Astro maintains a ChromaDB vector store that indexes markdowns and documents using local embeddings (BAAI/bge-small-en-v1.5). This enables semantic search across your knowledge base without external embedding API keys.

  • Local, CPU-based embeddings with no API key needed
  • Automatic indexing of content when it changes (where applicable)
  • Full rebuild via the Settings panel (POST /api/reindex)

HTTP API

Method Path Purpose
GET /api/search Semantic search over the vector store (query string parameters)
GET /api/stats Vector store statistics
POST /api/reindex Rebuild embeddings from existing data (Settings also triggers this)

MCP tools

Tool Role
search Semantic vector search across indexed content
get_stats Vector store statistics (same information as /api/stats)

Point MCP clients at /mcp/ on your Astro host. Other read/write tools on separate doc pages update content that is then reflected in the index after ingestion.