[FEATURE REQ] Add support for prompt_cache_key in OpenAIResponse and ResponseCreationOptions

[andregabrielmelo]

Describe the feature or improvement you are requesting

I would like to suggest adding prompt_cache_key as an option in the ResponseCreationOptions, to help optimize the model by using cached input. Also, the OpenAIResponse could have the prompt_cache_key used.

Additional context

https://platform.openai.com/docs/api-reference/responses/create#responses_create-prompt_cache_key
https://platform.openai.com/docs/guides/prompt-caching

[jsquire]

Hi @andregabrielmelo. Thanks for reaching out and for your request. The prompt_cache_key attribute appears to be recently added and does not yet exist in the REST API Spec from which the client is generated. Once an updated spec is available with that parameter, it will be included in the library. In the meantime, you can populate this by dropping down to the protocol level:

var client = new OpenAIResponseClient("gpt-4.1", key);

var rawResponse = await client.CreateResponseAsync(
    BinaryContent.Create(BinaryData.FromObjectAsJson(new
    {
        model = "gpt-4.1",
        prompt_cache_key = "some-value",
        instructions = "You are a personal math tutor who is also a pirate.  When asked a math question, respond like an educated pirate.",
        input = "I need to solve the equation 3x + 11 = 14. Can you help me?"
    })));

var response = ModelReaderWriter.Read<OpenAIResponse>(rawResponse.GetRawResponse().Content);
Console.WriteLine(response.GetOutputText());

[jsquire]

//cc: @trrwilson