D8R

Public alpha · 0.1.0a1 — a keyboard-first terminal workspace for exploring local data and writing structured queries. Pronounced “Datuhr”.

D8R combines a schema explorer, context-aware completion, source-specific saved query pages, reusable functions, and a results grid with CSV export. It is for analysts and developers who want to explore tables from the terminal and are willing to try a small query language instead of a general-purpose SQL editor.

Start with the bundled synthetic data or a local Parquet file. Remote connectors are available, but a connector or SQL dialect appearing in a picker is not a promise of live-backend compatibility. This is an early release, not a claim of production readiness. See the tested matrix and known limits.

Reproduce the demo · Full user guide · Alpha support and feedback

Why a backslash language?

The \ prefix gives the editor an explicit command and completion context: choose a table after \from, columns after \select, and a column, operator, then values after \where. The same structured document supports editing, validation, saved queries, and parameterized functions.

D8R parses that document into an AST and builds Ibis expressions. The document, data source, and SQL rendering dialect are separate choices:

• The source supplies the schema and the connection used by Run.
• The dialect selects SQL rendering for Compile, which does not execute the query or move it to another database.
• Backend and expression support still vary; SQL compilation is not evidence of successful execution on a vendor's service.

This is deliberately not an arbitrary SQL editor, and the language does not cover every SQL feature. The language contract and compatibility reference describe its scope. The terminal interface is built with Textual; no browser or application server is needed.

Install the alpha and run a query

Install Git and uv, then clone and select the alpha tag:

git clone https://github.com/harryvgiunta/d8r.git
cd d8r
git checkout v0.1.0-alpha.1
uv sync --locked
uv run d8r

Python 3.13 or newer is required; the repository pins its development Python in .python-version. uv creates a project-local .venv using uv.lock. For release-wheel installation and platform-specific qualifications, see the alpha guide.

A fresh workspace opens on demo, backed by bundled synthetic Parquet data and an in-process DuckDB engine. No database account or AI provider is needed. Try this query against that source:

\from events
\select event_type, count(*) as event_count
\group event_type
\order event_count desc
\limit 10

Press Ctrl+Enter (or F5) to run it and view the event counts. Press Ctrl+K (or F6) to compile it to SQL without executing it. Type \ to open completion; Enter or Tab accepts a suggestion. Ctrl+Q quits from the query workspace. The demo walkthrough provides the reproducible workflow behind the image above.

Open your own local Parquet file

1. Press Ctrl+O and choose Local Parquet file · DuckDB under Backend.

2. Enter an existing .parquet or .pq file path, then choose Add.

3. The file appears as the table data. Start with:

   \from data
   \limit 50
   

4. Press Ctrl+Enter. Use the schema explorer and completion to choose columns, then save query drafts in Pages or use Export CSV… in Results.

This entry path accepts one local file, not a URL, directory, or glob. Queries read it through DuckDB/Ibis without modifying the original file. The file is not uploaded by opening it; optional AI sharing is a separate boundary below. See local file behavior for persistence and reconnection details.

Optional AI and privacy

AI is not required to load data, edit, compile, or run queries. To enable it, open Settings → AI provider and configure an OpenAI-compatible Chat Completions endpoint, model, and key. D8R does not host a provider.

Sending a message shares the current document or function draft with your chosen provider. The assistant can also request schema/capability information, recent successful queries, and saved function definitions. Allow Sample Data, Test AI Queries, and Automatically Apply AI Edits are off by default. Opting into samples can send data values; opting into query tests permits bounded read-only execution. Turning a permission off does not erase previously shared conversation context. Use HTTPS for remote providers and review their policies.

Leave AI unconfigured when that context must remain local. Credentials use the OS keyring when remembered, but local query history, chats, drafts, and connection metadata are plaintext under ~/.d8r (or D8R_DATA_DIR). Protect those files, backups, exports, and diagnostic logs. Read the AI permissions and storage reference before use.

Limits, safety, and feedback

D8R runs with your user account's permissions; it is not a sandbox. Use trusted files and least-privileged database roles. Row limits cap returned rows, not query cost. Remote databases require their own infrastructure, drivers, and appropriate TLS configuration. Vendor-shaped bundled sources are mocks that execute locally, not live connections to those vendors.

Windows/Python 3.13 is the exercised native path for this alpha. macOS/Linux have not been exercised; PostgreSQL opt-in integration tests were not run; D1 HTTP coverage is mocked with SQLite; generic adapters are not live-certified. The alpha guide is the support matrix and known-limit reference.

Report bugs and usability feedback through GitHub Issues. Include the alpha version, OS/Python/terminal, backend, a minimal reproduction, and expected versus actual behavior. Remove credentials, private queries, schema, and data before sharing logs or screenshots. See feedback guidance.

The full user guide retains the detailed keyboard, workspace, language, functions, connections, AI, configuration, and developer references.

License

MIT · Copyright (c) 2026 Harry Giunta.