-
Notifications
You must be signed in to change notification settings - Fork 31
Open
Description
I developed a RAG web service with Azure Cognitive Services, Azure OpenAI API, Azure Cosmos Vector DB. I also built some of the basic blocks of Cosmos Graph DB, but wonder at what stage you integrate it with RAG? That is, does it make sense to do the following: (and if you can't answer, could you forward this to someone who knows):
- Generate Relationship (Node1->Edge->Node2) for the prompt text, resulting in a list of Relationships.
- For each Relationship and each node and edge: find the 1st-level related nodes and edges.
- For each found relationship concatenate nodes and the edge and create Embeddings out of it.
- Query the Cosmos vector DB and use the response for additional context for user prompt, along with the results of vector DB search (based on user prompt).
I might be wrong here, but I really would like to know at what stage you incorporate Graph DB querying into RAG.
Metadata
Metadata
Assignees
Labels
No labels