using-agent-brain

To perform multi-mode document search, use the hybrid or multi mode to fuse BM25 keyword hits with vector semantic retrieval, maximizing recall across diverse content types in your indexed codebases and documentation.

Yes, you can trace code dependencies and function calls by running a graph-based query. Graph mode explores function calls, imports, and dependencies within a codebase to surface related code paths across large repositories.

Yes, multi-mode document search requires a running Agent Brain server with indexed documents and pluggable providers for embeddings and summarization to execute BM25, vector, hybrid, and graph queries.

The best way to locate authentication references is using BM25 keyword search for exact term hits in code and API references, or vector semantic search to uncover related concepts and relationships across the indexed documentation.

Use graph mode instead of vector search when you need to explore structural relationships like function calls, imports, and dependencies within a codebase, rather than just uncovering semantically related content across documents.