GitHubCore manages PR status for all visible worktrees. It batches GitHub API calls, caches results on disk, and throttles refreshes to avoid rate limits.
PRStatusCacheService stores PRStatus objects keyed by {worktree-path}:{commit-hash}.
Invalidation rule: a cache entry is valid only when the worktree's current HEAD commit matches the cached commit. A new commit or a merge automatically causes a cache miss on the next poll.
This means:
- Reading PRs is fast (disk hit) when code hasn't changed.
- No manual cache busting is needed; git history drives freshness.
GitHubCore tracks which worktrees are currently visible on screen (visibleWorktrees). Only those paths are polled on each refresh cycle (~30 s). When the user scrolls, setVisibleWorktrees() updates the set.
Batch fetches call gh pr list or gh pr view via GitHubService. Results are merged into the pullRequests map and written to the disk cache.
Rate-limit avoidance: the refresh interval is deliberately long to avoid GitHub API rate limits in multi-project setups with many worktrees. See src/cores/GitHubCore.ts for current interval constants.
See concepts/status-model.md for the full PRStatus definition.
GitHubContext exposes createPR() and mergePR(). Both call through to GitHubService which shells out to gh pr create / gh pr merge. After a successful operation the cache entry for the affected worktree is cleared and a refresh is triggered.