Is your feature request related to a problem? Please describe.
GoModel implements /v1/audio/speech and /v1/audio/transcriptions, but not the third standard OpenAI audio endpoint, /v1/audio/translations (translate spoken audio in any language into English text). Users relying on the full OpenAI audio surface currently have a gap here.
Confirmed missing: no route or handler exists for it anywhere in the codebase (zero matches for audio-translation-related routes), while the sibling transcription endpoint is fully implemented.
Describe the solution you'd like
Add /v1/audio/translations, accepting the same multipart audio upload shape as the existing /v1/audio/transcriptions endpoint, but always returning English text output (no language parameter accepted on input, matching OpenAI's own Whisper translations behavior).
Since this is nearly a clone of the already-implemented transcription request/response handling and backend call, reusing that code path directly (same multipart parsing, same Whisper-family provider backend, forced English output) should keep this small.
Describe alternatives you've considered
None - the shape is dictated by OpenAI's existing API and by GoModel's own existing transcription implementation, which this should mirror closely.
Additional context
This is a good first issue: the sibling /v1/audio/transcriptions handler and adapter are a direct template to copy from, so most of the request-parsing and provider-call plumbing already exists in the codebase.
Is your feature request related to a problem? Please describe.
GoModel implements
/v1/audio/speechand/v1/audio/transcriptions, but not the third standard OpenAI audio endpoint,/v1/audio/translations(translate spoken audio in any language into English text). Users relying on the full OpenAI audio surface currently have a gap here.Confirmed missing: no route or handler exists for it anywhere in the codebase (zero matches for audio-translation-related routes), while the sibling transcription endpoint is fully implemented.
Describe the solution you'd like
Add
/v1/audio/translations, accepting the same multipart audio upload shape as the existing/v1/audio/transcriptionsendpoint, but always returning English text output (nolanguageparameter accepted on input, matching OpenAI's own Whispertranslationsbehavior).Since this is nearly a clone of the already-implemented transcription request/response handling and backend call, reusing that code path directly (same multipart parsing, same Whisper-family provider backend, forced English output) should keep this small.
Describe alternatives you've considered
None - the shape is dictated by OpenAI's existing API and by GoModel's own existing transcription implementation, which this should mirror closely.
Additional context
This is a good first issue: the sibling
/v1/audio/transcriptionshandler and adapter are a direct template to copy from, so most of the request-parsing and provider-call plumbing already exists in the codebase.