init - DuckBrain

duckbrain init [--instance <name>] [--sources slack,imessage|none] [--no-sync]

Creates the instance at ~/.duckbrain/<instance>/ (default instance: default): the DuckLake catalog (SQLite, WAL), the lake/ Parquet directory, the bronze envelope tables, and a per-instance state.sql agent surface.

Source picker

Then shows an interactive multi-select (arrow keys, space to toggle) of the sources that need a setup step. Selecting one runs its setup — installing the *crawl binary (Homebrew for most; go install for imsgcrawl) — and prints next steps:

Source	Setup	Status
Slack	install `slacrawl`	available
Notion	install `notcrawl`	available
WhatsApp	install `wacrawl`	available
Telegram	install `telecrawl`	available
Discord	install `discrawl`	available
Granola	`connect granola` (Pipedream broker)	available
iMessage	install `imsgcrawl` (go), grant Full Disk Access	available
Apple Photos	install `photoscrawl` (go)	available
GitHub issues/PRs	install `gitcrawl`	available
Contacts	install `clawdex`	available

Non-interactive: --sources slack,imessage skips the picker and runs just those setups; --sources none opts out. The prompt auto-skips when stdin isn’t a TTY, so scripts and buttons stay non-interactive. Sources with no setup aren’t listed — there’s nothing to do for them:

Session sources (Claude Code, Codex, OpenClaw) ingest once you’ve used those tools.
markdown reads the docs/ folder of whatever directory you run duckbrain from.
Apps (Gmail, GitHub, …) connect separately with duckbrain connect <app>.

Syncing

Syncing turns on by default — init ends by running sync on: streaming session ingest, a 6h sync schedule, and Claude Code session hooks. Opt out with --no-sync (turn on later with duckbrain sync on).