Bump astro from 6.4.8 to 7.1.1 - #44
Open
dependabot[bot] wants to merge 4 commits into
Open
Conversation
…nel (#43) * Add browser-local RAG assistant ("Ask about Carl") A recruiter-facing assistant on every page that answers from this site's own content, with retrieval and inference running entirely in the visitor's browser. No backend, no APIs, no analytics. Pipeline: question -> BM25F retrieval over a build-time corpus -> grounding gate -> local model writes prose -> source links rendered from retrieval metadata. The model only turns passages into sentences; it never decides whether an answer exists and never produces a link. Grounding: - Approved sources are declared in one place, the header of src/pages/assistant/corpus.json.ts. Nothing is crawled. - A question whose subject is absent from the corpus is declined by name before any model runs. - URLs, citations and addresses not present in the evidence are stripped in code. Model: Qwen3-0.6B q4f16 on WebGPU (~580 MB, cached). Chosen by measurement -- SmolLM2-360M contradicted its own evidence, and Qwen2.5-0.5B invented a project name, a definition and a framework. Falls back to quoting the closest passage when WebGPU, fp16 shaders or an unmetered connection are unavailable. Cost to a visitor who never opens it: ~5.4 KB gzipped and zero requests. The corpus, retriever and runtime are dynamically imported on first intent. Checks: npm run assistant:eval (25 retrieval cases), assistant:test (controller, grounding gate, and a provenance check asserting every word in the corpus comes from an approved source file). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Fix assistant model loading, launcher icon, and input collapse The local model never loaded for anyone. Loading transformers.js from jsDelivr's raw dist path threw before a single model byte was fetched: TypeError: Failed to resolve module specifier "onnxruntime-web/webgpu" The published dist/transformers.web.min.js keeps a bare import specifier that a browser cannot resolve without an import map. Switched to jsDelivr's `+esm` build, which rewrites those specifiers to absolute URLs. Verified end to end in headless Chrome: import, ONNX session creation, and generation all succeed. Also: - Launcher icon now fills the button, leaving a hairline of white, and reads as a swept three-blade propeller rather than petals. - The input could collapse to nothing. The autosizer ran while the panel was still display:none, where scrollHeight is 0, and wrote that back as a height — pushing the placeholder out of view. Added a CSS one-line floor, a guard for the unmeasurable case, and a re-measure when the panel opens. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Add a reset control to the assistant panel Clears the conversation and returns the panel to exactly how it looks on first open — intro text and suggestion chips included, by re-inserting a clone of the empty state captured at mount. Any in-flight generation is aborted, and the status line confirms the clear before handing back to the real model status. The button sits beside the close control and only appears once there is something to clear, so it is never a dead affordance. Two bugs surfaced while verifying this, both of which affected desktop: - Autofocus on open was gated on `pointer: fine`, which is false on any device with no pointing device at all, not just touch. Now gated on `pointer: coarse` so only a touch keyboard is spared. - The layout flush that `resizeInput` performs has to happen before focusing: focusing an element the browser still considers unrendered silently no-ops. Adds scripts/drive-assistant.mjs, which drives the real panel in headless Chrome over the DevTools Protocol and asserts open, ask, sources, reset and close. Dependency-free — Node's built-in fetch and WebSocket. Screenshots did not catch either bug above; this does. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * Upgrade assistant model from Qwen3-0.6B to Qwen3-1.7B - 1.7B model has real reasoning capacity for multi-turn context - Download size: ~1.4 GB (up from ~580 MB) - Relocate assistant modules from src/lib/assistant/ to src/assistant/ - Update banner text to 'the wait is worth it' - Update model labels throughout UI * Revert model back to Qwen3-0.6B — 1.7B download too slow for testing * feat: swap assistant from local ONNX to server-side llama.cpp via funnel - Add server-engine.ts: OpenAI-compatible SSE streaming client for the llama.cpp funnel API - Wire server engine into controller.ts with local ONNX fallback - Pass funnel URL/model via data attributes from Astro component - Add VITE_ASSISTANT_FUNNEL_URL and VITE_ASSISTANT_MODEL env vars - Update banner text to show Qwen3.6-35B-A3B-MTP · llama.cpp - Server engine is preferred; local ONNX drops to fallback on failure * chore: pass assistant funnel URL/model as secrets in deploy workflow * feat: add API key authentication for llama.cpp funnel - Add --api-key to llama.cpp startup script - Add LLAMA_API_KEY env file for systemd service - Server engine sends Authorization: Bearer header - Add VITE_ASSISTANT_API_KEY env var to Astro config - Pass API key through data attributes (component → ui → controller) - Add ASSISTANT_API_KEY secret to deploy workflow - GitHub secret already configured --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Bumps [astro](https://github.com/withastro/astro/tree/HEAD/packages/astro) from 6.4.8 to 7.1.1. - [Release notes](https://github.com/withastro/astro/releases) - [Changelog](https://github.com/withastro/astro/blob/main/packages/astro/CHANGELOG.md) - [Commits](https://github.com/withastro/astro/commits/astro@7.1.1/packages/astro) --- updated-dependencies: - dependency-name: astro dependency-version: 7.1.1 dependency-type: direct:production ... Signed-off-by: dependabot[bot] <support@github.com>
Contributor
Author
|
A newer version of astro exists, but since this PR has been edited by someone other than Dependabot I haven't updated it. You'll get a PR for the updated version as normal once this PR is merged. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bumps astro from 6.4.8 to 7.1.1.
Release notes
Sourced from astro's releases.
... (truncated)
Changelog
Sourced from astro's changelog.
... (truncated)
Commits
91c645a[ci] release (#17403)4b03702Normalize domain-based i18n request paths consistently (#17399)fdd673c[ci] release (#17387)d9f99e1Preserve CSP resource defaults for element directives (#17394)092da56Add more escaping for rendered content (#17393)186a1e7fix(logger): allows setting in updateConfig() (#17391)4407483fix(logger): useconsoleAPI in JSON logger for cross-runtime compatibility...16de021feat(logger): allow passing entrypoint as URL (#17389)ed71eaffix(logger): make AstroLoggerDestination not generic (#17390)30698a2feat: chunk data store (#17296)Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
@dependabot rebasewill rebase this PR@dependabot recreatewill recreate this PR, overwriting any edits that have been made to it@dependabot show <dependency name> ignore conditionswill show all of the ignore conditions of the specified dependency@dependabot ignore this major versionwill close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this minor versionwill close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this dependencywill close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)You can disable automated security fix PRs for this repo from the Security Alerts page.