Context

What is a flat vector search index?

• A flat vector search index is a new indexingMethod option on MongoDB Vector Search indexes that lets you choose between hnsw (default) and flat per vector field in the index definition.
  https://docs.google.com/document/d/1PFVMMq6W7m2ofaPw6U4KRCf2o9JQyKz_xrFsjzuVLFY https://www.mongodb.com/docs/atlas/atlas-vector-search/vector-search-type/
• When indexingMethod: "flat" is used, MongoDB performs an exhaustive search over the (typically filtered) subset of vectors instead of traversing an HNSW graph, which is specifically optimized for multitenant workloads with many tenants but relatively few vectors per tenant (≈<10K) where queries are always filtered by tenant/namespace.

What AI Frameworks do we need to ensure support of this tool?

• LangChain MongoDB integration (MongoDBAtlasVectorSearch): INTPYTHON-285 already added generic kwargs support to create_vector_search_index; we should explicitly verify and document that indexingMethod="flat" can be passed through unchanged.

• • Potentially filled through pymongo-search-utils.

Acceptance Criteria

• API support in the LangChain MongoDB Integration with a reach out to the PM team for a potential GTM push