MongoDB Atlas Vector Search

Full vector search requires M10+ clusters. The feature enables storing embeddings alongside document data with combined queries, filter by document fields and rank by vector similarity in single aggregation pipeline. Uses hierarchical navigable small world (HNSW) algorithm for approximate nearest neighbor search. Supports up to 4096 dimensions with float32 precision. Integration with LangChain, LlamaIndex, and major ML frameworks. For text-to-SQL contexts: natural fit for applications already using MongoDB that need semantic search, no separate vector database required. Key advantage is operational simplicity, single database for application data and AI features.