MongoDB Atlas Vector Search

Vectors are mathematical representations of data expressed as arrays of numbers that capture features or properties. In AI, embeddings encode complex data such as text, images, or audio into a format computers can process while retaining relationships and patterns.

Automated Embedding generates vector embeddings for your data directly inside MongoDB. When you define an autoEmbed field in a vector search index, MongoDB creates embeddings with a Voyage AI model at ingest, update, and query time, keeping vectors in sync without an external pipeline.

MongoDB Vector Search searches data based on semantic meaning captured in vectors, while MongoDB Search performs keyword search based on actual text and defined synonym mappings.

Yes. MongoDB Atlas is a fully managed, AI-ready data platform that includes text and vector search, letting you store operational data, metadata, and vector embeddings together. Vector Search is also available in MongoDB Enterprise Advanced and Community Edition.

It supports embeddings from any provider under the 4096-dimension limit. Automated Embedding supports the Voyage AI model family, and you can bring your own embeddings from any provider.

With Automated Embedding, inference, authentication, scaling, retries, and synchronization are managed by MongoDB, and models are billed per actual token usage. With bring-your-own embeddings, you handle billing, rate limits, and API key management.

KNN finds vectors near one another. ANN trades accuracy for performance and powers Atlas Vector Search using HNSW graphs. ENN returns exact closest vectors and can maintain sub-second latency for unfiltered queries up to 10,000 documents.

Yes. MongoDB Vector Search supports ingestion, indexing, and querying of scalar and binary quantized vectors, and provides options for automatic scalar and binary quantization of full-fidelity vectors.