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

[Integration] Request to support FAISSDocumentStore in 2.x #717

Open
v3ss0n opened this issue May 6, 2024 · 6 comments
Open

[Integration] Request to support FAISSDocumentStore in 2.x #717

v3ss0n opened this issue May 6, 2024 · 6 comments
Labels
contributions wanted! Looking for external contributions feature request Ideas to improve an integration

Comments

@v3ss0n
Copy link

v3ss0n commented May 6, 2024

Is your feature request related to a problem? Please describe.
We are having problem migrating current version to 2.0 due to no FAISSSupport
Describe the solution you'd like
Due to is simplicity , performance and developer experience , we and many other RAG solution provider are now using FAISS . So it would be nice to integrate it.

Describe alternatives you've considered
QDrant and many others are not approved by some organization yet because they are very new and not backed by big vendors.

@v3ss0n v3ss0n added the feature request Ideas to improve an integration label May 6, 2024
@masci
Copy link
Contributor

masci commented May 10, 2024

Hi @v3ss0n and thanks for the feature request.

Do you envision an exact porting of the FAISS document store from Haystack 1.x, with the accompanying persistent storage, or would an in-memory, pure-FAISS solution also work?

@v3ss0n
Copy link
Author

v3ss0n commented May 10, 2024

Exact porting is not necessary as long as we can use Fiass as DS , and yes of-course we are looking forward for both persistant and and in-memory DS but whatever work first.

@masci
Copy link
Contributor

masci commented May 10, 2024

I think having in-memory, basic support for FAISS can be done rather easily.
The problem with persistence is that FAISS alone can't manage metadata, so we need to pair it with another storage. We used sqlite in the previous version of Haystack and that turned out to be a mess, so this time it'd be better to pick a storage that's closer to the nature of FAISS - but this will require to prioritize this issue, and not sure this can happen soon.

@v3ss0n
Copy link
Author

v3ss0n commented May 10, 2024

yeah i agree , sqlite store was a mess and hard to manage. Lets start with in-memory.

@mlokhandwala
Copy link

Hi FAISS is more important then you might give it credit for, it is necessary for Haystack to move FAISS support to version 2. One of the most important attributes I like about FAISS That it uses local files and for document store, it can use SQL lite, This is a great advantage when we have small but many different isolated databases it may not be very scalable but there is a case to be made for small number of users requiring specialized RAG type applications.

@smach
Copy link

smach commented Sep 10, 2024

Would be very useful to have FAISS support in Haystack 2.0!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
contributions wanted! Looking for external contributions feature request Ideas to improve an integration
Projects
Development

No branches or pull requests

5 participants