Integrate chatgpt assistants / assistants api #3659
Replies: 3 comments 2 replies
-
|
Like you mentioned this would be appropriate as either a pipeline or a function, PR welcome! |
Beta Was this translation helpful? Give feedback.
-
|
I disagree - what they are doing is directly analogous to this project and would be a competing schema. This project aims to emulate via "pipelines" some of that same idea. Yes it takes some coding, as OpenAI's Assistants are a no-code alternative... But perhaps we can build a no-code solution here in a similar way. But this way it can be free, open-source and possibly self-hosted. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your replies. Are there any examples of getting open webui to work like openai assistants (with function calling knowledge)? Also, it would have been way easier if open webui directly supported adding open ai assistant just like adding another model. That way users can configure their openai assistant and use open webui as the UI for that assistant. |
Beta Was this translation helpful? Give feedback.
-
|
I see your point and I'm not against it at all! It actually would make it easy to use an OpenAI's Assistants. I just don't want Open WebUI to become little more than a wrapper for OpenAI. Instead I see it as a way to build your own system with similar capabilities - some coding required 😉 Personally, I'm working with LangGraph agents. These use "flow engineering" which is a new paradigm beyond "prompt I think Pipelines are what you're looking for. It's a separate Docker container that, once enabled, allows you to write your own AI model any way you want - function/tool calling, custom databases, external APIs, anything. It seems VERY new, yet under-documented and still being designed, to be honest. But it is exactly what I've been looking for. When you make a pipeline it appears to the end user as just another model they can choose from. Check Open WebUI's Pipelines GitHub for examples. Look for their Integrations folder. I feel it's what you're after. The OpenAI way of doing it is natural language only - no code. Pipelines seem way more powerful... some coding required. To be honest (and I haven't played around enough to know if they allow API access to their assistants) your pipeline code can just call their Assistants API perhaps. That may be what you're after.. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks so much for your detailed response. I will try pipelines |
Beta Was this translation helpful? Give feedback.
-
Describe the solution you'd like
I have multiple working chatgpt assistants that work well and has document search, function calling and all that. I would like to add the assistants id to open webui along with my openai api key. Then basically open webui can just behave like the UI.
Describe alternatives you've considered
Tried pipelines but it looks like I might have to write functions again? What other alternatives are there?
https://platform.openai.com/docs/assistants/overview
Beta Was this translation helpful? Give feedback.
All reactions