support cerebras.ai provider

[mikestaub]

I see here that Cerebras is a provider of this model, but I can't select it in roocode.

https://openrouter.ai/qwen/qwen3-32b/providers

Perhaps it is related to this?

#1919

I am blocked on using this provider, as we are now relying on openrouter exclusively for this provider.

#3247

[dosubot[bot]]

Cerebras.ai is not listed as a directly supported provider in Roo Code according to the official documentation. However, Roo Code does support any provider that offers an OpenAI-compatible API endpoint. If Cerebras.ai (via OpenRouter or otherwise) exposes such an endpoint, you may be able to use it by selecting "OpenAI Compatible" as the provider in Roo Code and entering the appropriate Base URL, API key, and model ID as described in the docs. The Base URL must be the endpoint provided by the service, not the official OpenAI endpoint. You should refer to Cerebras/OpenRouter documentation for the correct model names and endpoints to use in this configuration. Reference

I could not find any documentation about GitHub issue #1919 or pull request #3247, so I can't confirm their current status or whether they resolve your issue. If you are still unable to select Cerebras via OpenRouter in Roo Code, it may be due to a limitation or bug that is not yet documented.

To reply, just mention my name @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[mikestaub]

I was able to integrate the cerebras API by creating a local LiteLLM proxy, but the Qwen-3-32b model is simply not good enough to be productive. Once they support Deepseek R1 the speed will be incredible as I have success using deepseek R1 with roocode.

[deadcoder0904]

Would love this as its been added to Cline too - https://github.com/cline/cline/releases/tag/v3.17.6