I wrote a library specifically to support that: OpenAI.Responses — OpenAI.Responses v0.8.4 (but the downside, of course, is that unlike Mike’s library, you can’t really switch providers). You might try it while you are waiting for ReqLLM to add this functionality.
One of the reasons I did that (as opposed to just using existing solutions like LangChain) is that if you only need to work with OpenAI, Responses API is superior: cheaper on average as tokens caching is more efficient, provides additional functionality such as web search, background requests, automatic conversation building via previous_response_id, support for multiple developer messages, etc.
But the downside is that it’s hard to make it fit into the “Chat Completions” world view that every other provider (following the original OpenAI’s API) is supporting.
