Add Background & Queue Support for OpenAI

[ianhawes]

The OpenAI Responses API allows for background processing of messages. Currently, Prism filters out the provider options in Prism\Prism\Providers\OpenAI\Handlers\Text and Prism\Prism\Providers\OpenAI\Handlers\Structured to prevent background from being specified.

A queued response will look like this:

{
  "id": "resp_abc123xyz",
  "object": "response",
  "created_at": 1756644282,
  "status": "queued",
  "background": true,
  "error": null,
  "max_output_tokens": 4092,
  "model": "gpt-5-2025-08-07",
  "output": [],
  "store": true,
}

(some properties removed for brevity)

In order to support this setting, an additional check will need to be conducted in handle() to look for a status of queued and return the relevant response_id for future polling and retrieval.

Beyond supporting background in a request, it would be nice to also have support for:

• Confirming the authenticity of signed OpenAI webhooks
• Serializing and deserializing responses into Prism objects and vice-versa

[sixlive]

Let me stew on this for a few!

[RealFarway]

Subscribing for updates! Really interested in how this is going

[dcaswel]

We're looking to use Prism with a custom provider. When running speechToText on audio, that provider has a way to send them a webhook url and they return an ID that we store so we know what record to update when the webhook comes back. I know this isn't exactly the same thing as this issue but it would be nice if Prism had functionality to be able to handle things like asynchronous processing. This could include having the ability to put an ID in the Response objects and ways of getting things like a transcript from the provider.

[dcaswel]

FYI, I created a PR at #810 with an idea on how to implement an interface for asynchronous processes like what I was suggesting. I am open to suggestions on it.