Datasette

Datasette turns any SQLite database into an instantly-queryable JSON API and interactive web UI — zero configuration required. Point it at a .db file and every table becomes a REST endpoint (/db/table.json?_where=...), SQL queries run via the API (/db.json?sql=...), and faceting/full-text search are built in. For agents: publish internal SQLite datasets as queryable APIs without building custom CRUD layers, explore SQLite data programmatically via the REST API, or embed Datasette as a read-only data access layer in larger pipelines. Extremely lightweight (pip install datasette). Plugin ecosystem extends with GraphQL, authentication, visualization, and export plugins. Confidence is docs-derived.

Automatic REST API for every SQLite database and table. Key endpoints: GET /{db}/{table}.json lists rows (paginated; filtering via ?_where=column=value, ?_search=term; sorting via ?_sort=column). GET /{db}.json?sql={sql} runs arbitrary SQL queries and returns results as JSON ({columns: [...], rows: [[...], ...]}). GET /-/databases.json lists all loaded databases. GET /{db}.json lists tables. Pagination: ?_next= cursor from Link header. Formats: .json (default), .csv, .tsv. ?_shape= parameter controls JSON response shape (arrays, objects, array). ?_facet= parameter adds facet counts to response. Datasette configuration via metadata.json for table labels, access controls, and custom SQL views.

Datasette supports multiple authentication plugins. Default: no auth (all data public). datasette-auth-tokens plugin: API key via Authorization: Bearer token. datasette-github-auth and datasette-google-auth for OAuth. Per-table access control via allow blocks in metadata.json. HTTPS: operator-managed (deploy behind TLS proxy). For public read-only data publishing, no auth is common. For sensitive data, use authentication plugins + HTTPS. Datasette itself is read-only — no write API (by design); mutations must go via direct SQLite or separate write layer.

Datasette is a Python ASGI app — reliability is infrastructure-dependent. SQLite file must be readable; no concurrent writes from other processes while Datasette serves (use WAL mode for concurrent reads). SQL query timeout configurable (default: no limit; set sql_time_limit_ms in config). Immutable mode (datasette serve --immutable) prevents database modification. Error responses: JSON with {ok: false, error: "...", status: 404}. 403 for unauthorized access. Long-running SQL queries: timeout enforcement prevents abuse. No built-in SLA — self-hosted reliability is operator-managed. datasette-lite (WebAssembly, browser-only) for client-side deployments.

docs.datasette.io is excellent — complete reference, plugin documentation, deployment guides (Fly.io, Cloud Run, Heroku, Render), tutorial series, and plugin development guide. Created by Simon Willison (co-creator of Django) — documentation reflects deep care for developer experience. Getting started: pip install datasette; datasette serve mydata.db — running in 30 seconds. Large plugin ecosystem (100+ plugins on PyPI). Community via Datasette Discord, GitHub (8k+ stars), and creator's blog (simonwillison.net). Actively maintained. datasette-utils and sqlite-utils are companion CLI tools for preparing SQLite data.