Summary
- Build knowledge-grounded LLM applications with vector databases, semantic search, and retrieval strategies.
- Supports six vector database options (Pinecone, Weaviate, Milvus, Chroma, Qdrant, pgvector) and six embedding models optimized for different use cases and providers Covers five advanced retrieval patterns: hybrid search combining dense and sparse retrieval, multi-query generation, contextual compression, parent document retrieval, and HyDE (hypothetical document embeddings) Includes four document chunking strategies (recursive character, token-based, semantic, markdown header) and metadata filtering, MMR diversity balancing, and cross-encoder reranking for optimization Provides complete LangGraph implementation examples with async retrieval and generation nodes, plus evaluation metrics for measuring retrieval precision, recall, answer relevance, and faithfulness