[FEATURE]: enable responses API in openai compatible provider

[scratchmex]

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

We need an option for OpenAI compatible models like suggested in #5866 as well. I think it should be the default to use responses API for models that are intenfed to be used that way, like codex, same as proposed in above issue

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• [BUG]: GitHub Copilot Responses API routing for GPT-5+ #5866: [BUG]: GitHub Copilot Responses API routing for GPT-5+ - Already proposes adding a useResponsesApi option for OpenAI-compatible providers with similar intent
• [Feature Request] Support traditional Chat Completions API for OpenAI provider #7793: [Feature Request] Support traditional Chat Completions API for OpenAI provider - Related discussion on routing between Chat Completions and Responses APIs for OpenAI-compatible providers
• gpt-5.2-codex TypeError: sdk.responses is not a function. (In 'sdk.responses(modelID)', 'sdk.responses' is undefined) #8622: gpt-5.2-codex TypeError: sdk.responses is not a function - Related bug where responses API is expected but not available

Feel free to ignore if none of these address your specific case.

[Vonfry]

For responses api, you can use "@ai-sdk/openai" directly, IMO.

For example:

custom-provider = {
  npm = "@ai-sdk/openai";
  name = "OpenAI proxy";
    options = {
      apiKey = "{file:...}";
      baseURL = "https://.../v1";
    };
    models = {
      "gpt-5.2-codex" = {
         name = "GPT!";
        };
     };
 };

[DeluxeOwl]

The solution above doesn't show reasoning

[Vonfry]

Do you try following configurations:

options = {
  reasoningEffort = "xhigh";
  textVerbosity = "high";
  reasoningSummary = "detailed";
};