agentdb-vector-search

Enables fast, semantic document retrieval with 150x-12,500x speed boost over traditional methods.

• Implements semantic vector search using HNSW indexing and quantization
• Supports hybrid search combining vector similarity and metadata filtering
• Offers 150x-12,500x faster operations than traditional systems
• Integrates with RAG pipelines for context-aware querying

• HNSW indexing with sub-millisecond search (<100µs)
• 150x-12,500x faster operations than traditional solutions
• Supports multiple distance metrics: cosine, euclidean, dot product
• Offers quantization options: binary (32x), scalar (4x), product (8-16x)
• Hybrid search combining vector similarity and metadata filters
• MCP server integration for Claude Code and other agents
• Batch operations for efficient storage and retrieval
• Automatic embedding support with OpenAI or custom models

1. 1.Initialize vector database with `npx agentdb@latest init`
2. 2.Store documents using `storeWithEmbedding` or `insertPattern`
3. 3.Query with `query` CLI or `retrieveWithReasoning` API
4. 4.Use `hybridSearch` to combine vector similarity and metadata filters
5. 5.Integrate with RAG pipelines for context-aware generation
6. 6.Enable quantization (binary, scalar, product) for memory efficiency