fix: cohere models running

README.md

@@ -141,13 +141,15 @@ For more detailed examples and the explanation of other functionalities, refer t
 
 This plugin supports all currently available **Chat/Completion** and **Embeddings** models from GitHub Models. This plugin supports image input and multimodal models.
 
-Still in progress:
-1. Support for image output models
-
 ## API Reference
 
 You can find the full API reference in the [API Reference Documentation](https://xavidop.github.io/genkitx-github/)
 
+## Troubleshooting
+
+1. GPT o1-preview it is still in beta. It does not support system roles and the `temperature` and `topP` needs to be set to 1. See OpenAI annocement [here](https://openai.com/index/introducing-openai-o1-preview/)
+2. Cohere models only supports text output for now. Issue opened [here](https://github.com/orgs/community/discussions/142364).
+
 ## Contributing
 
 Want to contribute to the project? That's awesome! Head over to our [Contribution Guidelines](https://github.com/xavidop/genkitx-github/blob/main/CONTRIBUTING.md).

src/github_llms.ts

@@ -756,17 +756,19 @@ export function toGithubRequestBody(
       `${response_format} format is not supported for GPT models currently`,
     );
   }
+  const modelString = (request.config?.version || model.version || modelName) as string;
   const body = {
     body: {
       messages: githubMessages,
       tools: request.tools?.map(toGithubTool),
-      model: request.config?.version || model.version || modelName,
+      model: modelString,
       max_tokens: request.config?.maxOutputTokens,
       temperature: request.config?.temperature,
       top_p: request.config?.topP,
       n: request.candidates,
       stop: request.config?.stopSequences,
-      response_format: responseFormat,
+      // FIXME: coherence models don't support response_format for now
+      response_format: modelString.includes("cohere") ? "" : responseFormat,
     },
   } as any;