how to use gemini instead of class OpenAILLMProvider LLMProvider?

[jjstory24]

how to use gemini instead of class OpenAILLMProvider LLMProvider?

[sshh12]

Hey! you'll need to implement an instance of "LLMProvider" and then add it to LLM_PROVIDERS.

[dchoward9]

@jjstory24 if you were successful, could you kindly share your edits? I have tried to update the files to use both deepseek and gemini and were unsuccessful.

[sshh12]

Definitely considering adding deepseek to have an open-source backend thats supported. How do folks typically use these (like the API)? OpenRouter?

[dchoward9]

@sshh12 if you are able to do so and have the time, please do :)

I was able to use deepseek-chat with openrouter but each message and response took at least 10 minutes.

I'm not sure which all models support tool use / function calling, if that matters.

[sshh12]

oh darn deepseek-ai/DeepSeek-R1#9 yeah deepseek doesn't have function calling, will have to think about this a bit more

[gabelul]

would be great to be able to use custom openai compatible endpoints and to be able to use a model from there

[sshh12]

Makes sense, will add a config for a custom base url

[sshh12]

Added