A local graph memory layer for constrained RAG
The author says they built turbo-graph on top of turbovec. It adds GraphMemoryIndex to compact local vector search so it can handle tenant filters, source, time, tag, and graph limits. The author says it is not meant to replace vector DBs and describes it as an Alpha experiment for local or private RAG routes.
Key points
- turbo-graph is a fork built from turbovec.
- It adds GraphMemoryIndex for constrained RAG search.
- The post mentions tenant filters, source, time, and tag constraints.
- The author also mentions BM25 candidates, rerank, and explainability as real bottlenecks.
- The author says it does not replace vector DBs and is still Alpha.
Quick term guide
- GraphMemoryIndex
- A search structure meant to use both relationships and rules when finding stored information.
- vector search
- A search method that finds text with similar meaning, not only the same words.
- vector DBs
- Databases made to store and search the number-based representations used in vector search.
- vector DB
- A database designed to store embeddings and quickly find the most similar ones when you search.
- AI agents
- AI agents are AI tools that can carry out steps toward a goal, not just answer once.
- AI agent
- An AI program that can inspect information and suggest what to do next.
- agent memory
- Stored knowledge or records an AI agent uses when answering or acting.
- bottleneck
- A point where work gets stuck because one person or step cannot handle the volume, slowing down everything else.