chromadb-expert

name	chromadb-expert
description	Expert guidance for vector database management and RAG workflows using ChromaDB. Use when implementing semantic search, document embeddings, or agentic memory.

ChromaDB & Vector Expert

Vector Operations

• Embeddings: Use high-quality models (e.g., OpenAI, HuggingFace) consistent across ingestion and query.
• Collections: Group related data into specific collections. Use descriptive names.
• Metadata Filtering: Use metadata heavily to scope searches (e.g., user_id, file_type).

RAG Patterns

• Chunking: Use RecursiveCharacterTextSplitter with appropriate chunk_size and overlap.
• Retrieval: Implement hybrid search if possible. Use top_k appropriately to balance context window vs relevance.
• Persistence: Ensure ChromaDB is initialized with a persistent client (PersistentClient).