Skip to main content
Status: Available Full GitHub issue and pull-request discussion — thread titles + bodies, and (optionally) every comment — via gitcrawl, a local-first crawler that fetches from the GitHub API into a local SQLite database. DuckBrain attaches that database read-only and maps it to bronze, same pattern as Slack.

​
Per-repo, manual refresh

gitcrawl crawls one repo at a time (gitcrawl sync owner/repo) and has no sync-all, so DuckBrain doesn’t auto-refresh it — you crawl the repos you care about, then duckbrain sync git ingests whatever’s in the archive. gitcrawl sync fetches open issues + PRs by default; add --state all to backfill closed ones, and --include-comments to pull the comment threads (without it, only titles + bodies land).

​
What gets ingested

  • Two record shapes. One bronze row per thread (issue or PR: title + body) and one per comment (joined to its thread for context). Both upsert in place on edit.
  • Bot & deleted comments are dropped — CI/dependabot noise and tombstoned comments never reach the brain.
  • Classification tracks repo privacy — threads from a public repo are classification: public; anything from a private repo (or a repo whose visibility can’t be read) stays confidential, so a private thread never leaks.

​
Facet mapping

Distinct from GitHub via Pipedream, which pulls issue titles + commit messages + repo READMEs through managed OAuth. gitcrawl is the one that gets the full issue/PR discussion, local-first.