[+] Source separation server

.env.example

@@ -1 +1,2 @@
 MONGO_URL="mongodb://cat:meow@localhost:27017/"
+AUDIO_SEPARATOR_API="http://127.0.0.1:8000"

.gitignore

@@ -26,3 +26,5 @@ data/
 public/assets/fonts/cache
 .idea
 static/audio
+temp_audio
+.stfolder

scripts/README.md

@@ -0,0 +1,15 @@
+# Scripts
+
+## Setup
+
+```bash
+mamba create -n audio-separator python=3.13
+mamba activate audio-separator
+pip install "audio-separator[cpu]" fastapi uvicorn python-multipart
+```
+
+## Usage
+
+```bash
+python server.py
+```

scripts/server.py

@@ -0,0 +1,102 @@
+import logging
+import shutil
+import uuid
+import threading
+import uvicorn
+from pathlib import Path
+from fastapi import FastAPI, UploadFile, File, BackgroundTasks, HTTPException
+from fastapi.responses import FileResponse
+from audio_separator.separator import Separator
+
+# Configure logging
+logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
+logger = logging.getLogger(__name__)
+
+app = FastAPI()
+
+# Configuration
+MODEL_NAME = 'model_bs_roformer_ep_317_sdr_12.9755.ckpt'
+TEMP_DIR = Path('temp_audio')
+TEMP_DIR.mkdir(exist_ok=True)
+
+# Global state
+separator = None
+separator_lock = threading.Lock()
+jobs = {}
+
+def get_separator():
+    global separator
+    if separator is None:
+        logger.info("Initializing Separator...")
+        separator = Separator()
+        logger.info(f"Loading model {MODEL_NAME}...")
+        separator.load_model(model_filename=MODEL_NAME)
+        logger.info("Model loaded.")
+    return separator
+
+def process_separation(task_id: str, task_dir: Path, input_path: Path):
+    try:
+        with separator_lock:
+            jobs[task_id]['status'] = 'processing'
+            
+            sep = get_separator()
+            sep.output_dir = str(task_dir)
+            
+            logger.info(f"Separating {task_id}...")
+            output_files = sep.separate(str(input_path))
+            
+            results = {}
+            for file in output_files:
+                p = Path(file)
+                if 'Vocals' in p.name:
+                    results['vocals'] = str(p)
+                elif 'Instrumental' in p.name:
+                    results['instrumental'] = str(p)
+            
+            jobs[task_id]['results'] = results
+            jobs[task_id]['status'] = 'completed'
+            logger.info(f"Completed {task_id}")
+
+    except Exception as e:
+        logger.error(f"Error processing {task_id}: {e}")
+        jobs[task_id]['status'] = 'error'
+        jobs[task_id]['error'] = str(e)
+
+@app.post("/separate")
+async def separate(file: UploadFile = File(...), background_tasks: BackgroundTasks):
+    task_id = str(uuid.uuid4())
+    task_dir = TEMP_DIR / task_id
+    task_dir.mkdir(parents=True, exist_ok=True)
+    
+    input_path = task_dir / "input.mp3"
+    with open(input_path, "wb") as buffer:
+        shutil.copyfileobj(file.file, buffer)
+        
+    jobs[task_id] = {"status": "pending"}
+    background_tasks.add_task(process_separation, task_id, task_dir, input_path)
+    return {"task_id": task_id}
+
+@app.get("/status/{task_id}")
+async def get_status(task_id: str):
+    return jobs.get(task_id, {"status": "not_found"})
+
+@app.get("/result/{task_id}/{stem}")
+async def get_result(task_id: str, stem: str):
+    job = jobs.get(task_id)
+    if not job:
+        raise HTTPException(status_code=404, detail="Job not found")
+    
+    if job['status'] != 'completed':
+        raise HTTPException(status_code=400, detail="Job not completed")
+        
+    results = job.get('results', {})
+    file_path = results.get(stem)
+    
+    if not file_path or not Path(file_path).exists():
+        raise HTTPException(status_code=404, detail=f"Stem {stem} not found")
+        
+    return FileResponse(file_path)
+
+if __name__ == "__main__":
+    print("Starting Audio Separator API on port 8000...")
+    uvicorn.run(app, host="127.0.0.1", port=8000)