Vector + SQL
Vector search inside SQL. No glue code.
HNSW vector indexing with cosine, L2, and dot metrics — queryable directly from DataFusion-backed SQL. ML functions embedded. Arrow Flight streaming. 100M+ rows/sec per node.
Capabilities
What it does.
Every capability is production-grade. No flags, no betas — these ship the day you adopt Minds.
HNSW indices
Configurable M and ef_construction. Persistent. Filtered search supported.
3 distance metrics
Cosine · L2 · dot product. Per-index choice.
ANN inside SQL
`SELECT * FROM docs WHERE EMBEDDING <-> :q < 0.3 ORDER BY EMBEDDING <-> :q LIMIT 10`.
DataFusion
Cost-based + rule-based optimization. Vectorized execution.
Arrow Flight
Stream results between nodes at 100M+ rows/sec.
Parametrized everything
SQL injection structurally impossible — all paths prepared.
API
A few lines is all it takes.
The SDK reflects the conceptual model directly. No glue code, no orchestrator, no learning curve beyond the data shape.
exampletypescript
-- Vector + SQL together. Filter, then ANN, then aggregate.
SELECT
doc.title,
doc.published_at,
embedding <-> :query_vec AS distance
FROM resources doc
WHERE doc.tags @> ARRAY['finance']
AND doc.published_at >= '2026-01-01'
ORDER BY embedding <-> :query_vec
LIMIT 10;Specs
What it costs, what it guarantees.
Performance
- Vector search · 10K docs
- <10ms
- Analytics throughput
- 100M+ rows/sec per node
- Index types
- HNSW (configurable M, ef_construction)
- SQL engine
- DataFusion + Arrow
- Streaming
- Arrow Flight gRPC
Guarantees
- Durability
- fsync per WAL commit
- Isolation
- MVCC · snapshot
- Audit
- BLAKE3 Merkle chain
- Encryption
- AES-256-GCM per-namespace
- Concurrency
- 100K+ ops/sec
Build it on Minds.
Start with the SDK. Ship in an afternoon. Run anywhere — Akasha Cloud, on-prem, or air-gapped.