epic: Better Design abstraction for Remote AI (and Engineering abstraction)

[freelerobot]

Problem

• It's not clear to users they need to add an API key to chat with remote models.
• It's not clear to users the diff btw remote vs local model.
• Right now it just looks like its broken.

Success Criteria

1. RightPanel should indicate this model is not set up yet. Let's try the label API key needed
2. We should somehow link the user to settings > models to set it up

Additional context Add any other context or screenshots about the feature request here.

[imtuyethan]

Design (Deprecated)

https://www.figma.com/file/ytn1nRZ17FUmJHTlhmZB9f/Jan-App?type=design&node-id=1190-33531&mode=design&t=hHvTFu6BxvZx3oQh-4

[imtuyethan]

Archive original comment from @0xSage Problem

• It's not clear to users they need to add an API key to chat with remote models.
• It's not clear to users the diff btw remote vs local model.
• Right now it just looks like its broken.

Success Criteria

1. RightPanel should indicate this model is not set up yet. Let's try the label API key needed
2. We should somehow link the user to settings > models to set it up

Additional context Add any other context or screenshots about the feature request here.

[imtuyethan]

For review: (in settings): https://www.figma.com/file/ytn1nRZ17FUmJHTlhmZB9f/Jan-App?type=design&node-id=2160-191190&mode=design&t=frUQsREQGVbBH5uL-4 (in threads): https://www.figma.com/file/ytn1nRZ17FUmJHTlhmZB9f/Jan-App?type=design&node-id=1190-33531&mode=design&t=frUQsREQGVbBH5uL-4

[kha84]

I looked through the UI mockups and I can see there's one thing is missing - to allow user to specify it's own URL, for OpenAI-compatible models. Like for instance I do have a model running on my server via oobabooga (it provides OpenAI-compatible API out of the box) or via ollama+LiteLLM. And I want to be able to chat with that model from Jan - for that I need to be able to add a remotely inferenced model with OpenAI-compatible API on a custom URL.

[imtuyethan]

I looked through the UI mockups and I can see there's one thing is missing - to allow user to specify it's own URL, for OpenAI-compatible models. Like for instance I do have a model running on my server via oobabooga (it provides OpenAI-compatible API out of the box) or via ollama+LiteLLM. And I want to be able to chat with that model from Jan - for that I need to be able to add a remotely inferenced model with OpenAI-compatible API on a custom URL.

Thanks for flagging this @kha84. I'll add this user story as well!

[dan-menlo]

I am sunsetting this issue in favor for a more holistic redesign of the "Provider" abstraction:

https://www.notion.so/jan-ai/Provider-Abstraction-for-Local-and-Remote-AI-d54448ad5ce34cb2845d986870b9395e?pvs=4