Problem
Every developer (and every CI run) must do a full scan from scratch. On a 500-file Python project this might take 10-30 seconds — annoying in CI, and means agents in fresh environments wait before they can answer anything.
Proposed Change
Add export/import for a compressed graph snapshot:
codelens export-snapshot .codelens/graph.db.zst # write compressed snapshot
codelens import-snapshot .codelens/graph.db.zst # import + incremental delta
- Format: SQLite VACUUM INTO -> zstd compressed (~8-13x ratio)
- On first run with snapshot present: import snapshot, then only scan files changed since snapshot
- Auto-create .gitattributes with merge=ours to avoid binary merge conflicts
Benefits
- CI agents skip full scan, get instant structural knowledge
- New contributors onboard faster
- Snapshot is an artifact of the CI pipeline, always reflects latest merge
Problem
Every developer (and every CI run) must do a full scan from scratch. On a 500-file Python project this might take 10-30 seconds — annoying in CI, and means agents in fresh environments wait before they can answer anything.
Proposed Change
Add export/import for a compressed graph snapshot:
Benefits