Hybrid Storage & Vector Indexes

AnamDB uses Lance as its core transactional and analytical storage format, bringing vector search and columnar storage directly to the SQL + Datalog query kernel.

---

Lance Storage Features

• Arrow-native: Native compatibility with Apache Arrow memory formatting, allowing zero-copy data exchanges during execution phases.
• Vector Search: Fast Approximate Nearest Neighbor (ANN) index support built directly into table storage formats.
• Columnar Layout: Efficient analytical query speeds, ensuring fast projections and aggregates during Datalog evaluations.
• Versioned History: Instant snapshots and time-travel capabilities for tracking table history data.

---

Combining Structured & Vector Data

Because storage is handled natively via Lance, SQL queries can perform vector similarity search alongside standard relational filters and Datalog constraints:

SELECT id, title, vector_distance(embedding, [0.1, 0.2, ...]) AS dist
FROM articles
WHERE dist < 0.5
  AND status = 'published';