Simon Willison’s latest post introduces the idea of “Datasette Apps” — small, portable tools powered by SQLite and SQL, built on top of Datasette. Here’s a practical way to apply that concept today and ship useful internal tools fast. Source: simonwillison.net.
Why this matters
- Portable by default: A single SQLite file plus a lightweight server can power a full app.
- Fast to ship: Most functionality is SQL + a few configs instead of heavy frameworks.
- Cheap to run: Minimal infra; runs locally, in a container, or fully in-browser via WebAssembly.
- Great for “micro-tools”: internal dashboards, audits, mini-APIs, data reviews.
Build one in an afternoon
- Start with SQLite: model your data in a single
.db. Keep tables narrow and add indexes for common filters. Docs: sqlite.org. - Serve with Datasette: point it at your database and get tables, filters, and search out of the box. Docs: docs.datasette.io.
- Add “canned queries”: parameterized SQL for tasks users run often (e.g., ?start=&end=, ?owner=). This provides safe, reusable actions.
- Visualize quickly: use plugins like charts or maps to turn results into dashboards. Explore plugins on GitHub.
- Ship it: run as a Docker container or deploy to your favorite host. See Datasette’s deployment guide: Deploying Datasette.
- Option: run in the browser with Datasette Lite (WebAssembly) for zero-install, fully client-side demos.
Great use cases
- Review queues: triage content, flags, or support tickets with saved filters.
- Data quality checks: canned queries that highlight duplicates, nulls, or outliers.
- Audit and compliance: time-bounded queries with exports for auditors.
- Searchable reference: turn CSVs into a queryable, shareable internal index.
Tips and cautions
- Parameterize everything: avoid string interpolation; use named parameters to prevent SQL injection.
- Keep it explainable: document each canned query (inputs, logic, outputs) in the metadata.
- Design for read-mostly: SQLite excels at reads; batch writes or schedule ingests.
- Test with real data sizes: confirm indexes and query plans hold up as the dataset grows.
Takeaway
“Datasette Apps” are a smart pattern: assemble SQL-first micro-tools that are easy to deploy, cheap to run, and good enough to solve real problems—fast.
Get nuggets like this in your inbox. Subscribe to The AI Nuggets newsletter.
