Media
4 items
Description
Grab tables from any page and query them instantly with SQL — in-browser DuckDB analytics, with optional page scanning. Turn the web into your data warehouse. datajoi reads the page you're looking at and organizes what it finds into well-formed, typed datasets you can query with real SQL — or a plain English question — in a chat-style side panel. Everything runs on an in-browser DuckDB engine: your data stays on your machine, and everything you collect can leave as standard formats any tool can read. FIND DATA, GET DATASETS • One click scans the page and separates real datasets (tables with proper header rows) from page furniture — you only import what's worth keeping • Values are cleaned and typed on import: "$1,820", "(1,234)" and "63%" become real numbers, dates become dates, and totals rows are kept out of your aggregates. The original text is preserved alongside, and per-column notes tell you exactly what couldn't be converted • Pages with no real tables still yield data: combine infoboxes, layout tables and page metadata into one facts dataset per site that grows as you browse • Data files linked on the page (CSV, Excel, JSON, Parquet) import in the same click; local files drag in too • The panel follows your browsing: revisit a site and its saved datasets are waiting; visit a page with new data and a banner offers to review it — nothing is ever imported without you ASK ANYTHING • Write SQL or ask in plain English — questions are translated to SQL locally, and the query that ran is always shown • Typos, smart quotes and unbalanced parens are repaired, not rejected • DataStory summaries with inline charts, computed locally from your results (deterministic — no AI involved) • Full DuckDB SQL with schema-aware autocomplete; results grid built for 100k+ rows; saved queries and history survive restarts; dark mode YOUR DATA, IN REAL FORMATS • Every dataset is kept locally as typed Parquet in a per-site warehouse, so your collection survives restarts and restores instantly • Download any dataset as CSV, JSONL, a runnable SQL script, or Parquet • Download a whole site as a warehouse bundle: Parquet files plus a catalog.sql that opens directly in the DuckDB CLI — no datajoi needed DATAJOI PRO (OPTIONAL) Pro adds hosted publishing — put a site's warehouse at a stable URL that anyone with DuckDB can query remotely, and unpublish any time — plus managed page scanning (no API key needed), AI answers for questions the built-in translator can't parse, and unlimited sources and saved queries. Free covers everything else. Sign in with Google, upgrade from Settings; payment runs on Intuit QuickBooks. PRIVATE BY DESIGN The DuckDB engine is compiled to WebAssembly and ships inside the extension. Your tables, files, and queries are processed on your machine and stored in your browser — SQL never runs on a server, and there are no analytics or telemetry. Site access is requested one site at a time, only when you use Find Data, and is revocable in Settings. Data leaves your machine only when you act: exporting a file, publishing a warehouse (Pro), enabling page scanning, or signing in. HOW TO USE 1. Pin datajoi and click its icon on any page. 2. Hit "Find Data on this page" and pick what to import. 3. Ask questions, run SQL, export or publish what you need.