mirror of
https://github.com/open-webui/open-webui.git
synced 2026-03-17 20:43:32 -05:00
feat: support openai embeddings for rag #277
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @peperunas on GitHub (Feb 8, 2024).
Is your feature request related to a problem? Please describe.
It would be great to have support for OpenAI's embedding models. As of now, to the best of my knowledge, only local embeddings are supported.
Link: https://platform.openai.com/docs/guides/embeddings
@justinh-rahb commented on GitHub (Feb 8, 2024):
Believe you're correct. I'd go further and suggest to support the OpenAI API's embedding endpoint spec in general so that other compatible sources could be used.
@tjbck commented on GitHub (Feb 8, 2024):
Hi, I'm not entirely sure why or where this is useful, as we don't utilize any embeddings api directly and only use the
chat completionapi which handles embedding automatically. Could you guide me through and elaborate more on its potential use cases?@justinh-rahb commented on GitHub (Feb 8, 2024):
@tjbck right now the RAG feature uses an internal embedding model in the container to convert text into vectors, we're proposing to be able to use API embeddings models as alternative, some of them work significantly better than the default
all-MiniLM-L6-v2used right now.@tjbck commented on GitHub (Feb 8, 2024):
@justinh-rahb makes sense if that's what OP meant, thanks for the explanation!
@bozo32 commented on GitHub (Mar 29, 2024):
will this permit choice of index?
https://docs.llamaindex.ai/en/stable/module_guides/indexing/index_guide/
not sure, but for data sources that are somewhat internally structured (e.g. academic papers), when iterating through data sources the tree structure may perform differently.
@thienedits commented on GitHub (May 18, 2024):
Is it me or does this openai embedding model not work correctly? In the Document settings, OpenAI settings don't seem to work.
@yuzukumo commented on GitHub (Sep 7, 2024):
u need to restart the server for this setting to take effect