Proposal: Migrate from chat/completions to the responses API in OpenAIProvider

[jeancodogno]

The OpenAIProvider currently uses the legacy chat/completions endpoint, which is being phased out in favor of the more capable responses API. Updating to /v1/responses would bring several benefits:

• Full access to modern OpenAI features (e.g., prompt caching, structured outputs, multimodal input).
• A unified and simplified request/response format.
• Improved long-term compatibility with new models and future API updates.
• Better performance and reduced complexity compared to the older chat-based API.

I propose updating the provider to use the responses endpoint by default, with optional support for chat-style formatting if needed.

[ilvalerione]

It's already there as a separate provider: https://docs.neuron-ai.dev/the-basics/ai-provider#openairesponses

[jeancodogno]

Oh, great — thanks for pointing that out!
I actually hadn’t noticed the separate provider for the Responses API. My mistake 😅

Good to know it’s already implemented. I’ll take a look at OpenAIResponses. Thanks for the clarification!