Made Some Changes

app.py

@@ -0,0 +1,149 @@
+import os
+import tempfile
+import subprocess
+from fastapi import FastAPI, File, UploadFile, HTTPException
+from fastapi.responses import JSONResponse
+import aiohttp
+import numpy as np
+from datetime import datetime
+import logging
+
+# Setup
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+app = FastAPI(title="Emotion Detection API", docs_url="/docs")
+
+# Config - get from environment
+HF_TOKEN = os.getenv("HF_TOKEN", "")
+API_TOKEN = os.getenv("API_TOKEN", "test123")
+
+# Models - using only 2 for reliability
+MODELS = {
+    "wav2vec2_english": {
+        "url": "https://api-inference.huggingface.co/models/ehcalabres/wav2vec2-lg-xlsr-en-speech-emotion-recognition",
+        "weight": 0.7,
+    },
+    "gigam_emo": {
+        "url": "https://api-inference.huggingface.co/models/salute-developers/GigaAM-emo",
+        "weight": 0.3,
+    }
+}
+
+# Emotion mapping
+EMOTION_MAPPING = {
+    "angry": ["angry", "ang"],
+    "happy": ["happy", "hap"],
+    "sad": ["sad"],
+    "fear": ["fear"],
+    "surprise": ["surprise"],
+    "disgust": ["disgust"],
+    "neutral": ["neutral", "neu"]
+}
+
+@app.get("/health")
+async def health():
+    return {"status": "ok", "hf_token": bool(HF_TOKEN)}
+
+@app.get("/")
+async def root():
+    return {
+        "message": "Emotion Detection API",
+        "docs": "/docs",
+        "endpoints": ["POST /analyze"]
+    }
+
+@app.post("/analyze")
+async def analyze(file: UploadFile = File(...)):
+    """Analyze emotion from audio file"""
+    
+    # Check auth header
+    auth = file.headers.get("authorization", "")
+    if not auth or auth.replace("Bearer ", "") != API_TOKEN:
+        return JSONResponse(
+            status_code=401,
+            content={"error": "Invalid or missing Authorization header"}
+        )
+    
+    # Save uploaded file
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
+        content = await file.read()
+        tmp.write(content)
+        input_path = tmp.name
+    
+    try:
+        # Convert to proper format
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as out:
+            output_path = out.name
+        
+        subprocess.run([
+            "ffmpeg", "-i", input_path,
+            "-ar", "16000", "-ac", "1",
+            "-y", output_path
+        ], check=True, capture_output=True)
+        
+        # Read converted file
+        with open(output_path, "rb") as f:
+            audio_bytes = f.read()
+        
+        # Query models
+        headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+        results = {}
+        
+        async with aiohttp.ClientSession() as session:
+            for name, config in MODELS.items():
+                try:
+                    async with session.post(
+                        config["url"],
+                        headers=headers,
+                        data=audio_bytes,
+                        timeout=10
+                    ) as resp:
+                        if resp.status == 200:
+                            results[name] = await resp.json()
+                except Exception as e:
+                    logger.warning(f"{name} failed: {e}")
+        
+        # Simple ensemble
+        emotion_scores = {}
+        total_weight = 0
+        
+        for name, predictions in results.items():
+            weight = MODELS[name]["weight"]
+            total_weight += weight
+            
+            for pred in predictions:
+                label = pred.get("label", "").lower()
+                score = pred.get("score", 0)
+                
+                # Map to standard emotions
+                for std_emo, variations in EMOTION_MAPPING.items():
+                    if any(v in label for v in variations):
+                        emotion_scores[std_emo] = emotion_scores.get(std_emo, 0) + score * weight
+                        break
+        
+        # Normalize
+        if total_weight > 0:
+            emotion_scores = {k: v/total_weight for k, v in emotion_scores.items()}
+        
+        # Get primary emotion
+        primary = max(emotion_scores.items(), key=lambda x: x[1]) if emotion_scores else ("unknown", 0)
+        
+        return {
+            "primary_emotion": primary[0],
+            "confidence": round(primary[1], 3),
+            "all_emotions": {k: round(v, 3) for k, v in emotion_scores.items()},
+            "models_used": list(results.keys())
+        }
+        
+    except Exception as e:
+        logger.error(f"Error: {e}")
+        return JSONResponse(status_code=500, content={"error": str(e)})
+    finally:
+        # Cleanup
+        for path in [input_path, output_path]:
+            if os.path.exists(path):
+                os.unlink(path)
+
+# For Hugging Face
+from fastapi.middleware.cors import CORSMiddleware
+app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])