LLM Proxy

Appearance

Create translation ​

Translate an uploaded audio file into English. Multipart form data.

POST/v1/audio/translations

multipart/form-data. Streaming is not supported — the pipeline always runs non-streaming even if a stream field is sent.

Request
bash
curl http://localhost:8080/v1/audio/translations \
  -H "Authorization: Bearer $LLM_PROXY_KEY" \
  -F file=@recording-fr.mp3 \
  -F model=whisper-model \
  -F response_format=json
Request — OpenAI SDK
python
from openai import OpenAI

client = OpenAI(base_url="http://localhost:8080/v1", api_key="sk-your-proxy-key")
text = client.audio.translations.create(
    model="whisper-model",
    file=open("recording-fr.mp3", "rb"),
)
print(text.text)
Response
json
{"text": "Hello from the proxy!"}

Request fields ​

file (required), model (required), prompt, response_format (json default), temperature. No language and no timestamp_granularities fields.

Response formats ​

  • json → {"text"} (default)
  • text / srt / vtt → plain text
  • verbose_json → adds language, duration, segments

Provider support ​

openai and the openai-compatible family (except qwen/qwen-intl, which reject the route) — see the capability matrix.

Request
bash
curl http://localhost:8080/v1/audio/translations \
  -H "Authorization: Bearer $LLM_PROXY_KEY" \
  -F file=@recording-fr.mp3 \
  -F model=whisper-model \
  -F response_format=json
Request — OpenAI SDK
python
from openai import OpenAI

client = OpenAI(base_url="http://localhost:8080/v1", api_key="sk-your-proxy-key")
text = client.audio.translations.create(
    model="whisper-model",
    file=open("recording-fr.mp3", "rb"),
)
print(text.text)
Response
json
{"text": "Hello from the proxy!"}

MIT licensed

LLM Proxy