chromadb-embeddings - SKILL.md Agent Skill

name: chromadb-embeddings description: "Store and query vector embeddings using ChromaDB at {{CHROMADB_HOST}}:{{CHROMADB_PORT}}." metadata: openclaw: emoji: "🎨"

ChromaDB Embeddings

ChromaDB is available at http://{{CHROMADB_HOST}}:{{CHROMADB_PORT}} within the Docker network.

Create a Collection

curl -X POST "http://{{CHROMADB_HOST}}:{{CHROMADB_PORT}}/api/v1/collections" \
  -H "Content-Type: application/json" \
  -d '{"name": "my_collection", "metadata": {"hnsw:space": "cosine"}}'

Add Documents

curl -X POST "http://{{CHROMADB_HOST}}:{{CHROMADB_PORT}}/api/v1/collections/{collection_id}/add" \
  -H "Content-Type: application/json" \
  -d '{"ids": ["doc1"], "documents": ["Hello world"], "metadatas": [{"source": "test"}]}'

Query Similar Documents

curl -X POST "http://{{CHROMADB_HOST}}:{{CHROMADB_PORT}}/api/v1/collections/{collection_id}/query" \
  -H "Content-Type: application/json" \
  -d '{"query_texts": ["greeting"], "n_results": 5}'

Tips for AI Agents

ChromaDB can auto-generate embeddings if configured with an embedding function.
Use metadata filters to narrow search results.
Prefer cosine distance for normalized embeddings.