community: Fix retrieval node objects

[galshubeli]

• Description: This PR fixes conditions where FalkorDB queries contain FalkorDB node objects and converts them into a dictionary structure.
• Issue: When the retrieval contains FalkorDB objects, the Q&A step cannot handle this context as natural language.
• Dependencies: falkordb-py

If no one reviews your PR within a few days, please @-mention one of baskaryan, eyurtsev, ccurme, vbarda, hwchase17.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchain	⬜️ Ignored (Inspect)	Visit Preview		Mar 23, 2025 1:40pm

[ccurme]

Can you add a test that demonstrates what is being fixed? I assume the string representation for Node is nonsensical? This is a change in the interface so we need to demonstrate that this is a bug fix (i.e., the existing behavior is not useful).

[galshubeli]

Can you add a test that demonstrates what is being fixed? I assume the string representation for Node is nonsensical? This is a change in the interface so we need to demonstrate that this is a bug fix (i.e., the existing behavior is not useful).

You're right. This is a bug fix for the string representation of nodes. In other graph providers, it is implemented in the graph file (see the image below). I believe the correct place for this is when the user needs the query for context. Do you agree?

Could you explain how to add a test for this?

[ccurme]

@galshubeli For testing, we'd ideally add a unit test into a (new) file here. It should test what context is being supplied to a (fake) chat model. It looks like we return the context in intermediate steps, so you should be able to access that in the test.

The test should fail for the implementation that's currently on master, and demonstrate the impact of the change in this PR.

[ccurme]

@galshubeli we've now moved langchain-community into a standalone repo. Would you mind opening your PR there, along with a unit test?

https://github.com/langchain-ai/langchain-community