Use vault fallback while Datacore initializes

[NathanSkene]

Summary

• check Datacore's explicit api.core.initialized state before using its query index
• use Obsidian's vault and metadata cache while Datacore is still initializing or when a query throws
• trust empty Datacore results after initialization, avoiding duplicate full-vault scans for legitimate no-match queries
• cover title search, compatible-node lookup, and node enumeration with focused regression tests

Reproduction

Datacore exposes its API during plugin loading, but initializes its index asynchronously on workspace.onLayoutReady. During a first install, reindex, or cache rebuild, Discourse Graph can therefore see a loaded Datacore API whose core.initialized flag is still false. Queries in that window can return an empty or partial universe even though matching discourse nodes are already available through Obsidian's metadata cache.

The previous distribution-repository PR described the index as potentially “empty or stale.” The reproduced case is narrower: Datacore is present but has not finished initialization. This change does not treat an empty result from an initialized index as stale.

Validation

• pnpm --filter @discourse-graphs/obsidian test — 5 tests pass
• pnpm --filter @discourse-graphs/obsidian check-types
• targeted ESLint on the changed source and tests
• Obsidian production build (tsx scripts/build.ts)
• git diff --check

Replaces DiscourseGraphs/discourse-graph-obsidian#9, which targeted the generated distribution repository.

---

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
discourse-graph	Skipped		Jul 28, 2026 9:41am

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 1 additional finding.