Update app.py

app.py

@@ -419,12 +419,13 @@ async def chat(
 # ----------------------------
 # Speech-to-Text (STT) Endpoint
 # ----------------------------
-from openai import OpenAI
+from fastapi import UploadFile, File, Form
+from fastapi.responses import JSONResponse
+import azure.cognitiveservices.speech as speechsdk
 import tempfile
 import shutil
 import os
-
-openai_client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
+import traceback
 
 @app.post("/api/stt")
 async def speech_to_text(
@@ -432,42 +433,72 @@ async def speech_to_text(
     session_id: str = Form(...)
 ):
     """
-    Speech-to-Text using OpenAI Whisper (whisper-1)
-    Uses OPENAI_API_KEY from environment
+    Speech-to-Text using Azure Cognitive Services
     """
     try:
-        if not os.getenv("OPENAI_API_KEY"):
+        speech_key = os.getenv("AZURE_SPEECH_KEY")
+        speech_region = os.getenv("AZURE_SPEECH_REGION")
+
+        if not speech_key or not speech_region:
             return JSONResponse(
-                {"success": False, "message": "OPENAI_API_KEY not set"},
+                {"success": False, "message": "Azure Speech credentials not set"},
                 status_code=500
             )
 
-        # Save audio temporarily
+        # Save uploaded audio temporarily
         with tempfile.NamedTemporaryFile(delete=False, suffix=".webm") as tmp:
             audio_path = tmp.name
             shutil.copyfileobj(audio.file, tmp)
 
         try:
-            with open(audio_path, "rb") as audio_file:
-                transcript = openai_client.audio.transcriptions.create(
-                    file=audio_file,
-                    model="whisper-1",
-                )
+            # Azure Speech config
+            speech_config = speechsdk.SpeechConfig(
+                subscription=speech_key,
+                region=speech_region
+            )
+
+            speech_config.speech_recognition_language = "en-US"
+
+            # Azure prefers WAV, but WEBM works if codec is supported
+            audio_input = speechsdk.audio.AudioConfig(filename=audio_path)
 
-            text = transcript.text.strip()
+            recognizer = speechsdk.SpeechRecognizer(
+                speech_config=speech_config,
+                audio_config=audio_input
+            )
+
+            result = recognizer.recognize_once()
+
+            if result.reason == speechsdk.ResultReason.RecognizedSpeech:
+                text = result.text.strip()
+
+                if not text:
+                    return JSONResponse(
+                        {"success": False, "message": "No speech detected"},
+                        status_code=400
+                    )
 
-            if not text:
+                return {
+                    "success": True,
+                    "text": text,
+                    "result": text,
+                    "transcription": text
+                }
+
+            elif result.reason == speechsdk.ResultReason.NoMatch:
                 return JSONResponse(
-                    {"success": False, "message": "No speech detected"},
+                    {"success": False, "message": "No speech could be recognized"},
                     status_code=400
                 )
 
-            return {
-                "success": True,
-                "text": text,
-                "result": text,
-                "transcription": text
-            }
+            else:
+                return JSONResponse(
+                    {
+                        "success": False,
+                        "message": f"Azure STT failed: {result.reason}"
+                    },
+                    status_code=500
+                )
 
         finally:
             if os.path.exists(audio_path):