloud LLM API keys and Docker infrastructure. Your data goes to OpenAI/Anthropic. Uteke runs fully offline with local ONNX embeddings. No Docker, no Python, no API keys.
\n
\n \n
๐ค Building AI agents? Give them persistent memory without cloud dependencies. Your agent remembers user preferences, past decisions, and context โ across sessions, fully offline.
\n๐ฅ Working in a team? Use Rooms to share knowledge. Meeting notes, project decisions, architecture choices โ searchable by everyone, attributed by author.
\n๐ Building for privacy-sensitive domains? Healthcare, finance, legal โ data stays on your machine. No API calls, no telemetry, no cloud. Local embeddings (ONNX, 768d).
\nโจ๏ธ Power user who lives in the terminal? Uteke is your personal knowledge graph. Remember anything, recall by meaning, link related thoughts. All from the command line.
\n| Feature | \nWhat it does | \n
|---|---|
| ๐ง Hybrid Search | \nVector similarity + FTS5 full-text search, merged by Reciprocal Rank Fusion (RRF). Finds by meaning AND exact keywords. | \n
| ๐ Rooms | \nGroup memories by context (meetings, projects, clients) with author attribution. | \n
| โณ Time-travel | \nRecall memories as they existed at any point in time. uteke recall \"deploy\" --at 2025-01-15 | \n
| ๐ท๏ธ Rich Metadata | \nTags, entities, categories, key:value pairs on every memory. | \n
| ๐งฉ Memory Types | \nTyped categories (fact, procedure, decision, etc.) with auto-inference. | \n
| โ๏ธ Partial Updates | \nUpdate content, tags, metadata, importance, or type without full rewrite. | \n
| ๐ Citations | \nSource attribution on every memory (URL, file, user, import batch). | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ Relationship Graph | \nLink memories with typed edges (supersedes, contradicts, references). Auto-backlinks. | \n
| ๐ Cross-Entity Linking | \nBidirectional memoryโdocument references via [[doc-slug]] wikilinks. | \n
| ๐ค Cosine Auto-Linking | \nAutomatically creates similar_to edges between related memories. | \n
| ๐ Smart Decay | \nComposite importance scoring. Pin what matters, let stale memories fade. | \n
| ๐ Salience + Recency | \nDual-axis recall boost by memory type and age. | \n
| ๐ Orphan Detection | \nFind disconnected, low-importance memories for cleanup. | \n
| ๐ Dream Cycle | \nOne-command maintenance: lint โ backlinks โ dedup โ orphans. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ MCP Server | \nJSON-RPC over stdio + Streamable HTTP. Works with Claude Code, Cursor, Hermes. | \n
| ๐ฅ๏ธ Server Mode | \nPersistent daemon โ eliminates cold-start embedding load on every call. | \n
| ๐ Batch Import | \nImport entire directories with auto-strategy routing (document vs. memory extraction). | \n
| ๐ Document Engine | \nWiki/knowledge base with uteke doc create/get/list and auto-chunking. | \n
| ๐ฅ Import/Export | \nJSONL-based backup and restore. | \n
| ๐ View-Only API Keys | \nRead-only tokens for safe GET-only access to the server. | \n
| Feature | \nWhat it does | \n
|---|---|
| ๐ฆ Single Binary | \nZero dependencies. No Docker needed, no database server, no Python, no API keys. | \n
| ๐ Fully Offline | \nLocal ONNX embeddings (EmbeddingGemma Q4, 768d). No telemetry, no cloud. | \n
| โก Recall Cache | \nLRU cache eliminates redundant embedding for repeated queries. | \n
| ๐ฅ Tiered Memory | \nHot/Warm/Cold tracking with auto-cleanup of stale memories. | \n
| ๐ Embed Fallback | \nGracefully degrades to no-op embedder if local model fails (never crashes). | \n
| ๐ฅ Multi-Agent Namespaces | \nFully isolated memory per agent, zero overhead. | \n
| ๐ Benchmarks | \nBuilt-in uteke bench for perf testing. See results. | \n
// .mcp.json (Claude Code, Cursor)\n{ \"mcpServers\": { \"uteke\": { \"command\": \"uteke-mcp\" } } }\nFor Claude Desktop, Hermes, and HTTP transport, see MCP docs.
\n๐ Full documentation ยท CLI reference ยท Configuration
\ngraph LR\n Input[User Query] --> Embed[Local ONNX Embedder<br/>768d, EmbeddingGemma Q4]\n Embed --> HNSW[HNSW Vector Index<br/>usearch]\n Embed --> FTS5[FTS5 Full-Text<br/>SQLite]\n HNSW --> RRF[Reciprocal Rank Fusion<br/>k=60]\n FTS5 --> RRF\n RRF --> Results[Ranked Results]\n\n style Input fill:#4a9eff,color:#fff\n style Results fill:#4aff9e,color:#000\n style RRF fill:#ff9e4a,color:#fff\nHow hybrid search works:
\nEverything runs in-process. No network. No cloud. No server required (unless you want server mode).
\n\n \n
Mem0 and Letta are great โ but they require cloud API keys (OpenAI/LLM) and external infrastructure (Docker, Postgres, Qdrant). Your data gets sent to a cloud LLM provider. Uteke is a single binary with zero API keys. All embeddings run locally via ONNX. Your data never leaves your machine. See comparison table.
\n\nHow is Uteke different from AgentMemory?AgentMemory (25K stars) is a TypeScript/Node.js platform with 53 MCP tools and 12 auto-hooks. It's feature-rich but requires Docker + the iii-engine + LLM API keys. Uteke is Rust, zero dependencies, and works fully offline. If you want maximum integrations and don't mind cloud dependency โ AgentMemory. If you want privacy, speed, and zero setup โ Uteke.
\n\nHow is Uteke different from Engram?Engram (2.4K stars, Go) shares our philosophy: single binary, zero deps, MCP server, local-first. The key difference is search: Engram uses FTS5 only (keyword matching). Uteke uses hybrid search (HNSW vector similarity + FTS5 + Reciprocal Rank Fusion) โ meaning you can search by meaning, not just exact words. Uteke also adds rooms, time-travel, graph relationships, smart decay, document engine, and batch import.
\n\nWhat can Uteke remember?Anything text-based: decisions, meeting notes, code snippets, project context, personal notes, agent state. You can tag, categorize, and link memories. The --batch-dir flag lets you import entire document directories.
Yes. The embedding model (EmbeddingGemma Q4, 768d) downloads once (~188MB) on first run. After that, zero network calls. No telemetry. If the local model fails, Uteke degrades gracefully to a no-op embedder โ it never crashes and never calls a cloud API.
\n\nHow fast is recall?~45ms as a library (measured at 100โ10K memories). No network round-trip because everything is local. The LRU recall cache eliminates redundant embedding computation for repeated queries.
\n\nCan I use Uteke with my existing AI tools?Yes. Uteke ships with an MCP server that works with Claude Code, Cursor, and Hermes. You can also use the HTTP API directly in any language. See MCP setup โ
\n\nIs it production-ready?Uteke is at v0.7.3 with 206 tests, CI/CD on every commit, and benchmark harness. It's used in production by the CodeCora team and other early adopters. Still in 0.x โ expect rough edges, but the core is stable.
\ncargo build --workspace # Build\ncargo test --workspace # Test (206 tests)\ncargo clippy -- -D warnings # Lint\ncargo fmt # Format\nContributions welcome! Read CONTRIBUTING.md for the full guide.
\nApache License 2.0 โ use it, fork it, ship it.
\n\n Found this useful? โญ Star this repo โ it helps others discover Uteke.\n
\n\n \n
A free, offline-first, open source Point of Sale (POS) Progressive Web App built for Indonesian Micro, Small, and Medium Enterprises (UMKM). All data is stored locally on the user's device โ no server, no registration, no cost.
\n| Layer | \nTechnology | \n
|---|---|
| Framework | \nReact 18 + TypeScript | \n
| Build Tool | \nVite | \n
| Styling | \nTailwind CSS + shadcn/ui | \n
| Theming | \nnext-themes (dark mode) | \n
| Database | \nIndexedDB via Dexie.js | \n
| Charts | \nRecharts | \n
| Routing | \nReact Router DOM v6 | \n
| Forms & Validation | \nReact Hook Form + Zod | \n
| State | \n@tanstack/react-query | \n
| Icons | \nLucide React | \n
| i18n | \ni18next + react-i18next | \n
| Date | \ndate-fns (id, en-US, ms locales) | \n
| PWA | \nvite-plugin-pwa (Workbox) | \n
| Barcode | \nhtml5-qrcode (camera scanner + manual input) | \n
| Receipt | \nhtml2canvas (to PNG), Web Bluetooth Print (PWA), Bluetooth Classic (Android APK via Capacitor) | \n
| Font | \nPlus Jakarta Sans | \n
| Native Wrapper | \nCapacitor 8 (Android) | \n
# Clone the repository\ngit clone https://github.com/user/kasirgratisan.git\ncd kasirgratisan\n\n# Install dependencies\nnpm install\n\n# Start the development server\nnpm run dev\nThe app will be running at http://localhost:8080.
npm run build\nnpm run preview\nThis project can also run as a native Android app using Capacitor while keeping the PWA/web version working from the same codebase.
\nRequirements:
\nmacOS / Linux:
\nexport JAVA_HOME=\"/Applications/Android Studio.app/Contents/jbr/Contents/Home\"\nWindows (PowerShell):
\n$env:JAVA_HOME = \"C:\\Program Files\\Android\\Android Studio\\jbr\"\n\n\nAdjust the path if your Android Studio is installed elsewhere.
\n
npm run build\nnpx cap sync android\ncd android\n./gradlew assembleDebug\nOutput: android/app/build/outputs/apk/debug/app-debug.apk
npm run build\nnpx cap sync android\ncd android\n./gradlew bundleRelease\nOutput: android/app/build/outputs/bundle/release/app-release.aab
\n\nThe release AAB must be signed before uploading to Google Play. See Android signing docs.
\n
npm run cap:sync # build web bundle and sync Capacitor\nnpm run cap:android # build, sync, then open Android Studio\nnpm run cap:run # build, sync, then run on connected Android device/emulator\nsrc/\nโโโ App.tsx # Root component & routing\nโโโ main.tsx # Entry point\nโโโ index.css # Design tokens (HSL CSS variables)\nโโโ lib/\nโ โโโ db.ts # Dexie database schema, interfaces, seed data\nโ โโโ auth.ts # Multi-user auth helpers (PIN hashing, sessions, validation)\nโ โโโ utils.ts # Utility functions (cn, etc.)\nโ โโโ image-utils.ts # Image compression utility\nโ โโโ version-check.ts # Version check webhook\nโโโ components/\nโ โโโ layout/\nโ โ โโโ AppLayout.tsx # Main layout (responsive: max-w-lg mobile, max-w-6xl tablet/landscape)\nโ โ โโโ BottomNav.tsx # Bottom nav (5 tabs, center cashier CTA)\nโ โโโ Onboarding.tsx # First-run tutorial & store setup\nโ โโโ LoginScreen.tsx # Multi-user login (username + PIN)\nโ โโโ LockedPage.tsx # Permission-gated route fallback\nโ โโโ NavLink.tsx # Permission-aware nav link\nโ โโโ BackupReminder.tsx # Backup reminder & export utility\nโ โโโ Receipt.tsx # Receipt component (view, download, share, Bluetooth print)\nโ โโโ BarcodeScanner.tsx # Barcode/QR scanner with PWA-aware permission handling\nโ โโโ ThemeColorPicker.tsx # Accent color picker (8 presets)\nโ โโโ LanguageSwitcher.tsx # Language picker (ID, EN, MS)\nโ โโโ ui/ # shadcn/ui components (40+)\nโโโ i18n/\nโ โโโ index.ts # i18next initialization\nโ โโโ locales/\nโ โโโ id/ # Bahasa Indonesia\nโ โโโ en/ # English\nโ โโโ ms/ # Bahasa Malaysia\nโโโ pages/\nโ โโโ Dashboard.tsx # Home: stats, quick actions, low stock alerts\nโ โโโ Cashier.tsx # POS / cashier (barcode scan input, camera scanner, side-by-side cart on landscape)\nโ โโโ Products.tsx # Product CRUD (with description, SKU, units, photos)\nโ โโโ Reports.tsx # Sales reports & charts\nโ โโโ Settings.tsx # Settings (store, payments, categories, units, theme, backup, install PWA)\nโ โโโ Users.tsx # Multi-user management (owner only)\nโ โโโ Supplier.tsx # Supplier CRUD\nโ โโโ StockIn.tsx # Stock in + COGS calculation\nโ โโโ StockOut.tsx # Stock out\nโ โโโ StockReport.tsx # Stock movement reports\nโ โโโ TransactionHistory.tsx # Transaction history with open bill filter tabs\nโ โโโ NotFound.tsx # 404 page\nโโโ hooks/\n โโโ use-auth.tsx # Multi-user auth context (current user, permissions, login/logout)\n โโโ use-pwa-install.ts # PWA install prompt + standalone detection (incl. iOS)\n โโโ use-theme-color.ts # Accent color persistence\n โโโ use-mobile.tsx # Mobile breakpoint detection\n โโโ use-toast.ts # Toast helper\nAll data is stored locally in the browser using IndexedDB (via Dexie.js). No data is ever sent to any server.
\n| Table | \nDescription | \n
|---|---|
users | \nMulti-user accounts (owner/staff role, hashed PIN, granular permissions) | \n
categories | \nProduct categories (name, color, icon) | \n
products | \nMaster products (name, SKU, sell price, COGS, stock, unit, description) | \n
units | \nMaster units of measurement | \n
suppliers | \nSupplier data | \n
stockIns | \nStock-in records | \n
stockOuts | \nStock-out records | \n
hppHistory | \nCOGS change audit trail | \n
paymentMethods | \nPayment methods (Cash, Bank Transfer, QRIS, etc.) | \n
transactions | \nSales transactions (status: open/completed, customer name, table number, remarks) | \n
transactionItems | \nIndividual items within each transaction (per-item notes & discount) | \n
storeSettings | \nStore settings & app state (incl. multi-user toggle) | \n
When stock is received, COGS is automatically recalculated:
\nNew COGS = ((Old Stock ร Old COGS) + (New Qty ร Buy Price)) / (Old Stock + New Qty)\nGot suggestions, feature ideas, or found a bug? Submit and vote on our board:
\n\nJoin the Telegram group to discuss the app, ask questions, and share tips with other users:
\n๐ t.me/kasirgratisan
\nFreeKasir is proudly supported by:
\n\n![\"Sumopod\"](\"https://raw.githubusercontent.com/jipraks/kasirgratisan/main/public/sponsors/sumopod.png\")
\nWant to sponsor FreeKasir and have your logo featured here? Reach out at sponsorship@freekasir.com.
\nFreeKasir is built and maintained for free. If you find it useful, you can buy the developer a coffee to support continued development:
\n๐ traktir.jipraks.com
\nContributions are welcome! Here's how:
\ngit checkout -b feature/new-feature)git commit -m 'Add new feature')git push origin feature/new-feature)src/i18n/locales/{id,en,ms}/ JSON files inside the appropriate namespace (common, settings, products, reports, dashboard, onboarding)useTranslation('namespace') hook and t('key') in componentsi18n.language and NUMBER_LOCALES / CURRENCY_SYMBOL mapsdate-fns with locale from LOCALES mapshadcn/ui components from src/components/ui/toLocaleString('id-ID')useLiveQuery() from dexie-react-hooks for reactive data bindingcan() helper from useAuth() when multi-user is enabledBuilt with โค๏ธ for Indonesian small businesses.
\n\n
![\"python3.x\"](\"https://img.shields.io/badge/3.12.1-blue.svg?&logo=python&label=Python\"){kind=icon}
\n![\"Version](\"https://img.shields.io/pypi/v/pddiktipy?logo=Python&logoColor=white&label=PyPI&color=c125ff\")
![\"Downloads\"](\"https://img.shields.io/pepy/dt/pddiktipy?logo=PyPI&logoColor=white&label=Downloads&color=c125ff\")
![\"Author\"](\"https://img.shields.io/badge/Author-Ilham%20Riski-blue.svg?style=flat\"){kind=icon}
\n![\"License\"](\"https://img.shields.io/badge/License-MIT-green.svg\"){kind=icon}
\n\nLibrary Python untuk mengakses data PDDIKTI Kemdikbud dengan mudah, aman, dan terdokumentasi lengkap
\n
Wrapper API Python yang powerful dan user-friendly untuk mengambil data dari PDDIKTI Kemdikbud. Library ini menyediakan interface yang mudah digunakan untuk mengakses data mahasiswa, dosen, perguruan tinggi, dan program studi di Indonesia dengan dukungan type hints, error handling yang komprehensif, dan dokumentasi lengkap.
\npip install pddiktipy\nRequirements:
\nfrom pddiktipy import api\nfrom pprint import pprint\n\n# Menggunakan context manager (recommended)\nwith api() as client:\n # Cari semua data dengan keyword\n hasil = client.search_all('Unika Soegijapranata')\n pprint(hasil)\n \n # Cari mahasiswa spesifik\n mahasiswa = client.search_mahasiswa('Ilham Riski Wibowo')\n pprint(mahasiswa)\nLibrary ini menyediakan error handling yang komprehensif:
\nfrom pddiktipy import api\nfrom pddiktipy.exceptions import (\n ValidationError, \n APIConnectionError, \n APITimeoutError,\n PDDIKTIError\n)\n\ntry:\n with api() as client:\n # Ini akan raise ValidationError karena keyword kosong\n result = client.search_mahasiswa(\"\")\n \nexcept ValidationError as e:\n print(f\"Error validasi: {e}\")\nexcept APIConnectionError as e:\n print(f\"Error koneksi: {e}\")\nexcept APITimeoutError as e:\n print(f\"Request timeout: {e}\")\nexcept PDDIKTIError as e:\n print(f\"Error PDDIKTI API: {e}\")\nโก๏ธ API Documentation - 63 method API dengan dokumentasi komprehensif, contoh penggunaan, dan struktur data response
\nSemua response API menggunakan TypedDict untuk type safety dan konsistensi. Struktur data disesuaikan dengan konteks pendidikan Indonesia dan standar PDDIKTI.
\n๐ Changelog - Riwayat versi dan roadmap pengembangan
\n๐งช Testing Guide - Panduan testing dan quality assurance
\n๐ค Contributing Guide - Panduan berkontribusi pada proyek
\n๐ MIT License - Distributed under the MIT License
\nโญ Jika library ini membantu proyek Anda, jangan lupa untuk memberikan star di GitHub!
\n" }, + { + "fullName": "jipraks/kasirgratisan", + "name": "kasirgratisan", + "owner": "jipraks", + "ownerAvatarUrl": "https://avatars.githubusercontent.com/u/10278519?v=4", + "description": "Free, Open Source, Offline Point of Sales Apps", + "metaDescription": "Free, Open Source, Offline Point of Sales Apps", + "url": "https://github.com/jipraks/kasirgratisan", + "homepage": "", + "language": "TypeScript", + "stars": 112, + "forks": 56, + "topics": [], + "updatedAt": "2026-07-20T04:58:20Z", + "pushedAt": "2026-07-13T00:57:27Z", + "latestRelease": null, + "archived": false, + "licenseSpdx": "MIT", + "createdAt": "2026-02-12T11:02:43Z", + "openIssues": 0, + "openPullRequests": 1, + "subscribers": 1, + "communityHealth": 42, + "readmeHtml": "A free, offline-first, open source Point of Sale (POS) Progressive Web App built for Indonesian Micro, Small, and Medium Enterprises (UMKM). All data is stored locally on the user's device โ no server, no registration, no cost.
\n| Layer | \nTechnology | \n
|---|---|
| Framework | \nReact 18 + TypeScript | \n
| Build Tool | \nVite | \n
| Styling | \nTailwind CSS + shadcn/ui | \n
| Theming | \nnext-themes (dark mode) | \n
| Database | \nIndexedDB via Dexie.js | \n
| Charts | \nRecharts | \n
| Routing | \nReact Router DOM v6 | \n
| Forms & Validation | \nReact Hook Form + Zod | \n
| State | \n@tanstack/react-query | \n
| Icons | \nLucide React | \n
| i18n | \ni18next + react-i18next | \n
| Date | \ndate-fns (id, en-US, ms locales) | \n
| PWA | \nvite-plugin-pwa (Workbox) | \n
| Barcode | \nhtml5-qrcode (camera scanner + manual input) | \n
| Receipt | \nhtml2canvas (to PNG), Web Bluetooth Print (PWA), Bluetooth Classic (Android APK via Capacitor) | \n
| Font | \nPlus Jakarta Sans | \n
| Native Wrapper | \nCapacitor 8 (Android) | \n
# Clone the repository\ngit clone https://github.com/user/kasirgratisan.git\ncd kasirgratisan\n\n# Install dependencies\nnpm install\n\n# Start the development server\nnpm run dev\nThe app will be running at http://localhost:8080.
npm run build\nnpm run preview\nThis project can also run as a native Android app using Capacitor while keeping the PWA/web version working from the same codebase.
\nRequirements:
\nmacOS / Linux:
\nexport JAVA_HOME=\"/Applications/Android Studio.app/Contents/jbr/Contents/Home\"\nWindows (PowerShell):
\n$env:JAVA_HOME = \"C:\\Program Files\\Android\\Android Studio\\jbr\"\n\n\nAdjust the path if your Android Studio is installed elsewhere.
\n
npm run build\nnpx cap sync android\ncd android\n./gradlew assembleDebug\nOutput: android/app/build/outputs/apk/debug/app-debug.apk
npm run build\nnpx cap sync android\ncd android\n./gradlew bundleRelease\nOutput: android/app/build/outputs/bundle/release/app-release.aab
\n\nThe release AAB must be signed before uploading to Google Play. See Android signing docs.
\n
npm run cap:sync # build web bundle and sync Capacitor\nnpm run cap:android # build, sync, then open Android Studio\nnpm run cap:run # build, sync, then run on connected Android device/emulator\nsrc/\nโโโ App.tsx # Root component & routing\nโโโ main.tsx # Entry point\nโโโ index.css # Design tokens (HSL CSS variables)\nโโโ lib/\nโ โโโ db.ts # Dexie database schema, interfaces, seed data\nโ โโโ auth.ts # Multi-user auth helpers (PIN hashing, sessions, validation)\nโ โโโ utils.ts # Utility functions (cn, etc.)\nโ โโโ image-utils.ts # Image compression utility\nโ โโโ version-check.ts # Version check webhook\nโโโ components/\nโ โโโ layout/\nโ โ โโโ AppLayout.tsx # Main layout (responsive: max-w-lg mobile, max-w-6xl tablet/landscape)\nโ โ โโโ BottomNav.tsx # Bottom nav (5 tabs, center cashier CTA)\nโ โโโ Onboarding.tsx # First-run tutorial & store setup\nโ โโโ LoginScreen.tsx # Multi-user login (username + PIN)\nโ โโโ LockedPage.tsx # Permission-gated route fallback\nโ โโโ NavLink.tsx # Permission-aware nav link\nโ โโโ BackupReminder.tsx # Backup reminder & export utility\nโ โโโ Receipt.tsx # Receipt component (view, download, share, Bluetooth print)\nโ โโโ BarcodeScanner.tsx # Barcode/QR scanner with PWA-aware permission handling\nโ โโโ ThemeColorPicker.tsx # Accent color picker (8 presets)\nโ โโโ LanguageSwitcher.tsx # Language picker (ID, EN, MS)\nโ โโโ ui/ # shadcn/ui components (40+)\nโโโ i18n/\nโ โโโ index.ts # i18next initialization\nโ โโโ locales/\nโ โโโ id/ # Bahasa Indonesia\nโ โโโ en/ # English\nโ โโโ ms/ # Bahasa Malaysia\nโโโ pages/\nโ โโโ Dashboard.tsx # Home: stats, quick actions, low stock alerts\nโ โโโ Cashier.tsx # POS / cashier (barcode scan input, camera scanner, side-by-side cart on landscape)\nโ โโโ Products.tsx # Product CRUD (with description, SKU, units, photos)\nโ โโโ Reports.tsx # Sales reports & charts\nโ โโโ Settings.tsx # Settings (store, payments, categories, units, theme, backup, install PWA)\nโ โโโ Users.tsx # Multi-user management (owner only)\nโ โโโ Supplier.tsx # Supplier CRUD\nโ โโโ StockIn.tsx # Stock in + COGS calculation\nโ โโโ StockOut.tsx # Stock out\nโ โโโ StockReport.tsx # Stock movement reports\nโ โโโ TransactionHistory.tsx # Transaction history with open bill filter tabs\nโ โโโ NotFound.tsx # 404 page\nโโโ hooks/\n โโโ use-auth.tsx # Multi-user auth context (current user, permissions, login/logout)\n โโโ use-pwa-install.ts # PWA install prompt + standalone detection (incl. iOS)\n โโโ use-theme-color.ts # Accent color persistence\n โโโ use-mobile.tsx # Mobile breakpoint detection\n โโโ use-toast.ts # Toast helper\nAll data is stored locally in the browser using IndexedDB (via Dexie.js). No data is ever sent to any server.
\n| Table | \nDescription | \n
|---|---|
users | \nMulti-user accounts (owner/staff role, hashed PIN, granular permissions) | \n
categories | \nProduct categories (name, color, icon) | \n
products | \nMaster products (name, SKU, sell price, COGS, stock, unit, description) | \n
units | \nMaster units of measurement | \n
suppliers | \nSupplier data | \n
stockIns | \nStock-in records | \n
stockOuts | \nStock-out records | \n
hppHistory | \nCOGS change audit trail | \n
paymentMethods | \nPayment methods (Cash, Bank Transfer, QRIS, etc.) | \n
transactions | \nSales transactions (status: open/completed, customer name, table number, remarks) | \n
transactionItems | \nIndividual items within each transaction (per-item notes & discount) | \n
storeSettings | \nStore settings & app state (incl. multi-user toggle) | \n
When stock is received, COGS is automatically recalculated:
\nNew COGS = ((Old Stock ร Old COGS) + (New Qty ร Buy Price)) / (Old Stock + New Qty)\nGot suggestions, feature ideas, or found a bug? Submit and vote on our board:
\n\nJoin the Telegram group to discuss the app, ask questions, and share tips with other users:
\n๐ t.me/kasirgratisan
\nFreeKasir is proudly supported by:
\n\n\nWant to sponsor FreeKasir and have your logo featured here? Reach out at sponsorship@freekasir.com.
\nFreeKasir is built and maintained for free. If you find it useful, you can buy the developer a coffee to support continued development:
\n๐ traktir.jipraks.com
\nContributions are welcome! Here's how:
\ngit checkout -b feature/new-feature)git commit -m 'Add new feature')git push origin feature/new-feature)src/i18n/locales/{id,en,ms}/ JSON files inside the appropriate namespace (common, settings, products, reports, dashboard, onboarding)useTranslation('namespace') hook and t('key') in componentsi18n.language and NUMBER_LOCALES / CURRENCY_SYMBOL mapsdate-fns with locale from LOCALES mapshadcn/ui components from src/components/ui/toLocaleString('id-ID')useLiveQuery() from dexie-react-hooks for reactive data bindingcan() helper from useAuth() when multi-user is enabledBuilt with โค๏ธ for Indonesian small businesses.
\n\n Noise-canceling context and long-term memory for your AI agent โ lossy, but always reversible, and it never fabricates a result. Stop paying Claude to read 10,000 lines of terminal noise.\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Always reversible ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Every other compressor asks you to trust that what it cut didn't matter. OMNI doesn't ask โ it guarantees, and each guarantee is backed by code you can read:
\n| Guarantee | \nHow | \nProof | \n
|---|---|---|
| Get the original back, byte-for-byte | \neverything cut is archived in a local SQLite RewindStore (SHA-256 โ content); the agent gets a hash and calls omni_retrieve | \nHow it works | \n
| Never fabricates a result | \na distiller that parsed no signal returns the raw output, never a green no errors / passed string | \n#143 | \n
| Failures are never masked | \na command that exits non-zero passes through verbatim | \n#120 | \n
| Structured data is never touched | \nJSON / YAML / NDJSON / CSV pass through byte-for-byte | \npipeline::format | \n
| Numbers are measured, not aspirational | \n1,810 real traces replayed on the release binary โ and 63.6% of calls net zero, which we publish too | \nBenchmarks | \n
That is the one thing a bigger compression number can't buy: you can always recover the original, and it will never lie to your agent.
\nProblem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. The agent surfaces it through the omni_recall MCP tool before it repeats the mistake.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust, though the end-to-end cost is not zero.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
Yes, measurably, and the cost grows with your history. The distillation pipeline itself runs in single-digit milliseconds, but every hooked command also writes to the local RewindStore: a 496-byte git status takes ~82 ms against a fresh database and ~308 ms against a 97 MB one, and a 16.5 KB cargo test takes ~276 ms. Budget for it. OMNI_PASSTHROUGH=1 skips the pipeline entirely when you need the raw output back.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
\n
![\"Validate\"](\"https://github.com/wauputr4/agent-proxmox/actions/workflows/validate.yml/badge.svg\"){kind=icon}
Reusable AI agent skill for safe Proxmox VE host operations, LXC service handling, incident response, and maintenance logging.
\nThis repository packages proxmox-host-operator, a portable SKILL.md workflow for agents that operate small to medium Proxmox hosts with LXC containers, Docker Compose services, Cloudflare Tunnel, Tailscale, backups, and LVM thin storage.
The skill is based on roughly two months of hands-on best practices from managing a homelab Proxmox environment that runs production workloads: public websites, staging apps, shared databases, automation, AI tools, tunnels, backups, security hardening, and incident recovery. The source lessons were generalized so the skill is not tied to any one organization, domain, IP address, or service name.
\nAfter publishing this repository on GitHub, install it with the skills CLI:
\nnpx skills add wauputr4/agent-proxmox\nskills.sh lists GitHub-hosted skills automatically after users install them with the CLI.
\nskills/\n proxmox-host-operator/\n SKILL.md\n agents/openai.yaml\n references/\n incident-patterns.md\n activity-logging.md\n migration-playbook.md\n ops-logbook.md\n proxmox-runbooks.md\n security-hardening.md\n scripts/\n collect-proxmox-triage.sh\n new-log-entry.py\nskills.sh.json\nREADME.md\nCONTRIBUTING.md\nCODE_OF_CONDUCT.md\nSECURITY.md\nLICENSE\nUse this skill if your AI agent helps with:
\n.env files.Contributions are welcome. Good additions include generalized incident patterns, safer diagnostics, better rollback checklists, and examples from other Proxmox environments.
\nPlease avoid organization-specific hostnames, public IPs, credentials, or private service names in contributions. See CONTRIBUTING.md for the contribution workflow.
\nMIT. See LICENSE.
\n" diff --git a/src/data/revival.json b/src/data/revival.json index 6be11d6..23df244 100644 --- a/src/data/revival.json +++ b/src/data/revival.json @@ -200,7 +200,7 @@ "description": "Unofficial Python 3 API wrapper to retrieve data from PDDIKTI Kemdikbudristek.", "url": "https://github.com/IlhamriSKY/PDDIKTI-kemdikbud-API", "homepage": "https://pddikti.kemdiktisaintek.go.id/", - "stars": 112, + "stars": 113, "forks": 25, "language": "Python", "topics": [ @@ -210,7 +210,7 @@ ], "license": "NOASSERTION", "createdAt": "2021-04-21T07:59:57Z", - "updatedAt": "2026-07-22T15:35:48Z", + "updatedAt": "2026-07-22T17:59:08Z", "pushedAt": "2025-07-30T13:28:14Z", "archived": false, "disabled": false From d9dadea689bea86b3513e9c84d2fcf68e0495717 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Wed, 22 Jul 2026 20:35:26 +0000 Subject: [PATCH 21/25] Sync content data --- src/data/projects.json | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index 22247cf..f02f4e0 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -2094,8 +2094,8 @@ "stars": 6, "forks": 0, "topics": [], - "updatedAt": "2026-07-21T20:20:21Z", - "pushedAt": "2026-07-21T20:19:56Z", + "updatedAt": "2026-07-22T20:12:49Z", + "pushedAt": "2026-07-22T20:12:36Z", "latestRelease": null, "archived": false, "licenseSpdx": "MIT", From 6b7565ec909ccf9a9574473a871340b42c6ae6ee Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Wed, 22 Jul 2026 22:20:12 +0000 Subject: [PATCH 22/25] Sync content data --- src/data/projects.json | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index f02f4e0..e98daad 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -156,10 +156,10 @@ "url": "https://github.com/jipraks/yt-short-clipper", "homepage": "", "language": "Python", - "stars": 899, + "stars": 900, "forks": 279, "topics": [], - "updatedAt": "2026-07-22T11:22:16Z", + "updatedAt": "2026-07-22T22:15:38Z", "pushedAt": "2026-07-18T02:05:39Z", "latestRelease": { "name": "YT Short Clipper v2.0.5-beta", @@ -473,7 +473,7 @@ "url": "https://github.com/codecoradev/uteke", "homepage": "https://codecora.dev", "language": "Rust", - "stars": 125, + "stars": 126, "forks": 15, "topics": [ "ai", @@ -490,7 +490,7 @@ "sqlite", "vector-database" ], - "updatedAt": "2026-07-22T09:23:15Z", + "updatedAt": "2026-07-22T21:07:11Z", "pushedAt": "2026-07-21T23:43:24Z", "latestRelease": { "name": "Release v0.10.0", From c72610ba70e2e7d50fcada7a1ce850e8a96f85d9 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Thu, 23 Jul 2026 00:16:45 +0000 Subject: [PATCH 23/25] Sync content data --- src/data/projects.json | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index e98daad..d03b7ce 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -473,7 +473,7 @@ "url": "https://github.com/codecoradev/uteke", "homepage": "https://codecora.dev", "language": "Rust", - "stars": 126, + "stars": 127, "forks": 15, "topics": [ "ai", @@ -490,7 +490,7 @@ "sqlite", "vector-database" ], - "updatedAt": "2026-07-22T21:07:11Z", + "updatedAt": "2026-07-22T23:20:14Z", "pushedAt": "2026-07-21T23:43:24Z", "latestRelease": { "name": "Release v0.10.0", From 5e60680b648aae7215bbee3b0476d0f6aa5e61e7 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Thu, 23 Jul 2026 03:51:33 +0000 Subject: [PATCH 24/25] Sync content data --- src/data/projects.json | 43 +++++++++++++++++++++++------------------- 1 file changed, 24 insertions(+), 19 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index d03b7ce..423e024 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -130,7 +130,7 @@ "sistem-informasi-desa" ], "updatedAt": "2026-07-22T13:40:31Z", - "pushedAt": "2026-07-18T13:09:40Z", + "pushedAt": "2026-07-23T03:19:25Z", "latestRelease": { "name": "Rilis v2607.0.0", "tagName": "v2607.0.0", @@ -140,7 +140,7 @@ "archived": false, "licenseSpdx": "", "createdAt": "2016-05-21T10:55:38Z", - "openIssues": 355, + "openIssues": 350, "openPullRequests": 7, "subscribers": 110, "communityHealth": 50, @@ -424,7 +424,7 @@ "homepage": "", "language": "Python", "stars": 154, - "forks": 36, + "forks": 37, "topics": [], "updatedAt": "2026-07-22T06:40:34Z", "pushedAt": "2026-06-26T05:38:51Z", @@ -490,8 +490,8 @@ "sqlite", "vector-database" ], - "updatedAt": "2026-07-22T23:20:14Z", - "pushedAt": "2026-07-21T23:43:24Z", + "updatedAt": "2026-07-23T03:40:07Z", + "pushedAt": "2026-07-23T03:39:52Z", "latestRelease": { "name": "Release v0.10.0", "tagName": "v0.10.0", @@ -788,13 +788,13 @@ "tauri", "terminal" ], - "updatedAt": "2026-07-21T12:58:32Z", - "pushedAt": "2026-07-21T12:58:04Z", + "updatedAt": "2026-07-23T01:16:57Z", + "pushedAt": "2026-07-23T01:16:54Z", "latestRelease": { - "name": "TEDI v0.3.93", - "tagName": "v0.3.93", - "url": "https://github.com/IlhamriSKY/TEDI/releases/tag/v0.3.93", - "publishedAt": "2026-07-21T13:18:57Z" + "name": "TEDI v0.3.94", + "tagName": "v0.3.94", + "url": "https://github.com/IlhamriSKY/TEDI/releases/tag/v0.3.94", + "publishedAt": "2026-07-23T01:34:36Z" }, "archived": false, "licenseSpdx": "Apache-2.0", @@ -1298,8 +1298,8 @@ "stars": 24, "forks": 2, "topics": [], - "updatedAt": "2026-07-19T05:48:50Z", - "pushedAt": "2026-07-19T05:48:46Z", + "updatedAt": "2026-07-23T03:47:44Z", + "pushedAt": "2026-07-23T03:47:25Z", "latestRelease": null, "archived": false, "licenseSpdx": "MIT", @@ -1850,7 +1850,7 @@ "semantic-search" ], "updatedAt": "2026-07-03T02:36:07Z", - "pushedAt": "2026-07-14T09:11:34Z", + "pushedAt": "2026-07-23T02:39:54Z", "latestRelease": { "name": "OpenEmpiric v1.0.5", "tagName": "v1.0.5", @@ -2309,17 +2309,22 @@ "skills-sh", "tailscale" ], - "updatedAt": "2026-06-25T07:10:50Z", - "pushedAt": "2026-07-22T18:04:09Z", - "latestRelease": null, + "updatedAt": "2026-07-23T02:37:26Z", + "pushedAt": "2026-07-23T02:37:49Z", + "latestRelease": { + "name": "v0.1.0", + "tagName": "v0.1.0", + "url": "https://github.com/wauputr4/agent-proxmox/releases/tag/v0.1.0", + "publishedAt": "2026-07-23T02:37:49Z" + }, "archived": false, "licenseSpdx": "MIT", "createdAt": "2026-06-06T16:32:11Z", "openIssues": 0, - "openPullRequests": 2, + "openPullRequests": 1, "subscribers": 0, "communityHealth": 100, - "readmeHtml": "\n
Reusable AI agent skill for safe Proxmox VE host operations, LXC service handling, incident response, and maintenance logging.
\nThis repository packages proxmox-host-operator, a portable SKILL.md workflow for agents that operate small to medium Proxmox hosts with LXC containers, Docker Compose services, Cloudflare Tunnel, Tailscale, backups, and LVM thin storage.
The skill is based on roughly two months of hands-on best practices from managing a homelab Proxmox environment that runs production workloads: public websites, staging apps, shared databases, automation, AI tools, tunnels, backups, security hardening, and incident recovery. The source lessons were generalized so the skill is not tied to any one organization, domain, IP address, or service name.
\nAfter publishing this repository on GitHub, install it with the skills CLI:
\nnpx skills add wauputr4/agent-proxmox\nskills.sh lists GitHub-hosted skills automatically after users install them with the CLI.
\nskills/\n proxmox-host-operator/\n SKILL.md\n agents/openai.yaml\n references/\n incident-patterns.md\n activity-logging.md\n migration-playbook.md\n ops-logbook.md\n proxmox-runbooks.md\n security-hardening.md\n scripts/\n collect-proxmox-triage.sh\n new-log-entry.py\nskills.sh.json\nREADME.md\nCONTRIBUTING.md\nCODE_OF_CONDUCT.md\nSECURITY.md\nLICENSE\nUse this skill if your AI agent helps with:
\n.env files.Contributions are welcome. Good additions include generalized incident patterns, safer diagnostics, better rollback checklists, and examples from other Proxmox environments.
\nPlease avoid organization-specific hostnames, public IPs, credentials, or private service names in contributions. See CONTRIBUTING.md for the contribution workflow.
\nMIT. See LICENSE.
\n" + "readmeHtml": "\n
Reusable AI agent skill for safe Proxmox VE host operations, LXC service handling, incident response, and maintenance logging.
\nThis repository packages proxmox-host-operator, a portable SKILL.md workflow for agents that operate small to medium Proxmox hosts with LXC containers, Docker Compose services, Cloudflare Tunnel, Tailscale, backups, and LVM thin storage.
The skill is based on roughly two months of hands-on best practices from managing a homelab Proxmox environment that runs production workloads: public websites, staging apps, shared databases, automation, AI tools, tunnels, backups, security hardening, and incident recovery. The source lessons were generalized so the skill is not tied to any one organization, domain, IP address, or service name.
\nAfter publishing this repository on GitHub, install it with the skills CLI:
\nnpx skills add wauputr4/agent-proxmox\nskills.sh lists GitHub-hosted skills automatically after users install them with the CLI.
\nskills/\n proxmox-host-operator/\n SKILL.md\n agents/openai.yaml\n references/\n incident-patterns.md\n activity-logging.md\n migration-playbook.md\n ops-logbook.md\n proxmox-runbooks.md\n security-hardening.md\n scripts/\n collect-proxmox-triage.sh\n new-log-entry.py\nskills.sh.json\nREADME.md\nCONTRIBUTING.md\nCODE_OF_CONDUCT.md\nSECURITY.md\nLICENSE\nUse this skill if your AI agent helps with:
\n.env files.Contributions are welcome. Good additions include generalized incident patterns, safer diagnostics, better rollback checklists, and examples from other Proxmox environments.
\nPlease avoid organization-specific hostnames, public IPs, credentials, or private service names in contributions. See CONTRIBUTING.md for the contribution workflow.
\nMIT. See LICENSE.
\n" }, { "fullName": "faisalaffan/ragi-instant", From 708c13c5f0305cc2d5fc8f34ed72eb96f63a3364 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" <41898282+github-actions[bot]@users.noreply.github.com> Date: Thu, 23 Jul 2026 05:33:19 +0000 Subject: [PATCH 25/25] Sync content data --- src/data/projects.json | 98 +++++++++++++++++++++--------------------- 1 file changed, 49 insertions(+), 49 deletions(-) diff --git a/src/data/projects.json b/src/data/projects.json index 423e024..ab8f07d 100644 --- a/src/data/projects.json +++ b/src/data/projects.json @@ -130,7 +130,7 @@ "sistem-informasi-desa" ], "updatedAt": "2026-07-22T13:40:31Z", - "pushedAt": "2026-07-23T03:19:25Z", + "pushedAt": "2026-07-23T05:30:50Z", "latestRelease": { "name": "Rilis v2607.0.0", "tagName": "v2607.0.0", @@ -186,10 +186,10 @@ "url": "https://github.com/laravolt/indonesia", "homepage": "", "language": "PHP", - "stars": 670, + "stars": 671, "forks": 210, "topics": [], - "updatedAt": "2026-07-19T09:32:55Z", + "updatedAt": "2026-07-23T04:25:33Z", "pushedAt": "2026-03-03T06:32:45Z", "latestRelease": { "name": "v0.41", @@ -313,8 +313,8 @@ "token-reduction", "token-savings" ], - "updatedAt": "2026-07-22T17:36:51Z", - "pushedAt": "2026-07-22T17:58:39Z", + "updatedAt": "2026-07-23T05:00:52Z", + "pushedAt": "2026-07-23T05:00:38Z", "latestRelease": { "name": "v0.6.3", "tagName": "v0.6.3", @@ -324,8 +324,8 @@ "archived": false, "licenseSpdx": "MIT", "createdAt": "2026-03-15T03:38:04Z", - "openIssues": 21, - "openPullRequests": 2, + "openIssues": 20, + "openPullRequests": 1, "subscribers": 4, "communityHealth": 71, "readmeHtml": "\n Noise-canceling context and long-term memory for your AI agent โ lossy, but always reversible, and it never fabricates a result. Stop paying Claude to read 10,000 lines of terminal noise.\n
๐บ๐ธ English | ๐ฏ๐ต ๆฅๆฌ่ช | ๐จ๐ณ ็ฎไฝไธญๆ | ๐ธ๐ฆ ุงูุนุฑุจูุฉ | ๐ฎ๐ฉ Bahasa Indonesia | ๐ป๐ณ Tiแบฟng Viแปt | ๐ฐ๐ท ํ๊ตญ์ด
\n\n\n \n \n
\n\n58.9% fewer tokens on a real command mix ยท Cross-Session Memory ยท Format-safe ยท Always reversible ยท Fails open, never fabricates ยท Numbers you can reproduce
\n
Every AI coding assistant has two massive problems.
\n1. They read everything.
Build logs.
Docker logs.
CI logs.
Progress bars.
ANSI colors.
Thousands of tokens... to find one line. Claude isn't expensive. Your terminal is.
2. They forget everything.
Every time you restart Cursor, or switch from Claude Code to Windsurf, your agent gets amnesia. You have to re-explain the project goal. You have to remind them of the same framework gotchas over and over again.
OMNI fixes both.
\nProblem 1: Your terminal drowns out the signal
\nReal numbers, measured on tests/fixtures/ and replayed traces โ not aspirations:
| Command | \nWithout OMNI | \nWith OMNI | \nSaved | \n
|---|---|---|---|
cargo test (490 passed, 10 failed) | \n16.5 KB of per-test output | \nthe runner's own pass/fail summary | \n93% | \n
kubectl get pods (35 pods, 5 crashing) | \nthe full table | \n35 pods | 30 running, 5 error + the 5 failing pods named | \nโ | \n
git diff (multi-file) | \nlockfiles, whitespace, generated churn | \nthe code that actually changed | \n45% | \n
docker build (heavy cache noise) | \n9.2 KB of layer hashes and progress bars | \nthe build result, cache hits folded | \n37% | \n
\n\nThe honest caveat: OMNI compresses noisy successful output. A command that fails is passed through verbatim โ a hidden error is worse than an uncompressed one โ and structured output (JSON/YAML/CSV) is never touched. It earns its keep on repetitive tool chatter and gets out of the way everywhere else.
\n
Every other compressor asks you to trust that what it cut didn't matter. OMNI doesn't ask โ it guarantees, and each guarantee is backed by code you can read:
\n| Guarantee | \nHow | \nProof | \n
|---|---|---|
| Get the original back, byte-for-byte | \neverything cut is archived in a local SQLite RewindStore (SHA-256 โ content); the agent gets a hash and calls omni_retrieve | \nHow it works | \n
| Never fabricates a result | \na distiller that parsed no signal returns the raw output, never a green no errors / passed string | \n#143 | \n
| Failures are never masked | \na command that exits non-zero passes through verbatim | \n#120 | \n
| Structured data is never touched | \nJSON / YAML / NDJSON / CSV pass through byte-for-byte | \npipeline::format | \n
| Numbers are measured, not aspirational | \n1,810 real traces replayed on the release binary โ and 63.6% of calls net zero, which we publish too | \nBenchmarks | \n
That is the one thing a bigger compression number can't buy: you can always recover the original, and it will never lie to your agent.
\nProblem 2: Your agent forgets everything overnight
\nWithout OMNI: \"Please re-explain the project structure, the auth module is broken, and we use Postgres not MySQL.\"
With OMNI: The agent already knows. It picks up where you left off.
Without OMNI: Agent hits the same framework gotcha it already solved yesterday because it has no memory.
With OMNI: The fix is already stored. The agent surfaces it through the omni_recall MCP tool before it repeats the mistake.
Without OMNI: New IDE, new agent, zero context. You're starting from scratch.
With OMNI: Session summary is injected automatically. New agent is immediately up to speed.
The code you don't send to the AI is just as important as the code you do.
\nWhen you feed an AI megabytes of terminal noise, it suffers from context bloatโhallucinating fixes for the wrong warnings and burning your API budget on irrelevant output.
\nWhen you restart an agent and it has no memory, you lose hours re-establishing context that should have been preserved automatically.
\nOMNI solves both, invisibly:
\nThe honest headline, measured on the release binary against 1,810 real command\nexecutions replayed from one developer's actual usage:
\nThat second bullet is the number most tools in this category do not print. A tool\nthat claims to save 90% of every command is telling you it summarises output you\nneeded.
\n\nWhere the saving actually comes from, over the same 1,810 executions:
\n| Command | \nCalls | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|---|
cargo | \n29 | \n424 KB | \n13 KB | \n96.8% | \n
git | \n256 | \n5.9 MB | \n509 KB | \n91.3% | \n
ls | \n52 | \n71 KB | \n29 KB | \n59.5% | \n
kubectl | \n212 | \n4.4 MB | \n2.3 MB | \n48.0% | \n
find | \n39 | \n83 KB | \n53 KB | \n36.2% | \n
grep | \n184 | \n534 KB | \n385 KB | \n27.8% | \n
cat | \n85 | \n515 KB | \n468 KB | \n9.1% | \n
git and cargo carry the result; cat and grep are close to a no-op. OMNI\nearns its place on noisy, repetitive tooling output and gets out of the way\neverywhere else.
Single fixtures from tests/fixtures/, if you want to reproduce one by hand:
| Command / Context | \nInput | \nOutput | \nSaved | \n
|---|---|---|---|
cargo build (large, successful) | \n3,220 B | \n9 B | \n99.7% | \n
cargo test (490 passed, 10 failed) | \n16.5 KB | \n1,100 B | \n93.3% | \n
pytest (failures) | \n730 B | \n136 B | \n81.4% | \n
git status (dirty) | \n496 B | \n113 B | \n77.2% | \n
git diff (multi-file) | \n397 B | \n220 B | \n44.6% | \n
docker build (heavy noise) | \n9.2 KB | \n5.8 KB | \n37.2% | \n
kubectl get pods (mixed) | \n840 B | \n762 B | \n9.3% | \n
Latency is a real cost, not zero. OMNI runs on every hooked command, and the\nprice grows with your history: a 496-byte git status takes ~82 ms against a\nfresh database and ~308 ms against a 97 MB one. A 16.5 KB cargo test takes\n~276 ms. Budget for it.
To see your own actual token savings, just run omni stats after a few days of usage.
Omni is incredibly easy to set up. It natively integrates into your terminal.
\nmacOS / Linux:
\n# 1. Install via Homebrew\nbrew install fajarhide/tap/omni\n\n# 2. Setup Omni (Interactive Menu for Claude, VS Code, OpenCode, Codex, Antigravity)\nomni init\n\n# 3. Verify it's working\nomni doctor\n\n# 4. Or auto-fix any issues\nomni doctor --fix\n\n# 5. Check Current Status\nomni init --status\nUniversal Installer (macOS / Linux / WSL):
\ncurl -fsSL omni.weekndlabs.com/install | bash\nWindows (PowerShell):
\nirm omni.weekndlabs.com/install.ps1 | iex\nOMNI works seamlessly with the agentic tools you already use. It intercepts their terminal executions automatically.
\nOMNI isn't just a terminal filterโit's a cure for AI amnesia.
\nIf you've ever worked with an AI agent for more than an hour, you know the pain of context loss. You restart the agent, and suddenly it forgets what you were working on. It forgets the project goal. It starts making the exact same mistakes it made yesterday because it forgot the repository's undocumented quirks.
\nOMNI's Memory OS runs silently in the background to solve this:
\nomni goal): Set your North Star objective once. OMNI will relentlessly remind the agent of this exact priority on every single prompt, preventing it from drifting off-task.omni remember): Stop fixing the same hallucination. Agents can save project-specific rules, gotchas, and architecture decisions directly into OMNI's local SQLite backend. When they get stuck later, they automatically pull the exact answer back out via semantic search.Your agent gets smarter about your codebase every single day, and you never have to repeat yourself again.
\nOmni operates purely locally using a deterministic Read โ Guard โ Score โ Collapse โ Distill โ Persist pipeline.
flowchart LR\n Command[Raw Tool Output] --> Hook[Omni Hook]\n Hook --> Score[Scorer Engine]\n Score -->|Critical=1.0, Noise=0.1| Distill[Content Distiller]\n Distill --> Clean[Clean Context]\n Command --> SQLite[(RewindStore SQLite)]\nIf the AI really needs the dropped noise, OMNI's local SQLite RewindStore keeps the full uncompressed log safely hashed, allowing the agent to retrieve it anytime.
\n\nBuilt in Rust, though the end-to-end cost is not zero.
\n# Development\ncargo build --release\ncargo test --all\nmake fmt && make clippy\nDoes Omni permanently delete my logs?
No. The raw logs are compressed and stored locally in the SQLite RewindStore. The AI receives a hash and can retrieve the full log if needed.
Will this slow down my terminal?
Yes, measurably, and the cost grows with your history. The distillation pipeline itself runs in single-digit milliseconds, but every hooked command also writes to the local RewindStore: a 496-byte git status takes ~82 ms against a fresh database and ~308 ms against a 97 MB one, and a 16.5 KB cargo test takes ~276 ms. Budget for it. OMNI_PASSTHROUGH=1 skips the pipeline entirely when you need the raw output back.
Can I add my own filters?
Yes. You can teach OMNI to strip noise specific to your internal tools using TOML:
# ~/.omni/signals/custom.toml\n[filters.my_tool]\nmatch_command = \"^internal-tool\\\\b\"\nstrip_lines_matching = [\"^DEBUG\", \"syncing...\"]\nThis is a passion project built for the era of Agentic AI. Whether you're here to save money on tokens, test out free models, or help build the ultimate agentic toolbelt, contributions are always welcome!
\nmake ci and cargo build. Read our CONTRIBUTING.md for details.\n \n \n \n \n \n \n \n
TypeScript/Express REST API starter kit that ships with User Management, Role Management, and Authentication with JWTs, Prisma ORM, and PostgreSQL. The project is organized by feature with explicit service/repository layers so business logic stays separated from transport concerns.
\nsrc/\nโโโ app.ts # Express app bootstrap\nโโโ server.ts # Starts HTTP server\nโโโ config/ # Environment loader & Prisma client wrapper\nโโโ controllers/ # HTTP handlers grouped by module + shared helpers\nโโโ services/ # Business logic (auth/user) & mappers\nโโโ repositories/ # Prisma data access per module\nโโโ middleware/ # Cross-cutting middleware (auth, validation)\nโโโ routes/ # Express routers mounted under /auth and /users\nโโโ validations/ # Zod schemas for request validation\nโโโ views/ # Email templates\nโโโ errors/ # Custom AppError type & error utilities\nโโโ utils/ # Shared utilities (password, string, Zod helpers)\nโโโ types/ # Shared TS types & Express module augmentation\npnpm install\nCreate .env (never commit it) with the required settings:
# Server Configuration\nNODE_ENV=development\nFRONTEND_URL=http://localhost:8080\nPORT=5050\nLOG_LEVEL=info\n\n# Database\nDATABASE_URL=postgresql://USER:PASSWORD@HOST:PORT/DATABASE\n\n# JWT Configuration\nJWT_SECRET=super-secret\n\n# Bcrypt Configuration\nSALT_ROUNDS=10\n\n# Email Configuration\nEMAIL_HOST=smtp.gmail.com\nEMAIL_PORT=587\nEMAIL_USER=your-email@gmail.com\nEMAIL_PASSWORD=your-app-password\nEMAIL_FROM=noreply@yourapp.com\n\n# Token Configuration\nACCESS_TOKEN_EXPIRY=\"1h\"\nENABLE_REFRESH_TOKEN=true\nREFRESH_TOKEN_IN_JSON=true\nREFRESH_TOKEN_IN_COOKIE=true\nREFRESH_TOKEN_EXPIRY=7 * 24 * 60 * 60 * 1000 # 7 days in miliseconds\nCOOKIE_SAME_SITE=\"strict\" # strict, lax, none\nRESET_PASSWORD_TOKEN_EXPIRY=1 * 60 * 60 * 1000 # 1 hour in miliseconds\nEMAIL_VERIFICATION_TOKEN_EXPIRY=24 * 60 * 60 * 10000 # 24 hours in miliseconds\n\n# CORS Configuration\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080,https://yourdomain.com\nCORS_CREDENTIALS=true\n\n\nNote for Gmail: Use an App Password instead of your regular password. Enable 2FA and generate an App Password in Google Account Settings โ Security โ App passwords.
\n
prisma/schema.prisma.pnpm prisma migrate dev (for local) or pnpm prisma db push for quick sync.pnpm prisma generate. Output lands in src/generated/prisma.pnpm prisma db seedpnpm dev\nRuns tsx in watch mode, recompiling on changes. The API listens on PORT from the env file (defaults to 5050).
| Command | \nPurpose | \n
|---|---|
pnpm dev | \nStart the API in watch mode with tsx | \n
pnpm prisma migrate dev | \nCreate/apply migrations and regenerate Prisma client | \n
pnpm prisma generate | \nRegenerate Prisma client manually | \n
pnpm prisma db push | \nQuick sync schema to database without migrations | \n
pnpm prisma db seed | \nSeed the database | \n
pnpm build | \nBuild the API for production (bundles with tsup) | \n
pnpm start | \nStart the production server from dist/ | \n
pnpm test | \nRun all tests in watch mode | \n
pnpm test:unit | \nRun unit tests only | \n
pnpm test:integration | \nRun integration tests only (sequential) | \n
\n\nโ Production build: The repo currently runs via tsx; add a
\ntscbuild +startscript before deploying to production environments like Vercel/Node runtime functions.
The API includes interactive documentation powered by Swagger UI and OpenAPI 3.0 (swagger-jsdoc and swagger-ui-express).
/api-docs (accessible when the server is running)All endpoints respond with { status, message, data? } JSON payloads.\nFor paginated endpoints (like GET /users and GET /roles), the response also includes meta and links objects containing paging data and HATEOAS navigational URLs. They optionally accept query parameters: ?page=1&limit=10&sortBy=createdAt&sortOrder=asc&search=value.
/auth)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
POST | \n/auth/register | \nRegister a new user and send verification email | \nPublic | \nregisterUserSchema | \n
POST | \n/auth/login | \nVerify credentials and return JWT token | \nPublic | \n- | \n
POST | \n/auth/logout | \nLogout and revoke refresh token | \nRequired | \nlogoutSchema | \n
POST | \n/auth/refresh-token | \nRefresh access token using refresh token | \nPublic | \nrefreshTokenSchema | \n
POST | \n/auth/verify-email | \nVerify email address using token from email | \nPublic | \nverifyEmailSchema | \n
POST | \n/auth/resend-verification | \nResend verification email (rate limited: 3/10min) | \nPublic | \nresendVerificationSchema | \n
POST | \n/auth/forgot-password | \nRequest password reset email (rate limited: 3/15min) | \nPublic | \nforgotPasswordSchema | \n
POST | \n/auth/reset-password | \nReset password using token from email | \nPublic | \nresetPasswordSchema | \n
/users)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
POST | \n/users | \nCreate a user (admin-style) | \nRequired (users:create) | \ncreateUserSchema | \n
GET | \n/users | \nList all users | \nRequired (users:read) | \n- | \n
GET | \n/users/:id | \nFetch a user by ID | \nRequired (users:read) | \n- | \n
PUT | \n/users/:id | \nUpdate user fields (username, email, displayName) | \nRequired (users:update) | \nupdateUserSchema | \n
PUT | \n/users/:id/roles | \nAssign roles to a user | \nRequired (roles:assign) | \nassignRolesSchema | \n
PATCH | \n/users/password | \nUpdate current user's password | \nRequired (users:update) | \nupdatePasswordSchema | \n
DELETE | \n/users/:id | \nRemove a user | \nRequired (users:delete) | \n- | \n
/roles & /permissions)| Method | \nPath | \nDescription | \nAuth | \nValidation | \n
|---|---|---|---|---|
GET | \n/roles | \nList all roles | \nRequired (roles:read) | \n- | \n
GET | \n/roles/:id | \nFetch a role by ID | \nRequired (roles:read) | \n- | \n
POST | \n/roles | \nCreate a new role | \nRequired (roles:create) | \ncreateRoleSchema | \n
PUT | \n/roles/:id | \nUpdate an existing role | \nRequired (roles:update) | \nupdateRoleSchema | \n
DELETE | \n/roles/:id | \nRemove a role | \nRequired (roles:delete) | \n- | \n
GET | \n/permissions | \nList all system permissions | \nRequired (roles:read) | \n- | \n
Auth Required: Endpoints require Authorization: Bearer <token> header.
The API uses Zod for request validation with the following schemas:
\nregisterUserSchema: Validates user registration (username, email, displayName, password)createUserSchema: Validates user creation (username, email, displayName)updateUserSchema: Validates user updates (partial fields)assignRolesSchema: Validates assigning role IDs to a userupdatePasswordSchema: Validates password changes (currentPassword, newPassword, confirmPassword)createRoleSchema: Validates new role details and array of permission IDsupdateRoleSchema: Validates partial updates to a roleverifyEmailSchema: Validates email verification tokenresendVerificationSchema: Validates email for resending verificationforgotPasswordSchema: Validates email for password reset requestsresetPasswordSchema: Validates reset token and new password (min 8 chars, uppercase, lowercase, number)refreshTokenSchema: Validates refresh token from body or cookieslogoutSchema: Validates optional refresh token for revocation on logoutAll schemas include:
\nauthMiddleware (src/middleware/auth.middleware.ts): JWT verification and user authenticationrequireVerified (src/middleware/require-verified.middleware.ts): Ensures user email is verified before accessvalidate (src/middleware/validation.middleware.ts): Zod schema validation for request body/params/queryforgotPasswordLimiter (src/middleware/rate-limit.middleware.ts): Rate limiting for password reset (3 requests per 15 minutes)resendVerificationLimiter (src/middleware/rate-limit.middleware.ts): Rate limiting for resending verification emails (3 requests per 10 minutes)src/validations/<module>.validation.ts for request validation.src/routes/<module>.routes.ts and mount them in src/routes/index.ts.src/controllers/<module>.controller.ts.src/services/<module>.service.ts and reuse AppError for controlled failures.src/repositories/<module>.repository.ts.The API uses a custom AppError class for controlled error handling:
The API implements a robust, secure Refresh Token Rotation mechanism to safely extend user sessions without compromising security.
\nRefresh tokens are configured via environment variables in .env:
ENABLE_REFRESH_TOKEN=true\nREFRESH_TOKEN_IN_JSON=true\nREFRESH_TOKEN_IN_COOKIE=true\nCOOKIE_SAME_SITE=\"strict\" # strict, lax, none\nREFRESH_TOKEN_EXPIRY=604800000 # 7 days in milliseconds\n/auth/refresh-token, their old refresh token is immediately revoked and a new pair is issued. This provides Replay Protection.HTTP-Only cookie (for web clients to prevent XSS) and/or directly in the JSON response body (for mobile apps or specialized clients).COOKIE_SAME_SITE variable. Use strict when the frontend and backend are on the exact same domain, or lax/none (along with secure: true) if operating across subdomains or entirely different domains.The API includes Cross-Origin Resource Sharing (CORS) support to allow requests from different origins (e.g., frontend applications).
\nCORS is configured via environment variables in .env:
# Comma-separated list of allowed origins\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080,https://yourdomain.com\n\n# Allow credentials (cookies, authorization headers)\nCORS_CREDENTIALS=true\nContent-Type, Authorization, X-Requested-With\n\n[!WARNING]\nProduction Security
\n\n
\n- Never use
\n*(wildcard) forALLOWED_ORIGINSin production- Only add trusted domains to the whitelist
\n- Use HTTPS for production origins (e.g.,
\nhttps://yourdomain.com)- Regularly audit the allowed origins list
\n
\n\n[!IMPORTANT]\nCredentials Configuration
\n\n
\n- Set
\nCORS_CREDENTIALS=trueonly if you're using cookies or need to send authorization headers- When credentials are enabled, you cannot use wildcard origins
\n- Frontend must include
\ncredentials: 'include'(fetch) orwithCredentials: true(axios)
Development:
\nALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000,http://localhost:8080\nCORS_CREDENTIALS=true\nProduction:
\nALLOWED_ORIGINS=https://yourdomain.com,https://admin.yourdomain.com\nCORS_CREDENTIALS=true\nCORS Error: \"No 'Access-Control-Allow-Origin' header\"
\nALLOWED_ORIGINSsrc/app.tsCredentials Not Working:
\nCORS_CREDENTIALS=true in .envcredentials: 'include' or withCredentials: true*)The API uses Pino for structured JSON logging with comprehensive audit trails:
\nLogged Events:
\nLog Format:
\nExample Log:
\n{\n \"level\": 30,\n \"time\": 1702890637123,\n \"action\": \"user_login\",\n \"userId\": \"5ba52d7e-07f9-4b15-998f-fb1bf0885e7d\",\n \"email\": \"user@example.com\",\n \"msg\": \"User logged in successfully\"\n}\nThe project is configured to use tsup for efficient bundling.
pnpm builddist/src/server.ts and dependencies to dist/server.jsdist/views/emailspnpm startnode dist/server.jsbuild and start scripts in package.json.DATABASE_URL is set and migrations are run.Always run migrations in production before starting the app:
\npnpm prisma migrate deploy\nThe project uses Vitest for unit and integration testing.
\n# Run all tests (watch mode)\npnpm test\n\n# Run unit tests only\npnpm test:unit\n\n# Run integration tests only\npnpm test:integration\n\n# Run with coverage\npnpm exec vitest run --coverage\nTests are organized in tests/ with separate directories for unit and integration tests:
tests/\nโโโ unit/ # Unit tests (mocked dependencies)\nโ โโโ controllers/\nโ โโโ services/\nโ โโโ middleware/\nโ โโโ utils/\nโโโ integration/ # Integration tests (real database)\n โโโ helpers/ # Test utilities (DB reset)\n โโโ repositories/ # Repository tests\n โโโ routes/ # Route/endpoint tests\nIntegration tests run against a real PostgreSQL database. Ensure your DATABASE_URL points to a test database that can be safely cleared between tests.
\n\n" }, - { - "fullName": "rizukirr/hyprsimple", - "name": "hyprsimple", - "owner": "rizukirr", - "ownerAvatarUrl": "https://avatars.githubusercontent.com/u/230007457?v=4", - "description": "Minimal aesthetic arch linux + hyprland dotfiles", - "metaDescription": "Minimal aesthetic arch linux + hyprland dotfiles", - "url": "https://github.com/rizukirr/hyprsimple", - "homepage": "", - "language": "Shell", - "stars": 19, - "forks": 2, - "topics": [ - "arch-linux", - "dotfiles", - "hyprland", - "hyprland-config", - "linux", - "omarchy" - ], - "updatedAt": "2026-07-21T12:07:06Z", - "pushedAt": "2026-07-21T12:06:23Z", - "latestRelease": { - "name": "v0.2.2", - "tagName": "v0.2.2", - "url": "https://github.com/rizukirr/hyprsimple/releases/tag/v0.2.2", - "publishedAt": "2026-06-29T12:37:11Z" - }, - "archived": false, - "licenseSpdx": "", - "createdAt": "2025-10-04T10:27:45Z", - "openIssues": 0, - "openPullRequests": 0, - "subscribers": 1, - "communityHealth": 28, - "readmeHtml": "[!WARNING]\nIntegration tests truncate all tables before each test. Do not run against a production database.
\n
Minimal Hyprland dotfiles for Arch Linux. Clean, functional, no bloat.
\n\n\n[!Note]\nThis dotfile have builtin muslimtify. A prayer time notification daemon for Linux. Run
\nmuslimtify-removeto uninstall it (package, daemon, waybar module, and CSS). Runmuslimtify-addto re-enable it later. Both commands are idempotent and back up your waybar config to.bakbefore editing.
![\"Home](\"https://raw.githubusercontent.com/rizukirr/hyprsimple/main/assets/image1.png\")
| Power Menu | \nTerminal | \n
|---|---|
![\"Power](\"https://raw.githubusercontent.com/rizukirr/hyprsimple/main/assets/image5.png\") | \n![\"Terminal\"](\"https://raw.githubusercontent.com/rizukirr/hyprsimple/main/assets/image2.png\") | \n
| Menu Launcher | \nTheme Switcher | \n
|---|---|
![\"Menu](\"https://raw.githubusercontent.com/rizukirr/hyprsimple/main/assets/image3.png\") | \n![\"Theme](\"https://raw.githubusercontent.com/rizukirr/hyprsimple/main/assets/image4.png\") | \n
git clone https://github.com/rizukirr/hyprsimple.git\ncd hyprsimple\n./install.sh\n\n\n[!WARNING]\nThese dotfiles have only been tested on a fresh Arch Linux install where Hyprland was selected\nas the desktop during installation. Coming from another desktop environment or compositor\n(KDE, GNOME, etc.) is untested and may require manual cleanup.
\n
If you run into a problem installing hyprsimple, please open an issue โ thank you!
\nTo see the available network interfaces, run wifi. To connect to a network, run wifi <network name> for example wifi \"MY NETWORK\"
Press SUPER + / for interactive viewer with fuzzy search.
| Key | \nAction | \n
|---|---|
SUPER + T | \nOpen terminal (Ghostty) | \n
SUPER + B | \nOpen browser (Brave) | \n
SUPER + A | \nApp launcher (Rofi) | \n
SUPER + F | \nFile manager (Nautilus) | \n
SUPER + O | \nNotes (Obsidian) | \n
SUPER + S | \nAndroid Studio | \n
SUPER + E | \nEmoji picker | \n
SUPER + V | \nClipboard history | \n
SUPER + M | \nColor picker | \n
| Key | \nAction | \n
|---|---|
SUPER + Q | \nKill active window | \n
SUPER + W | \nToggle floating | \n
SUPER + SHIFT + J | \nToggle split (dwindle) | \n
SUPER + H / J / K / L | \nMove focus left / down / up / right | \n
SUPER + SHIFT + Arrow | \nResize window | \n
SUPER + LMB drag | \nMove window | \n
SUPER + RMB drag | \nResize window | \n
| Key | \nAction | \n
|---|---|
SUPER + [1-9, 0] | \nSwitch to workspace 1-10 | \n
SUPER + SHIFT + [1-9, 0] | \nMove window to workspace 1-10 | \n
SUPER + SHIFT + S | \nMove window to scratchpad | \n
SUPER + Scroll | \nCycle through workspaces | \n
| Key | \nAction | \n
|---|---|
SUPER + SHIFT + T | \nSwitch theme | \n
SUPER + SHIFT + W | \nPick wallpaper from current theme | \n
SUPER + ALT + W | \nCycle to next wallpaper | \n
| Key | \nAction | \n
|---|---|
Print | \nScreenshot current monitor | \n
SUPER + Print | \nScreenshot active window | \n
SUPER + ALT + Print | \nScreenshot selected region | \n
SUPER + CTRL + Print | \nScreenshot region to clipboard | \n
| Key | \nAction | \n
|---|---|
SUPER + R | \nRecord region with mic audio | \n
SUPER + SHIFT + R | \nRecord fullscreen with mic audio | \n
SUPER + ALT + R | \nRecord region with system audio | \n
SUPER + SHIFT + ALT + R | \nRecord fullscreen with system audio | \n
SUPER + CTRL + R | \nRecord region without audio | \n
SUPER + CTRL + SHIFT + R | \nRecord fullscreen without audio | \n
| Key | \nAction | \n
|---|---|
Volume Up / Down | \nAdjust volume | \n
Mute | \nToggle mute | \n
Mic Mute | \nToggle microphone mute | \n
Play / Pause | \nMedia play/pause | \n
Next / Prev | \nMedia next/previous track | \n
Brightness Up / Down | \nAdjust screen brightness | \n
Kbd Brightness Up / Down | \nAdjust keyboard backlight | \n
| Key | \nAction | \n
|---|---|
SUPER + ESC | \nPower menu | \n
SUPER + SHIFT + L | \nLock screen | \n
SUPER + X | \nExit Hyprland | \n
CTRL + ESC | \nToggle waybar | \n
SUPER + N | \nToggle nightlight | \n
SUPER + D | \nDismiss notifications | \n
SUPER + SHIFT + I | \nToggle idle lock | \n
SUPER + F10 | \nSwitch audio output | \n
SUPER + SHIFT + M | \nToggle monitor mirroring | \n
SUPER + CTRL + V | \nToggle virtual mirror | \n
SUPER + / | \nShow all keybindings | \n
Helper scripts live in .local/bin (installed to ~/.local/bin, which is on PATH).\nMost are wired to keybindings or waybar; all can also be run directly from a terminal.
| Script | \nDescription | \n
|---|---|
audio-switch.sh | \nCycle through available audio output devices | \n
volume-notify.sh | \nShow the current PipeWire volume via a dunst notification | \n
record-audio.sh | \nRecord audio from the default input to ~/Music | \n
| Script | \nDescription | \n
|---|---|
brightness-notify.sh | \nShow the current screen brightness via a dunst notification | \n
keyboard-brightness.sh | \nControl the keyboard backlight (up / down / cycle) | \n
toggle-nightlight.sh | \nToggle a warm screen temperature via hyprsunset | \n
theme-switcher.sh | \nSwitch theme via rofi picker, or apply one directly by name | \n
theme-apply-templates.sh | \nGenerate themed app configs from a theme's colors.toml | \n
wallpaper-switcher.sh | \nSwitch or cycle wallpaper within the current theme | \n
live-wallpaper-toggle.sh | \nToggle live wallpaper (cycle backgrounds vs. static) | \n
monitor-mirror-toggle.sh | \nToggle extend vs. mirror mode for an external monitor | \n
virtual-mirror-toggle.sh | \nMirror a monitor into a window (via wl-mirror) for screen sharing | \n
| Script | \nDescription | \n
|---|---|
screenshot.sh | \nTake a screenshot (clipboard / window / region / monitor) | \n
screen-record.sh | \nStart/stop screen recording (region or output; mic, internal, or no audio) | \n
screen-record-active.sh | \nReport whether a screen recording is currently running | \n
| Script | \nDescription | \n
|---|---|
wifi.sh | \nList and connect to WiFi networks | \n
wifi-powersave.sh | \nToggle WiFi power saving (on / off) | \n
hotspot.sh | \nCreate a WiFi hotspot with internet sharing | \n
setup-dns.sh | \nConfigure the DNS provider (Cloudflare / Google / DHCP) | \n
| Script | \nDescription | \n
|---|---|
battery-monitor.sh | \nLow-battery notifications and automatic brightness reduction | \n
bluetooth-toggle.sh | \nToggle Bluetooth adapter power | \n
toggle_cpu_mode.sh | \nSwitch CPU governor between performance and powersave | \n
toggle-idle.sh | \nToggle hypridle (lock-on-idle) on/off | \n
hypr-logout.sh | \nGracefully close all windows and stop the Hyprland session | \n
| Script | \nDescription | \n
|---|---|
capslock-notify.sh | \nNotify on Caps Lock state changes | \n
notification-dismiss.sh | \nDismiss all dunst notifications | \n
| Script | \nDescription | \n
|---|---|
search.sh | \nFuzzy file finder (ripgrep + fzf) that opens the result in nvim | \n
search_by_keyword.sh | \nFuzzy content search (ripgrep + fzf) that opens the match in nvim | \n
show-keybindings.sh | \nShow all Hyprland keybindings in a rofi fuzzy-search menu | \n
| Script | \nDescription | \n
|---|---|
hyprsimple-muslimtify.sh | \nAdd or remove the muslimtify prayer-times integration | \n
waybar-muslimtify.sh | \nProvide the waybar module output (next prayer + tooltip) for muslimtify | \n
These are sourced by other files rather than run directly.
\n| Script | \nDescription | \n
|---|---|
bashrc.sh / zsh.sh / fish.fish | \nPer-shell init (zoxide, fzf, starship, aliases) sourced from your shell's rc file | \n
terminal.sh | \nDetect your login shell and wire the matching init script into its rc file | \n
hypr-helpers.sh | \nShared hyprpaper helper functions used by the wallpaper scripts | \n
Troubleshooting and known issues (NVIDIA boot hang, Plymouth blank-screen splash, and\nmore) are documented in FAQ.md.
\nMIT
\n" - }, { "fullName": "rahmanef63/open-silong", "name": "open-silong", @@ -1426,7 +1389,7 @@ "url": "https://github.com/rahmanef63/open-silong", "homepage": "https://silong-os.vercel.app/", "language": "TypeScript", - "stars": 18, + "stars": 19, "forks": 4, "topics": [ "block-editor", @@ -1440,7 +1403,7 @@ "self-hosted", "workspace" ], - "updatedAt": "2026-07-18T02:50:07Z", + "updatedAt": "2026-07-23T04:22:49Z", "pushedAt": "2026-07-17T09:32:16Z", "latestRelease": { "name": "v1.0.0 โ First public release", @@ -1457,6 +1420,43 @@ "communityHealth": 100, "readmeHtml": "Open-source, self-hostable collaborative workspace โ inspired by Notion & Obsidian.
\n![\"Release\"](\"https://img.shields.io/badge/release-v1.0.0-success.svg\"){kind=icon}
\n![\"License:](\"https://img.shields.io/badge/License-MIT-blue.svg\"){kind=icon}
\n![\"Stack\"](\"https://img.shields.io/badge/Next.js-16-black?logo=next.js\"){kind=icon}
![\"React\"](\"https://img.shields.io/badge/React-19-61dafb?logo=react\"){kind=icon}
![\"Convex\"](\"https://img.shields.io/badge/Convex-1.36-cc4444\"){kind=icon}
![\"Tailwind\"](\"https://img.shields.io/badge/Tailwind-v4-38bdf8?logo=tailwindcss\"){kind=icon}
![\"PRs](\"https://img.shields.io/badge/PRs-welcome-brightgreen.svg\"){kind=icon}
Live demo ยท\nDocs ยท\nContributing ยท\nSecurity
\nA block-based workspace for notes, docs, and lightweight databases,\nwith an Obsidian-style knowledge graph on top. Built for teams that\nwant to own their data: self-host the full stack with Docker\nCompose, or run on Convex Cloud free tier. MIT licensed. No vendor\nlock-in.
\n\n\nInspired by Notion & Obsidian.\nopen-silong is an independent, clean-room project โ not affiliated with,\nendorsed by, or connected to Notion Labs, Inc. or Dynalist Inc. It borrows\nideas (the block editor, the knowledge graph), never code or brand assets.\nSee
\nTRADEMARKS.md.
![\"open-silong](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/dashboard.png\")
| Block editor | \nDatabase โ Table | \nDatabase โ Board | \n
|---|---|---|
![\"Block](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/editor.png\") | \n![\"Database](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/database.png\") | \n![\"Database](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/database-board.png\") | \n
| Template gallery | \nAdmin panel | \nCommand palette | \n
![\"Template](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/templates.png\") | \n![\"Admin](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/admin.png\") | \n![\"Command](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/command-palette.png\") | \n
| Dark mode | \nMobile | \nFirst-run setup | \n
|---|---|---|
![\"Dashboard](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/dashboard-dark.png\") | \n![\"Mobile](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/mobile-home.png\") | \n![\"First-run](\"https://raw.githubusercontent.com/rahmanef63/open-silong/main/docs/media/setup.png\") | \n
\n\nCaptured live on silong-os.vercel.app,\nsigned in as the workspace superadmin (demo workspace seeded from the\n
\n/setupwizard).
[[wikilink]], @mention, #tag, and database row โ with backlinks\nand unresolved \"ghost\" nodes. Live d3-force layout with tunable\nforces, cluster tinting, and focus/neighbourhood highlighting.@page mentions,\npresence indicators, version snapshots./mcp/v1).One-click: Deploy with Vercel โ\nonly asks for CONVEX_DEPLOY_KEY (create a project at\ndashboard.convex.dev โ Settings โ Deploy\nKeys). The build deploys the Convex functions, provisions the auth keys,\nand injects NEXT_PUBLIC_CONVEX_URL automatically. Your first visit\nlands on the /setup wizard: claim the owner (superadmin) account and\nseed the template gallery + demo workspace in one click.
Local development:
\ngit clone https://github.com/rahmanef63/open-silong.git\ncd open-silong\npnpm install\ncp .env.example .env.local # fill NEXT_PUBLIC_CONVEX_URL after step 4\nnpx convex dev # creates Convex Cloud project, prints URL\npnpm dev # http://localhost:3000\nConvex Cloud free tier covers small teams. Full walk-through in\nDEPLOY.md.
git clone https://github.com/rahmanef63/open-silong.git\ncd open-silong\ncp .env.example .env.local # fill INSTANCE_*, JWT_*, POSTGRES_URL\ndocker compose up -d # Convex backend on port 3210\npnpm install\npnpm exec convex deploy --yes # push schema + functions\npnpm dev # http://localhost:3000\nFull Dokploy + Traefik + Postgres + S3 setup in\nDEPLOY.md.
convex/auth.ts already wires Google โ just provide credentials:
# 1. Google Cloud Console โ APIs & Services โ Credentials โ Create OAuth 2.0\n# client (Web app). Authorized redirect URI:\n# https://<your-CONVEX_SITE_ORIGIN>/api/auth/callback/google\n# 2. Set on Convex backend\npnpm exec convex env set AUTH_GOOGLE_ID <client-id>.apps.googleusercontent.com\npnpm exec convex env set AUTH_GOOGLE_SECRET <client-secret>\nThe \"Sign in with Google\" button in /auth activates automatically\nonce those two env vars are set. Step-by-step including consent screen\nsetup + adding GitHub/Apple/Discord providers:\nDEPLOY.md#google-oauth-sign-in-optional.
silong-os.vercel.app โ public demo.\nLands you straight in a guest workspace (no sign-up needed), or create\nan email + password account to keep your data. Instance is shared.
\nLooking for the UI as a localStorage-only starter? The same editor\nships as notion-page-clone-os in the\nrahman-resources template\nmarketplace:
npx rahman-resources@latest add notion-page-clone-os\n| Layer | \nChoice | \nWhy | \n
|---|---|---|
| Frontend | \nNext 16 (App Router) + React 19 | \nRSC, streaming, file-based routing | \n
| Styling | \nTailwind v4 + shadcn/ui | \nTheme tokens, primitives, dark mode | \n
| Backend | \nConvex 1.36 (self-hostable) | \nRealtime, optimistic, typed end-to-end | \n
| Auth | \n@convex-dev/auth | \nMagic-link, OAuth-ready, no Clerk | \n
| Storage | \nConvex file storage OR S3 adapter | \nPluggable per slice | \n
| Search | \nConvex full-text index | \nNo external search service | \n
| Deploy | \nDocker Compose + Traefik (self-host) OR Convex Cloud | \nPick your trade-off | \n
A one-screen system view. The full set โ data model, auth/authz flow, slice\ngraph, and the memory-graph pipeline โ lives in\ndocs/architecture/diagrams.md.
flowchart LR\n B[\"Browser<br/>Next 16 ยท React 19\"] --> P[\"proxy.ts<br/>optimistic auth gate\"]\n B -- \"reactive queries\" --> C[\"Convex backend<br/>queries ยท mutations<br/>in-handler authz\"]\n P --> C\n C --> S[\"schema.ts ยท 32 tables\"]\n S --> DB[(\"Postgres / Convex Cloud\")]\n C --> F[(\"Files: Convex blob / S3\")]\n A[\"AI agents\"] -- \"Notion-canonical JSON\" --> H[\"MCP HTTP surface\"] --> C\nRepository layout:
\napp/ Next 16 App Router routes\n dashboard/* Authenticated surfaces (pages, db, settings, โฆ)\n share/[id] Public read-only share surface\n preview/* Marketing + sandbox\n\nfrontend/\n slices/<name>/ Vertical feature slices โ see docs/api/slices.md\n shared/ Cross-slice primitives, providers, store hooks\nproxy.ts Convex auth optimistic gate (not the security boundary)\n\nconvex/\n features/<name>/ Per-feature backend (schema + queries + mutations)\n _shared/ Auth helpers, rate limit, workspace gates\n http.ts Public HTTP routes (share, MCP)\n mcp/ MCP HTTP surface (Notion-canonical JSON)\n\ndocker-compose.yml Convex self-hosted (port 3210, Traefik-frontable)\nThe codebase follows a slice architecture: each feature lives in\nfrontend/slices/<name>/ with optional convex/features/<name>/\nmirror. Cross-slice imports go through the barrel only. See\nCONTRIBUTING.md for the rules.
| Topic | \nWhere | \n
|---|---|
| Per-slice API + UX docs | \ndocs/api/ | \n
| Architecture diagrams (system ยท data model ยท flows) | \ndocs/architecture/diagrams.md | \n
| Deploy walkthroughs (cloud + self-host + Dokploy) | \nDEPLOY.md | \n
| Slice catalog (every feature in one page) | \ndocs/api/slices.md | \n
| Architecture decisions + audit notes | \ndocs/audit/ | \n
| Contributing guide | \nCONTRIBUTING.md | \n
| Security policy | \nSECURITY.md | \n
| Code of Conduct | \nCODE_OF_CONDUCT.md | \n
| Trademarks + inspiration + legal notes | \nTRADEMARKS.md | \n
| Changelog | \nCHANGELOG.md | \n
See docs/notion-clone/ROADMAP.md\nfor the full backlog.
Bug reports, feature ideas, doc fixes, and code PRs are all welcome.\nRead CONTRIBUTING.md for dev setup, slice\narchitecture, and PR conventions.
By participating, you agree to the\nCode of Conduct.
\nFound a vulnerability? Please don't open a public issue. Email\nsecurity@rahmanef.com or use a private GitHub Security Advisory โ\nsee SECURITY.md for SLAs and scope.
MIT ยฉ 2026 Rahman Effendi and open-silong contributors.
\nopen-silong is an independent open-source project. It is not\naffiliated with, sponsored by, endorsed by, or associated with Notion\nLabs, Inc. or Dynalist Inc. (the maker of Obsidian) in any way.
It is inspired by Notion (the block editor +\nlightweight databases) and Obsidian (the\nlocal-first knowledge graph). \"Notion\" and \"Obsidian\" are trademarks of\ntheir respective owners, used here only in a nominative / descriptive\nsense to identify familiar UI patterns โ analogous to how an \"iPhone\ncase\" advertises compatibility without claiming any link to Apple.
\nopen-silong is a clean-room implementation built independently on\nConvex, Next.js,\nshadcn/ui, and the open-source\nd3-force layout. No proprietary Notion or\nObsidian code, design files, brand assets, or trade secrets are used.\nImport/export adapters target documented public file formats purely for\ninteroperability.
\nA full plain-language explanation โ the idea/expression distinction,\nnominative fair use, and international (EU/CJEU) anchors โ is in\nTRADEMARKS.md. If you represent a rights holder\nand have a good-faith concern, reach us via the email in\nSECURITY.md โ we will adjust naming, disclaimers, or\nsurfaces in good faith.
\n\n" }, + { + "fullName": "rizukirr/hyprsimple", + "name": "hyprsimple", + "owner": "rizukirr", + "ownerAvatarUrl": "https://avatars.githubusercontent.com/u/230007457?v=4", + "description": "Minimal aesthetic arch linux + hyprland dotfiles", + "metaDescription": "Minimal aesthetic arch linux + hyprland dotfiles", + "url": "https://github.com/rizukirr/hyprsimple", + "homepage": "", + "language": "Shell", + "stars": 19, + "forks": 2, + "topics": [ + "arch-linux", + "dotfiles", + "hyprland", + "hyprland-config", + "linux", + "omarchy" + ], + "updatedAt": "2026-07-21T12:07:06Z", + "pushedAt": "2026-07-21T12:06:23Z", + "latestRelease": { + "name": "v0.2.2", + "tagName": "v0.2.2", + "url": "https://github.com/rizukirr/hyprsimple/releases/tag/v0.2.2", + "publishedAt": "2026-06-29T12:37:11Z" + }, + "archived": false, + "licenseSpdx": "", + "createdAt": "2025-10-04T10:27:45Z", + "openIssues": 0, + "openPullRequests": 0, + "subscribers": 1, + "communityHealth": 28, + "readmeHtml": "This notice is not legal advice; consult a qualified attorney for\nguidance specific to your jurisdiction and use.
\n
Minimal Hyprland dotfiles for Arch Linux. Clean, functional, no bloat.
\n\n\n[!Note]\nThis dotfile have builtin muslimtify. A prayer time notification daemon for Linux. Run
\nmuslimtify-removeto uninstall it (package, daemon, waybar module, and CSS). Runmuslimtify-addto re-enable it later. Both commands are idempotent and back up your waybar config to.bakbefore editing.
| Power Menu | \nTerminal | \n
|---|---|
| \n | \n
| Menu Launcher | \nTheme Switcher | \n
|---|---|
| \n | \n
git clone https://github.com/rizukirr/hyprsimple.git\ncd hyprsimple\n./install.sh\n\n\n[!WARNING]\nThese dotfiles have only been tested on a fresh Arch Linux install where Hyprland was selected\nas the desktop during installation. Coming from another desktop environment or compositor\n(KDE, GNOME, etc.) is untested and may require manual cleanup.
\n
If you run into a problem installing hyprsimple, please open an issue โ thank you!
\nTo see the available network interfaces, run wifi. To connect to a network, run wifi <network name> for example wifi \"MY NETWORK\"
Press SUPER + / for interactive viewer with fuzzy search.
| Key | \nAction | \n
|---|---|
SUPER + T | \nOpen terminal (Ghostty) | \n
SUPER + B | \nOpen browser (Brave) | \n
SUPER + A | \nApp launcher (Rofi) | \n
SUPER + F | \nFile manager (Nautilus) | \n
SUPER + O | \nNotes (Obsidian) | \n
SUPER + S | \nAndroid Studio | \n
SUPER + E | \nEmoji picker | \n
SUPER + V | \nClipboard history | \n
SUPER + M | \nColor picker | \n
| Key | \nAction | \n
|---|---|
SUPER + Q | \nKill active window | \n
SUPER + W | \nToggle floating | \n
SUPER + SHIFT + J | \nToggle split (dwindle) | \n
SUPER + H / J / K / L | \nMove focus left / down / up / right | \n
SUPER + SHIFT + Arrow | \nResize window | \n
SUPER + LMB drag | \nMove window | \n
SUPER + RMB drag | \nResize window | \n
| Key | \nAction | \n
|---|---|
SUPER + [1-9, 0] | \nSwitch to workspace 1-10 | \n
SUPER + SHIFT + [1-9, 0] | \nMove window to workspace 1-10 | \n
SUPER + SHIFT + S | \nMove window to scratchpad | \n
SUPER + Scroll | \nCycle through workspaces | \n
| Key | \nAction | \n
|---|---|
SUPER + SHIFT + T | \nSwitch theme | \n
SUPER + SHIFT + W | \nPick wallpaper from current theme | \n
SUPER + ALT + W | \nCycle to next wallpaper | \n
| Key | \nAction | \n
|---|---|
Print | \nScreenshot current monitor | \n
SUPER + Print | \nScreenshot active window | \n
SUPER + ALT + Print | \nScreenshot selected region | \n
SUPER + CTRL + Print | \nScreenshot region to clipboard | \n
| Key | \nAction | \n
|---|---|
SUPER + R | \nRecord region with mic audio | \n
SUPER + SHIFT + R | \nRecord fullscreen with mic audio | \n
SUPER + ALT + R | \nRecord region with system audio | \n
SUPER + SHIFT + ALT + R | \nRecord fullscreen with system audio | \n
SUPER + CTRL + R | \nRecord region without audio | \n
SUPER + CTRL + SHIFT + R | \nRecord fullscreen without audio | \n
| Key | \nAction | \n
|---|---|
Volume Up / Down | \nAdjust volume | \n
Mute | \nToggle mute | \n
Mic Mute | \nToggle microphone mute | \n
Play / Pause | \nMedia play/pause | \n
Next / Prev | \nMedia next/previous track | \n
Brightness Up / Down | \nAdjust screen brightness | \n
Kbd Brightness Up / Down | \nAdjust keyboard backlight | \n
| Key | \nAction | \n
|---|---|
SUPER + ESC | \nPower menu | \n
SUPER + SHIFT + L | \nLock screen | \n
SUPER + X | \nExit Hyprland | \n
CTRL + ESC | \nToggle waybar | \n
SUPER + N | \nToggle nightlight | \n
SUPER + D | \nDismiss notifications | \n
SUPER + SHIFT + I | \nToggle idle lock | \n
SUPER + F10 | \nSwitch audio output | \n
SUPER + SHIFT + M | \nToggle monitor mirroring | \n
SUPER + CTRL + V | \nToggle virtual mirror | \n
SUPER + / | \nShow all keybindings | \n
Helper scripts live in .local/bin (installed to ~/.local/bin, which is on PATH).\nMost are wired to keybindings or waybar; all can also be run directly from a terminal.
| Script | \nDescription | \n
|---|---|
audio-switch.sh | \nCycle through available audio output devices | \n
volume-notify.sh | \nShow the current PipeWire volume via a dunst notification | \n
record-audio.sh | \nRecord audio from the default input to ~/Music | \n
| Script | \nDescription | \n
|---|---|
brightness-notify.sh | \nShow the current screen brightness via a dunst notification | \n
keyboard-brightness.sh | \nControl the keyboard backlight (up / down / cycle) | \n
toggle-nightlight.sh | \nToggle a warm screen temperature via hyprsunset | \n
theme-switcher.sh | \nSwitch theme via rofi picker, or apply one directly by name | \n
theme-apply-templates.sh | \nGenerate themed app configs from a theme's colors.toml | \n
wallpaper-switcher.sh | \nSwitch or cycle wallpaper within the current theme | \n
live-wallpaper-toggle.sh | \nToggle live wallpaper (cycle backgrounds vs. static) | \n
monitor-mirror-toggle.sh | \nToggle extend vs. mirror mode for an external monitor | \n
virtual-mirror-toggle.sh | \nMirror a monitor into a window (via wl-mirror) for screen sharing | \n
| Script | \nDescription | \n
|---|---|
screenshot.sh | \nTake a screenshot (clipboard / window / region / monitor) | \n
screen-record.sh | \nStart/stop screen recording (region or output; mic, internal, or no audio) | \n
screen-record-active.sh | \nReport whether a screen recording is currently running | \n
| Script | \nDescription | \n
|---|---|
wifi.sh | \nList and connect to WiFi networks | \n
wifi-powersave.sh | \nToggle WiFi power saving (on / off) | \n
hotspot.sh | \nCreate a WiFi hotspot with internet sharing | \n
setup-dns.sh | \nConfigure the DNS provider (Cloudflare / Google / DHCP) | \n
| Script | \nDescription | \n
|---|---|
battery-monitor.sh | \nLow-battery notifications and automatic brightness reduction | \n
bluetooth-toggle.sh | \nToggle Bluetooth adapter power | \n
toggle_cpu_mode.sh | \nSwitch CPU governor between performance and powersave | \n
toggle-idle.sh | \nToggle hypridle (lock-on-idle) on/off | \n
hypr-logout.sh | \nGracefully close all windows and stop the Hyprland session | \n
| Script | \nDescription | \n
|---|---|
capslock-notify.sh | \nNotify on Caps Lock state changes | \n
notification-dismiss.sh | \nDismiss all dunst notifications | \n
| Script | \nDescription | \n
|---|---|
search.sh | \nFuzzy file finder (ripgrep + fzf) that opens the result in nvim | \n
search_by_keyword.sh | \nFuzzy content search (ripgrep + fzf) that opens the match in nvim | \n
show-keybindings.sh | \nShow all Hyprland keybindings in a rofi fuzzy-search menu | \n
| Script | \nDescription | \n
|---|---|
hyprsimple-muslimtify.sh | \nAdd or remove the muslimtify prayer-times integration | \n
waybar-muslimtify.sh | \nProvide the waybar module output (next prayer + tooltip) for muslimtify | \n
These are sourced by other files rather than run directly.
\n| Script | \nDescription | \n
|---|---|
bashrc.sh / zsh.sh / fish.fish | \nPer-shell init (zoxide, fzf, starship, aliases) sourced from your shell's rc file | \n
terminal.sh | \nDetect your login shell and wire the matching init script into its rc file | \n
hypr-helpers.sh | \nShared hyprpaper helper functions used by the wallpaper scripts | \n
Troubleshooting and known issues (NVIDIA boot hang, Plymouth blank-screen splash, and\nmore) are documented in FAQ.md.
\nMIT
\n" + }, { "fullName": "RafiulM/warungos", "name": "warungos", @@ -1591,8 +1591,8 @@ "stars": 14, "forks": 2, "topics": [], - "updatedAt": "2026-07-22T12:27:39Z", - "pushedAt": "2026-07-22T12:26:04Z", + "updatedAt": "2026-07-23T05:17:32Z", + "pushedAt": "2026-07-23T05:17:02Z", "latestRelease": { "name": "v1.2.54", "tagName": "v1.2.54", @@ -1850,7 +1850,7 @@ "semantic-search" ], "updatedAt": "2026-07-03T02:36:07Z", - "pushedAt": "2026-07-23T02:39:54Z", + "pushedAt": "2026-07-23T04:31:41Z", "latestRelease": { "name": "OpenEmpiric v1.0.5", "tagName": "v1.0.5",