Datasette

// Overview

Datasette is an open source multi-tool for exploring and publishing data. You point it at a SQLite database file and it starts a web server that turns that file into a website: every table gets a browsable page with sorting and filtering, every query gets a shareable URL, and every page has a matching JSON endpoint. Nothing has to be modelled or configured first — the schema already in the file is the interface.

The second half of the tool is publishing. Datasette ships commands that package a database together with the server and deploy it to a host such as Heroku or Google Cloud Run, so a dataset that was a local file becomes a public, linkable site. A metadata file carries licence and source attribution alongside the data, which is the part that matters for journalism and archival work. Datasette Lite goes further and runs the whole thing as WebAssembly in the browser, with no server at all.

Datasette is maintained by Simon Willison and needs Python 3.8 or later. A large plugin ecosystem extends it with authentication, visualisations, extra output formats and more, and it sits at the centre of a family of related tools — sqlite-utils for building and reshaping the databases it reads, and the Dogsheep personal-analytics projects built on top of both. Its audience is data journalists, museum curators, archivists and researchers, and in LLM work it is a convenient way to browse a scraped or logged corpus before that corpus is embedded or evaluated.

// What it does

• Serve any SQLite database as a website where each table is a browsable, sortable, filterable page
• Get a JSON API for every table and every custom SQL query, with no extra code
• Publish a database to Heroku or Google Cloud Run with a single publish command
• Attach licence and source attribution to a dataset through a metadata file
• Run entirely in the browser through Datasette Lite, a WebAssembly build that needs no server
• Extend the server with plugins for authentication, charts, extra export formats and custom pages

// Getting started

Install from PyPI or Homebrew, then point the server at a database file. Datasette requires Python 3.8 or higher.

pip install datasette
# or on macOS
brew install datasette

datasette serve path/to/database.db

curl 'http://localhost:8001/database/table.json'

datasette publish cloudrun path/to/database.db --service my-data
datasette publish heroku path/to/database.db

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Publish a public, linkable copy of a dataset with its licence and source recorded alongside it
• Explore an unfamiliar SQLite file in a browser instead of writing throwaway SQL scripts
• Give a dataset a read-only JSON API so other tools and notebooks can query it over HTTP
• Browse and spot-check a scraped or logged corpus before embedding, sampling or evaluating it for LLM work

// How Datasette compares

Datasette alongside other open-source data wrangling tools AI/TLDR tracks, ranked by GitHub stars.