Advanced: Performance

GrafitoDB performance depends on SQLite configuration, index usage, and query patterns. This page summarizes practical tips.

General Tips

• Use transactions for bulk inserts (db.begin_transaction() / db.commit()).
• Keep properties lean; avoid large blobs in node properties.
• Prefer targeted labels and relationship types to reduce scan size.

Indexing

• Create property indexes for frequently filtered properties.
• Use full-text search for text-heavy filtering instead of CONTAINS.
• Use vector indexes for semantic search at scale.

Text Search (FTS5)

• Ensure your SQLite build supports FTS5.
• Rebuild the FTS index after bulk updates if needed.

Vector Indexing

• Choose an ANN backend appropriate for your dataset size.
• Persist vector indexes on disk for faster startups.
• Rebuild indexes after large batch updates.

Query Planning

• Filter early (labels/types/properties) and avoid wide graph traversals.
• Keep shortest-path searches bounded where possible.

Operational Notes

• For large workloads, use a file-backed database instead of :memory:.
• Consider WAL mode in SQLite for better concurrent reads.