main.py

# main.py
from fastapi import FastAPI, WebSocket, WebSocketDisconnect
from inference import VideoProcessor
import threading
import queue
import asyncio
import json

app = FastAPI()

@app.websocket("/ws/inference")
async def websocket_endpoint(websocket: WebSocket):
    await websocket.accept()
    processor = VideoProcessor()
    result_queue = queue.Queue()
    
    try:
        # 1. Nhận link RTSP từ Laptop
        data = await websocket.receive_text()
        config = json.loads(data)
        rtsp_url = config.get("url")
        
        print(f"Received request for: {rtsp_url}")
        
        # 2. Chạy Model trong luồng riêng (Background Thread)
        # Để không chặn WebSocket
        process_thread = threading.Thread(
            target=processor.start_processing, 
            args=(rtsp_url, result_queue)
        )
        process_thread.start()

        # 3. Vòng lặp gửi kết quả về Laptop
        while True:
            # Kiểm tra queue xem có kết quả mới từ Model không
            try:
                # Non-blocking get
                result = result_queue.get_nowait()
                await websocket.send_json(result)
            except queue.Empty:
                # Nếu không có kết quả, ngủ xíu để đỡ tốn CPU
                await asyncio.sleep(0.1)
                
            # Kiểm tra nếu client ngắt kết nối thì dừng model
            if not process_thread.is_alive():
                await websocket.send_json({"status": "Processing finished"})
                break

    except WebSocketDisconnect:
        print("Client disconnected")
        processor.running = False # Ra lệnh dừng model
        process_thread.join()
    except Exception as e:
        print(f"Error: {e}")
        processor.running = False