Firestore vectorstore

[juanfe88]

Hello,

I'm exploring the implementation of a RAG system using Firestore as the vector store. Langchain's official documentation directed me to this repository within the googleapis organization: https://github.com/googleapis/langchain-google-firestore-python.

However, I've noticed this repository has seen little activity this year. Would it be possible for its functionality to be absorbed into this main repository? This would help maintain a unified source and interface for Langchain's Google Cloud integrations.

[lkuligin]

are there any specific features you're missing in that stand-alone library?

[juanfe88]

Yes for example, the retrieval with distance scores is not yet implemented and I have submitted a PR to do so. Also in an extended version of the firestore vectorestore I am using in a private repo, I have been playing with the implementation of the async methods (eg. asimiliarity_search_by_vector_with_score) using the async client from firesotre and I was planning on also submit a PR to add these implementations to the package

[kurtisvg]

Looks like the PR was reviewed and has gotten some attention.

I'm not sure it makes sense to move the Firestore integrations into this repo unless @lkuligin is also interested in taking responsibility for maintaining them as well. From a technical perspective, it also seems best to keep them in separate packages and avoid dependency bloat.

[juanfe88]

Yup, it recently got reviewed, but the development process overall seems a bit stale in that repo (last commit on main was from 3 months ago). However, I do understand that adding another object that uses yet another GCP service could be quite a hastle to maintain from a dependency point of view. Nonetheless, it would still be great if the addition of Firestore support in this repo could help the development move forward