Create app.py

app.py

@@ -0,0 +1,144 @@
+# ======================================================
+# HCL AI VOICE DETECTION API
+# Hugging Face Spaces (FastAPI)
+# ======================================================
+
+import base64
+import io
+import logging
+import librosa
+import torch
+
+from fastapi import FastAPI, HTTPException, Depends, Security
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.security.api_key import APIKeyHeader
+from pydantic import BaseModel
+
+from transformers import AutoFeatureExtractor, AutoModelForAudioClassification
+
+# ======================================================
+# CONFIGURATION
+# ======================================================
+API_KEY_NAME = "access_token"
+API_KEY_VALUE = "HCL_SECURE_KEY_2026"
+
+MODEL_ID = "facebook/wav2vec2-base-960h"
+TARGET_SR = 16000
+
+# ======================================================
+# LOGGING
+# ======================================================
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger("voice-detection")
+
+# ======================================================
+# DEVICE & MODEL LOADING (RUNS ON STARTUP)
+# ======================================================
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+logger.info(f"Using device: {DEVICE}")
+
+feature_extractor = AutoFeatureExtractor.from_pretrained(MODEL_ID)
+model = AutoModelForAudioClassification.from_pretrained(
+    MODEL_ID,
+    num_labels=2
+).to(DEVICE)
+
+model.eval()
+logger.info("Model loaded successfully")
+
+# ======================================================
+# FASTAPI APP
+# ======================================================
+app = FastAPI(
+    title="HCL AI Voice Detection API",
+    version="1.0.0"
+)
+
+api_key_header = APIKeyHeader(name=API_KEY_NAME, auto_error=False)
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# ======================================================
+# SCHEMAS
+# ======================================================
+class AudioRequest(BaseModel):
+    audio_base64: str
+
+
+class PredictionResponse(BaseModel):
+    classification: str
+    confidence_score: float
+
+
+# ======================================================
+# SECURITY
+# ======================================================
+async def verify_api_key(api_key: str = Security(api_key_header)):
+    if api_key != API_KEY_VALUE:
+        raise HTTPException(status_code=403, detail="Invalid API Key")
+    return api_key
+
+
+# ======================================================
+# CORE LOGIC
+# ======================================================
+def decode_audio(b64_audio: str) -> bytes:
+    try:
+        return base64.b64decode(b64_audio.split(",")[-1])
+    except Exception:
+        raise HTTPException(status_code=400, detail="Invalid Base64 audio")
+
+
+def analyze_voice(audio_bytes: bytes) -> tuple[str, float]:
+    audio, _ = librosa.load(
+        io.BytesIO(audio_bytes),
+        sr=TARGET_SR,
+        mono=True
+    )
+
+    inputs = feature_extractor(
+        audio,
+        sampling_rate=TARGET_SR,
+        return_tensors="pt"
+    )
+
+    inputs = {k: v.to(DEVICE) for k, v in inputs.items()}
+
+    with torch.inference_mode():
+        logits = model(**inputs).logits
+        probs = torch.softmax(logits, dim=-1)
+
+    confidence, prediction = torch.max(probs, dim=-1)
+    label = "AI_GENERATED" if prediction.item() == 1 else "HUMAN"
+
+    return label, round(confidence.item(), 4)
+
+
+# ======================================================
+# ENDPOINTS
+# ======================================================
+@app.get("/health")
+def health():
+    return {"status": "ok", "device": DEVICE}
+
+
+@app.post(
+    "/predict",
+    response_model=PredictionResponse
+)
+async def predict(
+    request: AudioRequest,
+    _: str = Depends(verify_api_key)
+):
+    audio_bytes = decode_audio(request.audio_base64)
+    label, score = analyze_voice(audio_bytes)
+
+    return {
+        "classification": label,
+        "confidence_score": score
+    }