Transcription-Only Mode in the Realtime API

[jhaeussler-tricat]

Feature Request

Is your feature request related to a problem? Please describe.

I'm unable to use transcription-only mode with the OpenAI realtime API via the package because only the realtime session type appears to be implemented. According to the OpenAI API documentation, transcription-only mode should be available by sending a session update with type: "transcription" and a corresponding configuration, but there is currently no built-in support for this. This limits my ability to use the API for scenarios where I only need transcriptions without additional real-time features.

Describe the solution you'd like

I would like the package to support transcription-only mode for the realtime API. Specifically, it should be possible to:

• Create a new session with type: "transcription" and the required configuration.
• Update an existing realtime session to switch to transcription-only mode by sending a session update request with type: "transcription".
• Have clear API methods or parameters for specifying whether a session is in realtime or transcription-only mode.

Describe alternatives you've considered

Currently, the only workaround is to manually craft and send HTTP session update requests outside of the package, which is cumbersome and loses the benefits of integration provided by the library.

Additional context

Relevant documentation: https://platform.openai.com/docs/guides/realtime-transcription This feature would help users who want transcription without TTS or other real-time features, reducing latency and potentially lowering costs.

[StephenHodgson]

Pretty sure this should already be possible in the latest release, but I'll double check.

[StephenHodgson]

@jhaeussler-tricat I looked into it but this is a pretty big breaking change I'll have to wait until 9.x to make