Notebook

Notebook is a format where SQL is split into cells, each runs independently, results stick around between runs. Like Jupyter, but no Python — just SQL and Markdown.

Why

The regular SQL editor is great for one-off queries. Notebook is for when you:

Prepare an ad-hoc report: 5–10 queries in the flow "load → clean → aggregate → export".
Want to keep an incident postmortem: "here's the slow query, here's the EXPLAIN, here's the fix".
Share an analysis with a teammate: one file with SQL + Markdown commentary.

Create

+ New tab → Notebook or from the tab dropdown: New Notebook.

An empty document opens with one empty cell.

Cells

Two types:

SQL cell

▶ Run cell button or ⌘+Enter — executes and shows the result below.
Run all — executes top to bottom in order.
The result persists between runs (saved with the notebook file).

Markdown cell

Text notes with markdown support: headings, lists, links, inline code.
Double-click to enter edit mode. Click outside to render.

Toggle between types via the icon in the cell's top-right (SQL ↔ Text).

Cell management

Icons left of each cell:

+ add a cell above / below
↑↓ move
× delete
▶▶ run from this cell to the bottom

Saving

File → Save notebook or ⌘+S. File extension — .ide99-nb (it's JSON: list of cells + last results + connection reference).

To open one — File → Open notebook or drag&drop the file onto the ide99 window.

What's in the file

A report.ide99-nb file is plain JSON, drop it into git:

{
  "version": 1,
  "title": "Q4 revenue analysis",
  "connection_hint": "db_prod",
  "cells": [
    { "type": "markdown", "body": "# Daily revenue\n\nLast 30 days:" },
    { "type": "sql", "body": "SELECT ... LIMIT 30;", "result": { "rows": 30, "duration_ms": 142 } }
  ]
}

connection_hint — the connection name at save time. When opening on another machine, ide99 looks for a connection with that name; if not found — it prompts you to pick one.

What's not saved

Result contents (just metadata: row count, duration). Safe to share via git — no data leakage.
Connection passwords (those live in keychain, not in the file).

If you want to export results with data — use Export on a specific cell (CSV, JSON, Excel).

Tip: on-call workflow

A template for incident postmortems:

Markdown cell: link to incident + date
SQL: SELECT * FROM events WHERE event_at BETWEEN ... ORDER BY ts;
SQL: EXPLAIN ANALYZE of the slow query
Markdown: root cause + fix
SQL: confirmation that things work after the fix

Save as 2026-05-22-incident-orders.ide99-nb, commit to your team's repo.

SQL editor — regular mode with one tab and a result panel.
Query history — global history, separate from notebooks.