Unknown Context
Graph retrieval augmented generation (Graph RAG, https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/) is emerging as a powerful complement to traditional vector search methods.
Key points:
- Leverages structured data in graph databases (nodes and relationships)
- Enhances depth and context of retrieved information
- Complements existing vector search methods
This provides a unique opportunity for MongoDB to put forth our differentiator of being a developer data platform and being able to provide a graph DB support in addition to Vector DB support. With GraphRAG, the knowledge graph is shallow enough to be represented in MongoDB document model, and the graph triplet be lookuped via the $graphLookup syntax. We should explore implementing this approach to improve our information retrieval capabilities.
Next steps:
- Implement Graph RAG in LangChain (requested by Cisco)
- Implement Graph RAG support in LlamaIndex (requested by GE HC)
- Provide a proof-of-concept implementation
- Identify any performance/ resource limitations and potential challenges when using MongoDB for GraphRAG