[BUG] New feature for Post Process does not propose any provider

[lstep]

Bug Description

When first time activated the Post Process in the debug screen, The Provider field for Post Process does not provided any option to choose from ("no options found"). Had to stop, and then restart the application to have the providers.

System Information

App Version: 0.6.0 (debian package verison)

Operating System:
Ubuntu 24.04.3 LTS

[rodalpho]

Also would like Gemini API as well.

[SmetDenis]

Hello. @cjpais

I tried specifying my LiteLLM host (OpenAI Compatible), but it didn't work. I also tried specifying Ollama host (it proxies requests to openai/litellm), but that didn't work either.

It would be useful to be able to change the OpenAI host to a custom one.

Thank you in advance.

[lstep]

Hello. @cjpais

I tried specifying my LiteLLM host (OpenAI Compatible), but it didn't work. I also tried specifying Ollama host (it proxies requests to openai/litellm), but that didn't work either.

It would be useful to be able to change the OpenAI host to a custom one.

Thank you in advance.

You can change the OpenAI host to a custom one, just choose "Custom" in the drop down menu, that's what I'm doing (but connecting to a llama.cpp server), it works fine

[SmetDenis]

You can change the OpenAI host to a custom one, just choose "Custom" in the drop down menu, that's what I'm doing (but connecting to a llama.cpp server), it works fine

Yes, thank you. That's the first thing I tried, but for some reason, LiteLLM isn't working, even though I use it in many applications. It doesn't see the list of models, even if I explicitly specify the model name.

[cjpais]

I see you wrote https, I'm not familiar with litellm, but definitely if you are hosting it locally it's http and not https. I've connected without issue to LMStudio

In regards to the issue itself, this is probably a similar issue as reported in #96. It's likely how the settings code generically works, this is an issue and needs a fix

[SmetDenis]

@cjpais hello.
Thank you for response.

Can I hope that this will ever work with https? Because, it seems to me, it's quite logical, especially since it has the same API as OpenAI. In other words, it would be great to have the ability to change the domain for OpenAI, super simple changes.

Alternative option is to use the Ollama protocol, but it is completely different. Not like OpenAI or OpenRouter.

[cjpais]

It supports https, but your local provider may not. Im guessing you have a configuration issue somewhere as my local providers work without issue. This is a debug feature and comes with little support. It is intended for the most advanced users. If you want to debug, I recommend building from source and submitting a PR to fix.

Let's keep this thread to the original issue which was opened.