Fix transcriptions endpoint

extensions/openai/script.py

@@ -38,6 +38,8 @@ from .typing import (
     CompletionResponse,
     DecodeRequest,
     DecodeResponse,
+    TranscriptionsRequest,
+    TranscriptionsResponse,
     EmbeddingsRequest,
     EmbeddingsResponse,
     EncodeRequest,
@@ -53,6 +55,8 @@ from .typing import (
     to_dict
 )
 
+from io import BytesIO
+
 params = {
     'embedding_device': 'cpu',
     'embedding_model': 'sentence-transformers/all-mpnet-base-v2',
@@ -176,12 +180,13 @@ def handle_billing_usage():
     return JSONResponse(content={"total_usage": 0})
 
 
-@app.post('/v1/audio/transcriptions', dependencies=check_key)
+@app.post('/v1/audio/transcriptions', response_model=TranscriptionsResponse, dependencies=check_key)
-async def handle_audio_transcription(request: Request):
+async def handle_audio_transcription(request: Request, request_data: TranscriptionsRequest = Depends(TranscriptionsRequest.as_form)):
     r = sr.Recognizer()
 
-    form = await request.form()
+    file = request_data.file
-    audio_file = await form["file"].read()
+    audio_file = await file.read()
+    audio_file = BytesIO(audio_file)
     audio_data = AudioSegment.from_file(audio_file)
 
     # Convert AudioSegment to raw data
@@ -189,8 +194,8 @@ async def handle_audio_transcription(request: Request):
 
     # Create AudioData object
     audio_data = sr.AudioData(raw_data, audio_data.frame_rate, audio_data.sample_width)
-    whisper_language = form.getvalue('language', None)
+    whisper_language = request_data.language
-    whisper_model = form.getvalue('model', 'tiny')  # Use the model from the form data if it exists, otherwise default to tiny
+    whisper_model = request_data.model  # Use the model from the form data if it exists, otherwise default to tiny
 
     transcription = {"text": ""}
 
@@ -200,10 +205,11 @@ async def handle_audio_transcription(request: Request):
             transcription["text"] = await run_in_executor(partial)
 
     except sr.UnknownValueError:
-        print("Whisper could not understand audio")
+        logger.warning("Whisper could not understand audio")
         transcription["text"] = "Whisper could not understand audio UnknownValueError"
+
     except sr.RequestError as e:
-        print("Could not request results from Whisper", e)
+        logger.warning("Could not request results from Whisper", e)
         transcription["text"] = "Whisper could not understand audio RequestError"
 
     return JSONResponse(content=transcription)

extensions/openai/typing.py

@@ -3,6 +3,7 @@ import time
 from typing import Dict, List
 
 from pydantic import BaseModel, Field
+from fastapi import UploadFile, Form
 
 
 class GenerationOptions(BaseModel):
@@ -128,6 +129,25 @@ class ChatPromptResponse(BaseModel):
     prompt: str
 
 
+class TranscriptionsRequest(BaseModel):
+    file: UploadFile
+    language: str | None = Field(default=None)
+    model: str = Field(default='tiny')
+
+    @classmethod
+    def as_form(
+        cls,
+        file: UploadFile = UploadFile(...),
+        language: str | None = Form(None),
+        model: str = Form('tiny'),
+    ) -> 'TranscriptionsRequest':
+        return cls(file=file, language=language, model=model)
+
+
+class TranscriptionsResponse(BaseModel):
+    text: str
+
+
 class EmbeddingsRequest(BaseModel):
     input: str | List[str] | List[int] | List[List[int]]
     model: str | None = Field(default=None, description="Unused parameter. To change the model, set the OPENEDAI_EMBEDDING_MODEL and OPENEDAI_EMBEDDING_DEVICE environment variables before starting the server.")