Source: supabase.com/blog/explorer-and-notebooks-investigations-your-team-can-rerun
Explorer lets users pick a source, write SQL against either their database or their logs, and see each result as a table or a chart. Supabase Assistant, the chat panel in the dashboard, can write a query from a plain-language request in the same editor, where it can be read and changed before it runs. Log queries carry their own time range, so queries from both sources can be compared side by side.
Notebooks combine queries, charts and markdown notes as cells in one document. A notebook saves its queries rather than their results, so running it again returns current data. The Assistant can write and run notebook cells and save them, but every create and edit waits for approval, and it can also change a single cell. Explorer works without the Assistant, and notebooks can be written entirely by hand.
Notebooks live in the project's supabase folder, so they can be versioned and reviewed in a pull request. Users can create one in the dashboard and pull it down with the CLI, or write one locally and push it up. Results are not stored in the file, so the same notebook works against any branch. The notebooks commands are available in the beta version of the CLI.
Why it matters if you build agents
Agents can run a saved notebook as part of their ongoing checks, such as repeating a database query after a deploy to see whether numbers moved as expected. Because notebooks are plain files in the repo, their queries can be reviewed like any other code. Developers using the Assistant should note that its notebook changes do not save until they are approved.
Get stories like this every weekday
One short email with the day's agent tooling news.