Dot Product Pro is a self-hosted vector search platform. Ingest data, build indexes, and serve similarity queries — on your hardware, in your network, with no external dependencies.
Send vectors or raw text via HTTP. Dot Product Pro handles embedding generation or accepts pre-computed vectors directly. JSON or binary wire format.
Builds and maintains an HNSW index on disk. Persistent across restarts. Incremental updates without full rebuilds. Configurable for memory vs. recall tradeoff.
Query by vector or text. Returns ranked results with scores. Supports metadata filtering. Latency in the single-digit milliseconds on modest hardware.
One executable. No runtime dependencies. Ships with everything it needs.
Designed for environments where network connectivity is intermittent or absent. Works air-gapped.
Low memory footprint. Runs on ARM. Tested on devices with constrained resources.
Standard REST interface. Any language, any stack. OpenAPI spec included.
Indexes survive restarts. Local disk or volume mount. No external database required.
In-browser retrieval over local embeddings. Ships as a standalone module.
# run locally docker run -p 8080:8080 -v ./data:/data dotproduct/server # ingest curl -X POST localhost:8080/vectors -d '{"id":"doc_1","vector":[0.12,0.87,...],"meta":{"src":"readme.md"}}' # query curl -X POST localhost:8080/query -d '{"vector":[0.11,0.85,...],"top_k":10}'
One instance. Full functionality. No expiry.
Unlimited vectors, multi-index, priority support.
Multi-node, replication, and access controls.