Feature Request: Support OpenAI speech-to-text interface /v1/audio/transcriptions

[candrews]

Prerequisites

• I am running the latest code. Mention the version if possible as well.
• I carefully followed the README.md.
• I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• I reviewed the Discussions, and have a new and useful enhancement to share.

Feature Description

Now that we have support for Gemma 4 STT (#21421), we should consider implementing OpenAI's explicit speech-to-text API. Documentation is here https://platform.openai.com/docs/guides/speech-to-text

Example of v1/audio/transcriptions

from openai import OpenAI
client = OpenAI()

audio_file= open("/path/to/file/audio.mp3", "rb")
transcription = client.audio.transcriptions.create(
    model="whisper-1", 
    file=audio_file
)

print(transcription.text)

The llama.cpp web interface already supports uploading audio for transcription but it's not the same API.

Supporting streaming transcriptions would be ideal, but non-streaming would be a great start.

Motivation

Implementing the /v1/audio/transcriptions API will allow a wide variety of existing clients to use llama.cpp for STT.

This enhancement would particularly benefit Home Assistant users. Currently, HA users need to run a separate STT model (such as whisper.cpp or parakeet) and LLM (such as Gemma), which requires more resources and complexity. With this enhancement, HA users could just run Gemma 4 E2B/E4B and use it for both STT and LLM purposes.

Possible Implementation

No response

[ngxson]

This enhancement would particularly benefit Home Assistant users.

Same motivation for my hassio setup, I'll implement this soon