Support for OpenAI Audio / Transcription Endpoints

[mathetake]

Discussed in #1498

^{Originally posted by tanthcstt November 5, 2025}
Does Envoy Gateway Support for OpenAI Audio / Transcription Endpoints

[tanthcstt]

I’m working on it. I’d like to do it.

[tanthcstt]

I have no experience with Kubernetes; it was a little bit hard for me to deploy and test, but finally it works. I will create a PR soon. I’m trying to understand what I did in this repo first :) AI has helped me a lot.

[tanthcstt]

@mathetake
/v1/audio/transcriptions and /v1/audio/speech work with the OpenAI provider. Do I need to create a translator for Gemini audio understanding? It’s a little different, but both can be used for audio transcription. Gemini uses the generateContent endpoint with a prompt, while OpenAI has a separate endpoint (/audio/transcriptions).

[johnugeorge]

@tanthcstt You may start without translation(passthrough for Openai provider). Translators can be added later. Take a look at image generation api.

[tanthcstt]

Do all tests need to pass before creating a PR?
@johnugeorge