juzer09's picture
Upload 2 files
6af9be3 verified
#!/usr/bin/env python3
"""
Madverse Music API
AI Music Detection Service
"""
# Configure numba before any other imports to avoid caching issues
import os
os.environ.setdefault('NUMBA_DISABLE_JIT', '1')
os.environ.setdefault('NUMBA_CACHE_DIR', '/app/.cache/numba')
from fastapi import FastAPI, HTTPException, BackgroundTasks, Header, Depends
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
from pydantic import BaseModel, HttpUrl
import torch
import soundfile as sf
import scipy.signal
import tempfile
import requests
from pathlib import Path
import time
from typing import Optional, Annotated, List, Union
import uvicorn
import asyncio
from contextlib import asynccontextmanager
import socket
import numpy as np
# Global model variable
model = None
@asynccontextmanager
async def lifespan(app: FastAPI):
"""Application lifespan management"""
# Startup
global model
try:
from sonics import HFAudioClassifier
print("🔄 Loading Madverse Music AI model...")
# Set cache directory to a writable location
cache_dir = "/app/.cache" if os.path.exists("/app") else "./cache"
os.makedirs(cache_dir, exist_ok=True)
# Load model with explicit cache directory
model = HFAudioClassifier.from_pretrained(
"awsaf49/sonics-spectttra-alpha-120s",
cache_dir=cache_dir
)
model.eval()
print("✅ Model loaded successfully!")
except Exception as e:
print(f"❌ Failed to load model: {e}")
import traceback
traceback.print_exc()
raise
yield
# Shutdown
print("🔄 Shutting down...")
# Initialize FastAPI app with lifespan
app = FastAPI(
title="Madverse Music API",
description="AI-powered music detection API to identify AI-generated vs human-created music",
version="1.0.0",
docs_url="/",
redoc_url="/docs",
lifespan=lifespan
)
# API Key Configuration
API_KEY = os.getenv("MADVERSE_API_KEY", "madverse-music-api-key-2024") # Default key for demo
async def verify_api_key(x_api_key: Annotated[Union[str, None], Header()] = None):
"""Verify API key from header"""
if x_api_key is None:
raise HTTPException(
status_code=401,
detail="Missing API key. Please provide a valid X-API-Key header."
)
if x_api_key != API_KEY:
raise HTTPException(
status_code=401,
detail="Invalid API key. Please provide a valid X-API-Key header."
)
return x_api_key
class MusicAnalysisRequest(BaseModel):
urls: List[HttpUrl]
def check_api_key_first(request: MusicAnalysisRequest, x_api_key: Annotated[Union[str, None], Header()] = None):
"""Check API key before processing request"""
if x_api_key is None:
raise HTTPException(
status_code=401,
detail="Missing API key. Please provide a valid X-API-Key header."
)
if x_api_key != API_KEY:
raise HTTPException(
status_code=401,
detail="Invalid API key. Please provide a valid X-API-Key header."
)
return request
class FileAnalysisResult(BaseModel):
url: str
success: bool
classification: Optional[str] = None # "Real" or "Fake"
confidence: Optional[float] = None # 0.0 to 1.0
probability: Optional[float] = None # Raw sigmoid probability
raw_score: Optional[float] = None # Raw model output
duration: Optional[float] = None # Audio duration in seconds
message: str
processing_time: Optional[float] = None
error: Optional[str] = None
class MusicAnalysisResponse(BaseModel):
success: bool
total_files: int
successful_analyses: int
failed_analyses: int
results: List[FileAnalysisResult]
total_processing_time: float
message: str
class ErrorResponse(BaseModel):
success: bool
error: str
message: str
def cleanup_file(file_path: str):
"""Background task to cleanup temporary files"""
try:
if os.path.exists(file_path):
os.unlink(file_path)
except:
pass
def download_audio(url: str, max_size_mb: int = 100) -> str:
"""Download audio file from URL with size validation"""
try:
print(f"🔽 Downloading audio from: {url}")
# Check if URL is accessible
response = requests.head(str(url), timeout=10)
print(f"📊 Head response status: {response.status_code}")
# Check content size
content_length = response.headers.get('Content-Length')
if content_length:
size_mb = int(content_length) / (1024 * 1024)
print(f"📏 File size: {size_mb:.2f}MB")
if int(content_length) > max_size_mb * 1024 * 1024:
raise HTTPException(
status_code=413,
detail=f"File too large. Maximum size: {max_size_mb}MB"
)
# Download file
print("🔽 Starting download...")
response = requests.get(str(url), timeout=30, stream=True)
response.raise_for_status()
print(f"✅ Download response status: {response.status_code}")
# Create temporary file
with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as tmp_file:
downloaded_size = 0
for chunk in response.iter_content(chunk_size=8192):
downloaded_size += len(chunk)
if downloaded_size > max_size_mb * 1024 * 1024:
os.unlink(tmp_file.name)
raise HTTPException(
status_code=413,
detail=f"File too large. Maximum size: {max_size_mb}MB"
)
tmp_file.write(chunk)
print(f"💾 Downloaded {downloaded_size} bytes to {tmp_file.name}")
return tmp_file.name
except requests.exceptions.RequestException as e:
error_msg = f"Failed to download audio: {str(e)}"
print(f"❌ Download error: {error_msg}")
raise HTTPException(
status_code=400,
detail=error_msg
)
except Exception as e:
error_msg = f"Error downloading file: {str(e)}"
print(f"❌ Unexpected download error: {error_msg}")
raise HTTPException(
status_code=500,
detail=error_msg
)
def classify_audio(file_path: str) -> dict:
"""Classify audio file using the AI model"""
try:
print(f"🎵 Loading audio file: {file_path}")
# Check if file exists
if not os.path.exists(file_path):
raise ValueError(f"Audio file not found: {file_path}")
# Check file size
file_size = os.path.getsize(file_path)
print(f"📏 Audio file size: {file_size} bytes")
if file_size == 0:
raise ValueError("Audio file is empty")
# Load audio with soundfile
print("🔊 Loading audio with soundfile...")
audio, sr = sf.read(file_path)
print(f"🎼 Audio loaded: {len(audio)} samples at {sr}Hz, duration: {len(audio)/sr:.2f}s")
if len(audio) == 0:
raise ValueError("Audio file contains no audio data")
# Convert to mono if stereo
if audio.ndim > 1:
print("🔀 Converting stereo to mono...")
audio = np.mean(audio, axis=1)
# Resample to 16kHz if needed (model requirement)
target_sr = 16000
if sr != target_sr:
print(f"🔄 Resampling from {sr}Hz to {target_sr}Hz...")
# Calculate the number of samples after resampling
num_samples = int(len(audio) * target_sr / sr)
audio = scipy.signal.resample(audio, num_samples)
sr = target_sr
print(f"✅ Resampled: {len(audio)} samples at {sr}Hz")
# Convert to tensor and add batch dimension
print("🧮 Converting to tensor...")
audio_tensor = torch.FloatTensor(audio).unsqueeze(0)
print(f"📊 Tensor shape: {audio_tensor.shape}")
# Get prediction
print("🤖 Running model inference...")
with torch.no_grad():
output = model(audio_tensor)
print(f"📈 Model output: {output}")
# Convert logit to probability using sigmoid
prob = torch.sigmoid(output).item()
print(f"📊 Sigmoid probability: {prob}")
# Classify: prob < 0.5 = Real, prob >= 0.5 = Fake
if prob < 0.5:
classification = "Real"
confidence = (1 - prob) * 2 # Convert to 0-1 scale
else:
classification = "Fake"
confidence = (prob - 0.5) * 2 # Convert to 0-1 scale
result = {
"classification": classification,
"confidence": min(confidence, 1.0), # Cap at 1.0
"probability": prob,
"raw_score": output.item(),
"duration": len(audio) / sr
}
print(f"✅ Classification result: {result}")
return result
except Exception as e:
error_msg = f"Error analyzing audio: {str(e)}"
print(f"❌ Audio analysis error: {error_msg}")
import traceback
print(f"🔍 Traceback: {traceback.format_exc()}")
raise HTTPException(
status_code=500,
detail=error_msg
)
async def process_single_url(url: str) -> FileAnalysisResult:
"""Process a single URL and return result"""
start_time = time.time()
temp_file = None
try:
print(f"🚀 Processing URL: {url}")
# Download audio file
temp_file = download_audio(url)
print(f"✅ Download completed: {temp_file}")
# Classify audio
result = classify_audio(temp_file)
print(f"✅ Classification completed: {result}")
# Calculate processing time
processing_time = time.time() - start_time
# Prepare response
emoji = "🎤" if result["classification"] == "Real" else "🤖"
message = f'{emoji} Detected as {result["classification"].lower()} music'
return FileAnalysisResult(
url=str(url),
success=True,
classification=result["classification"],
confidence=result["confidence"],
probability=result["probability"],
raw_score=result["raw_score"],
duration=result["duration"],
message=message,
processing_time=processing_time
)
except Exception as e:
processing_time = time.time() - start_time
error_msg = str(e)
print(f"❌ Processing failed for {url}: {error_msg}")
import traceback
print(f"🔍 Full traceback: {traceback.format_exc()}")
return FileAnalysisResult(
url=str(url),
success=False,
message=f"❌ Failed to process: {error_msg}",
processing_time=processing_time,
error=error_msg
)
finally:
# Cleanup file in background
if temp_file:
try:
print(f"🧹 Cleaning up temporary file: {temp_file}")
os.unlink(temp_file)
except Exception as cleanup_error:
print(f"⚠️ Failed to cleanup {temp_file}: {cleanup_error}")
@app.post("/analyze", response_model=MusicAnalysisResponse)
async def analyze_music(
request: MusicAnalysisRequest = Depends(check_api_key_first)
):
"""
Analyze music from URL(s) to detect if it's AI-generated or human-created
- **urls**: Array of direct URLs to audio files (MP3, WAV, FLAC, M4A, OGG)
- Returns classification results for each file
- Processes files concurrently for better performance when multiple URLs provided
"""
start_time = time.time()
if not model:
raise HTTPException(
status_code=503,
detail="Model not loaded. Please try again later."
)
if len(request.urls) > 50: # Limit processing
raise HTTPException(
status_code=400,
detail="Too many URLs. Maximum 50 files per request."
)
if len(request.urls) == 0:
raise HTTPException(
status_code=400,
detail="At least one URL is required."
)
try:
# Process all URLs concurrently with limited concurrency
semaphore = asyncio.Semaphore(5) # Limit to 5 concurrent downloads
async def process_with_semaphore(url):
async with semaphore:
return await process_single_url(str(url))
# Create tasks for all URLs
tasks = [process_with_semaphore(url) for url in request.urls]
# Wait for all tasks to complete
results = await asyncio.gather(*tasks, return_exceptions=True)
# Process results and handle any exceptions
processed_results = []
successful_count = 0
failed_count = 0
for i, result in enumerate(results):
if isinstance(result, Exception):
# Handle exception case
processed_results.append(FileAnalysisResult(
url=str(request.urls[i]),
success=False,
message=f"❌ Processing failed: {str(result)}",
error=str(result)
))
failed_count += 1
else:
processed_results.append(result)
if result.success:
successful_count += 1
else:
failed_count += 1
# Calculate total processing time
total_processing_time = time.time() - start_time
# Prepare summary message
total_files = len(request.urls)
if total_files == 1:
# Single file message
if successful_count == 1:
message = processed_results[0].message
else:
message = processed_results[0].message
else:
# Multiple files message
if successful_count == total_files:
message = f"✅ Successfully analyzed all {total_files} files"
elif successful_count > 0:
message = f"⚠️ Analyzed {successful_count}/{total_files} files successfully"
else:
message = f"❌ Failed to analyze any files"
return MusicAnalysisResponse(
success=successful_count > 0,
total_files=total_files,
successful_analyses=successful_count,
failed_analyses=failed_count,
results=processed_results,
total_processing_time=total_processing_time,
message=message
)
except Exception as e:
raise HTTPException(
status_code=500,
detail=f"Internal server error during processing: {str(e)}"
)
@app.get("/health")
async def health_check():
"""Health check endpoint"""
return {
"status": "healthy",
"model_loaded": model is not None,
"service": "Madverse Music API"
}
@app.get("/info")
async def get_info():
"""Get API information"""
return {
"name": "Madverse Music API",
"version": "1.0.0",
"description": "AI-powered music detection to identify AI-generated vs human-created music",
"model": "SpecTTTra-α (120s)",
"accuracy": {
"f1_score": 0.97,
"sensitivity": 0.96,
"specificity": 0.99
},
"supported_formats": ["MP3", "WAV", "FLAC", "M4A", "OGG"],
"max_file_size": "100MB",
"max_duration": "120 seconds",
"authentication": {
"required": True,
"type": "API Key",
"header": "X-API-Key",
"example": "X-API-Key: your-api-key-here"
},
"usage": {
"curl_example": "curl -X POST 'http://localhost:8000/analyze' -H 'X-API-Key: your-api-key' -H 'Content-Type: application/json' -d '{\"url\":\"https://example.com/song.mp3\"}'"
}
}
def find_available_port(start_port: int = 8000, max_attempts: int = 10) -> int:
"""Find an available port starting from start_port"""
import random
import time
# Add some randomization to avoid race conditions
time.sleep(random.uniform(0.1, 0.5))
for port in range(start_port, start_port + max_attempts):
try:
# Try to bind to the port with proper error handling
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
s.bind(('0.0.0.0', port))
s.listen(1)
print(f"✅ Port {port} is available")
return port
except OSError as e:
print(f"❌ Port {port} is busy: {e}")
continue
# If no port found, raise an exception
raise RuntimeError(f"No available port found in range {start_port}-{start_port + max_attempts - 1}")
if __name__ == "__main__":
try:
# Check if we're in a Hugging Face environment
is_hf_space = os.getenv('SPACE_ID') is not None
hf_port = os.getenv('PORT') # HF Spaces sets this
if is_hf_space and hf_port:
# Use HF Spaces assigned port
port = int(hf_port)
print(f"🤗 Running in Hugging Face Spaces on port {port}")
elif is_hf_space:
print("🤗 Running in Hugging Face Spaces environment")
# Use standard HF Spaces port
port = 7860
else:
# Find an available port for local development
port = find_available_port(8000, 10)
print(f"🚀 Starting server on port {port}")
# For HF Spaces, don't use the retry logic as it might cause issues
if is_hf_space:
uvicorn.run(app, host="0.0.0.0", port=port)
else:
# Add retry logic for local development
max_retries = 3
for attempt in range(max_retries):
try:
uvicorn.run(app, host="0.0.0.0", port=port)
break # If successful, break out of retry loop
except OSError as e:
if "Address already in use" in str(e) and attempt < max_retries - 1:
print(f"⚠️ Port {port} became busy, trying next port...")
port = find_available_port(port + 1, 10)
print(f"🔄 Retrying on port {port}")
else:
raise
except RuntimeError as e:
print(f"❌ {e}")
print("💡 Suggestions:")
print(" 1. Wait a moment and try again (another instance might be shutting down)")
print(" 2. Manually specify a different port:")
print(" uvicorn app:app --host 0.0.0.0 --port 8001")
print(" 3. Check for running processes: ps aux | grep python")
except KeyboardInterrupt:
print("\n🛑 Server stopped by user")
except Exception as e:
print(f"❌ Failed to start server: {e}")
import traceback
traceback.print_exc()