Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Data references are no longer given in query output #77

Open
simonepstein opened this issue Oct 14, 2024 · 2 comments
Open

Data references are no longer given in query output #77

simonepstein opened this issue Oct 14, 2024 · 2 comments

Comments

@simonepstein
Copy link

Hi there,

I noticed that the data references are no longer included in the query response, this is since commit 8bc7522

I wondered if you could explain the reason for removing these references; I think the context specific evidence is useful for the user to review when evaluating the response from the llm.

@gusye1234
Copy link
Owner

Yeah, I implemented this feature at first.
But I found out that GraphRAG will attach the reference id, which is a unique identifier inside the storage. So for most of the users, that identifier is pretty meaningless because you have to look into the internals of nano-graphrag's storage to get the corresponding data.
I think the default version of nano-graphrag should output something basic, so I remove the data references. But I'm thinking adding the referred data as an output of query, so advanced users can pack their own answer styles.

@simonepstein
Copy link
Author

simonepstein commented Oct 18, 2024

I naively reverted the local_rag_response prompt to return the data references (i.e. "This is an example sentence supported by multiple data references [Data: (record ids); (record ids)]." ...) and then if I return the response with the context data, I can at least link from the response to the top level of data references that were used to form the response.

It might be useful to be able to dig into the graph further, for example to see sources that were used to form a community, but I imagine this could get confusing quite quickly. Maybe navigating the context data is more of a graph visualisation problem.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants