python-detector-worker/app.py

from typing import Any, Dict
import os
import json
import time
import queue
import torch
import cv2
import numpy as np
import base64
import logging
import threading
import requests
import asyncio
import psutil
import zipfile
import ssl
import urllib3
import subprocess
import tempfile
import redis
from urllib.parse import urlparse
from requests.adapters import HTTPAdapter
from urllib3.util.ssl_ import create_urllib3_context
from fastapi import FastAPI, WebSocket, HTTPException
from fastapi.websockets import WebSocketDisconnect
from fastapi.responses import Response
from websockets.exceptions import ConnectionClosedError
from ultralytics import YOLO

# Import shared pipeline functions
from siwatsystem.pympta import load_pipeline_from_zip, run_pipeline, cleanup_camera_stability, cleanup_pipeline_node
from siwatsystem.model_registry import get_registry_status, cleanup_registry
from siwatsystem.mpta_manager import get_or_download_mpta, release_mpta, get_mpta_manager_status, cleanup_mpta_manager

app = FastAPI()

# Global dictionaries to keep track of models and streams
# "models" now holds a nested dict: { camera_id: { modelId: model_tree } }
models: Dict[str, Dict[str, Any]] = {}
streams: Dict[str, Dict[str, Any]] = {}
# Store session IDs per display
session_ids: Dict[str, int] = {}
# Track shared camera streams by camera URL
camera_streams: Dict[str, Dict[str, Any]] = {}
# Map subscriptions to their camera URL
subscription_to_camera: Dict[str, str] = {}
# Store latest frames for REST API access (separate from processing buffer)
latest_frames: Dict[str, Any] = {}
# Store cached detection dict after successful pipeline completion
cached_detections: Dict[str, Dict[str, Any]] = {}
# Enhanced caching system for LPR integration
session_detections: Dict[str, Dict[str, Any]] = {}  # session_id -> detection data
session_to_camera: Dict[str, str] = {}              # session_id -> camera_id
detection_timestamps: Dict[str, float] = {}         # session_id -> timestamp (for cleanup)
# Track frame skipping for pipeline buffer after detection
frame_skip_flags: Dict[str, bool] = {}
# Track camera connection states for immediate error handling
camera_states: Dict[str, Dict[str, Any]] = {}
# Track session ID states and pipeline modes per camera
session_pipeline_states: Dict[str, Dict[str, Any]] = {}
# Store full pipeline results for caching
cached_full_pipeline_results: Dict[str, Dict[str, Any]] = {}

with open("config.json", "r") as f:
    config = json.load(f)

poll_interval = config.get("poll_interval_ms", 100)
reconnect_interval = config.get("reconnect_interval_sec", 5)
TARGET_FPS = config.get("target_fps", 10)
poll_interval = 1000 / TARGET_FPS
logging.info(f"Poll interval: {poll_interval}ms")
max_streams = config.get("max_streams", 5)
max_retries = config.get("max_retries", 3)

# Configure logging
logging.basicConfig(
    level=logging.INFO,  # Set to INFO level for less verbose output
    format="%(asctime)s [%(levelname)s] %(name)s: %(message)s",
    handlers=[
        logging.FileHandler("detector_worker.log"),  # Write logs to a file
        logging.StreamHandler()  # Also output to console
    ]
)

# Create a logger specifically for this application
logger = logging.getLogger("detector_worker")
logger.setLevel(logging.DEBUG)  # Set app-specific logger to DEBUG level

# Create WebSocket communication logger
ws_logger = logging.getLogger("websocket_comm")
ws_logger.setLevel(logging.INFO)
ws_handler = logging.FileHandler("websocket_comm.log", encoding='utf-8')
ws_formatter = logging.Formatter("%(asctime)s [%(levelname)s] %(message)s")
ws_handler.setFormatter(ws_formatter)
ws_logger.addHandler(ws_handler)
ws_logger.propagate = False  # Don't propagate to root logger

# Ensure all other libraries (including root) use at least INFO level
logging.getLogger().setLevel(logging.INFO)

logger.info("Starting detector worker application")
logger.info(f"Configuration: Target FPS: {TARGET_FPS}, Max streams: {max_streams}, Max retries: {max_retries}")
ws_logger.info("WebSocket communication logging started - TX/RX format")
logger.info("WebSocket communication will be logged to websocket_comm.log")

# Ensure the models directory exists
os.makedirs("models", exist_ok=True)
logger.info("Ensured models directory exists")

# Constants for heartbeat and timeouts
HEARTBEAT_INTERVAL = 2  # seconds

# Global Redis connection for LPR integration
redis_client_global = None
lpr_listener_thread = None
cleanup_timer_thread = None
lpr_integration_started = False

# Redis connection helper functions
def get_redis_config_from_model(camera_id: str) -> Dict[str, Any]:
    """Extract Redis configuration from loaded model pipeline"""
    try:
        for model_id, model_tree in models.get(camera_id, {}).items():
            if hasattr(model_tree, 'get') and 'redis_client' in model_tree:
                # Extract config from existing Redis client
                client = model_tree['redis_client']
                if client:
                    return {
                        'host': client.connection_pool.connection_kwargs['host'],
                        'port': client.connection_pool.connection_kwargs['port'],
                        'password': client.connection_pool.connection_kwargs.get('password'),
                        'db': client.connection_pool.connection_kwargs.get('db', 0)
                    }
    except Exception as e:
        logger.debug(f"Could not extract Redis config from model: {e}")
    
    # Fallback - try to read from pipeline.json directly
    try:
        pipeline_dirs = []
        models_dir = "models"
        if os.path.exists(models_dir):
            for root, dirs, files in os.walk(models_dir):
                if "pipeline.json" in files:
                    with open(os.path.join(root, "pipeline.json"), 'r') as f:
                        config = json.load(f)
                        if 'redis' in config:
                            return config['redis']
    except Exception as e:
        logger.debug(f"Could not read Redis config from pipeline.json: {e}")
    
    return None

def create_redis_connection() -> redis.Redis:
    """Create Redis connection using config from pipeline"""
    global redis_client_global
    
    if redis_client_global is not None:
        try:
            redis_client_global.ping()
            return redis_client_global
        except:
            redis_client_global = None
    
    # Find any camera with a loaded model to get Redis config
    redis_config = None
    for camera_id in models.keys():
        redis_config = get_redis_config_from_model(camera_id)
        if redis_config:
            break
    
    if not redis_config:
        logger.error("No Redis configuration found in any loaded models")
        return None
    
    try:
        redis_client_global = redis.Redis(
            host=redis_config['host'],
            port=redis_config['port'],
            password=redis_config.get('password'),
            db=redis_config.get('db', 0),
            decode_responses=True,
            socket_connect_timeout=5,
            socket_timeout=5
        )
        redis_client_global.ping()
        logger.info(f"✅ Connected to Redis for LPR at {redis_config['host']}:{redis_config['port']}")
        return redis_client_global
    except Exception as e:
        logger.error(f"❌ Failed to connect to Redis for LPR: {e}")
        redis_client_global = None
        return None

# LPR Integration Functions
def process_license_result(lpr_data: Dict[str, Any]):
    """Process incoming LPR result and update backend"""
    try:
        # Enhanced debugging for LPR data reception
        logger.info("=" * 60)
        logger.info("🚗 LPR SERVICE DATA RECEIVED")
        logger.info("=" * 60)
        logger.info(f"📥 Raw LPR data: {json.dumps(lpr_data, indent=2)}")
        
        session_id = str(lpr_data.get('session_id', ''))
        license_text = lpr_data.get('license_character', '')
        
        logger.info(f"🔍 Extracted session_id: '{session_id}'")
        logger.info(f"🔍 Extracted license_character: '{license_text}'")
        logger.info(f"📊 Current cached sessions count: {len(session_detections)}")
        logger.info(f"📊 Available session IDs: {list(session_detections.keys())}")
        
        # Find cached detection by session_id
        if session_id not in session_detections:
            logger.warning("❌ LPR SESSION ID NOT FOUND!")
            logger.warning(f"   Looking for session_id: '{session_id}'")
            logger.warning(f"   Available sessions: {list(session_detections.keys())}")
            logger.warning(f"   Session count: {len(session_detections)}")
            
            # Additional debugging - show session timestamps
            if session_detections:
                logger.warning("📅 Available session details:")
                for sid, timestamp in detection_timestamps.items():
                    age = time.time() - timestamp
                    camera = session_to_camera.get(sid, 'unknown')
                    logger.warning(f"   Session {sid}: camera={camera}, age={age:.1f}s")
            else:
                logger.warning("   No cached sessions available - worker may not have processed any detections yet")
            
            logger.warning("💡 Possible causes:")
            logger.warning("   1. Session expired (TTL: 10 minutes)")
            logger.warning("   2. Session ID mismatch between detection and LPR service")
            logger.warning("   3. Detection was not cached (no sessionId from backend)")
            logger.warning("   4. Worker restarted after detection but before LPR result")
            return
        
        # Get the original detection data
        detection_data = session_detections[session_id].copy()
        camera_id = session_to_camera.get(session_id, 'unknown')
        
        logger.info("✅ LPR SESSION FOUND!")
        logger.info(f"   📹 Camera ID: {camera_id}")
        logger.info(f"   ⏰ Session age: {time.time() - detection_timestamps.get(session_id, 0):.1f} seconds")
        
        # Show original detection structure before update
        original_license = detection_data.get('data', {}).get('detection', {}).get('licensePlateText')
        logger.info(f"   🔍 Original licensePlateText: {original_license}")
        logger.info(f"   🆕 New licensePlateText: '{license_text}'")
        
        # Update licensePlateText in detection
        if 'data' in detection_data and 'detection' in detection_data['data']:
            detection_data['data']['detection']['licensePlateText'] = license_text
            
            logger.info("🎯 LICENSE PLATE UPDATE SUCCESS!")
            logger.info(f"   ✅ Updated detection for session {session_id}")
            logger.info(f"   ✅ Set licensePlateText = '{license_text}'")
            
            # Show full detection structure after update
            detection_dict = detection_data['data']['detection']
            logger.info("📋 Updated detection dictionary:")
            logger.info(f"   carModel: {detection_dict.get('carModel')}")
            logger.info(f"   carBrand: {detection_dict.get('carBrand')}")
            logger.info(f"   bodyType: {detection_dict.get('bodyType')}")
            logger.info(f"   licensePlateText: {detection_dict.get('licensePlateText')} ← UPDATED")
            logger.info(f"   licensePlateConfidence: {detection_dict.get('licensePlateConfidence')}")
        else:
            logger.error("❌ INVALID DETECTION DATA STRUCTURE!")
            logger.error(f"   Session {session_id} has malformed detection data")
            logger.error(f"   Detection data keys: {list(detection_data.keys())}")
            if 'data' in detection_data:
                logger.error(f"   Data keys: {list(detection_data['data'].keys())}")
            return
        
        # Update timestamp to indicate this is an LPR update
        detection_data['timestamp'] = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
        
        # Update all caches with new data
        session_detections[session_id] = detection_data.copy()
        cached_detections[camera_id] = detection_data.copy()
        
        # CRITICAL: Also update the pipeline state cached detection dict (used by lightweight mode)
        if camera_id in session_pipeline_states:
            pipeline_state = session_pipeline_states[camera_id]
            current_cached_dict = pipeline_state.get("cached_detection_dict", {})
            
            # Update the pipeline cached detection dict with new license plate
            updated_dict = current_cached_dict.copy() if current_cached_dict else {}
            updated_dict['licensePlateText'] = license_text
            
            pipeline_state["cached_detection_dict"] = updated_dict
            logger.info(f"✅ LPR: Updated pipeline state cached_detection_dict for camera {camera_id}")
            logger.debug(f"🔍 Pipeline cached dict now: {updated_dict}")
        else:
            logger.warning(f"⚠️ Camera {camera_id} not found in session_pipeline_states - pipeline cache not updated")
        
        logger.info("📡 SENDING UPDATED DETECTION TO BACKEND")
        logger.info(f"   📹 Camera ID: {camera_id}")
        logger.info(f"   📨 Updated licensePlateText: '{license_text}'")
        logger.info("   🔄 Updated both cache systems:")
        logger.info(f"     1️⃣ cached_detections[{camera_id}] ✅")
        logger.info(f"     2️⃣ session_pipeline_states[{camera_id}].cached_detection_dict ✅")
        
        # Log the full message being sent
        logger.info("📋 Updated detection data in cache:")
        logger.info(json.dumps(detection_data, indent=2))
        
        logger.info("✅ ALL CACHES UPDATED!")
        logger.info(f"   🎯 Lightweight mode will now use updated licensePlateText")
        logger.info(f"   📤 Backend will receive: licensePlateText = '{license_text}'")
        logger.info("   🔄 Both cache systems synchronized with LPR data")
        
        logger.info("=" * 60)
        logger.info("🏁 LPR PROCESSING COMPLETE")
        logger.info(f"   Session: {session_id}")
        logger.info(f"   License: '{license_text}'")
        logger.info(f"   Status: ✅ SUCCESS - DETECTION CACHE UPDATED")
        logger.info("=" * 60)
            
    except Exception as e:
        logger.error("=" * 60)
        logger.error("❌ LPR PROCESSING FAILED")
        logger.error("=" * 60)
        logger.error(f"Error: {e}")
        import traceback
        logger.error(f"Traceback: {traceback.format_exc()}")
        logger.error("=" * 60)

# LPR integration now uses cached detection mechanism instead of direct WebSocket sending

def license_results_listener():
    """Background thread to listen for LPR results from Redis"""
    logger.info("🎧 Starting LPR listener thread...")
    
    while True:
        try:
            redis_client = create_redis_connection()
            if not redis_client:
                logger.error("❌ No Redis connection available for LPR listener")
                time.sleep(10)
                continue
            
            pubsub = redis_client.pubsub()
            pubsub.subscribe("license_results")
            logger.info("✅ LPR listener subscribed to 'license_results' channel")
            
            for message in pubsub.listen():
                try:
                    if message['type'] == 'message':
                        logger.info("🔔 REDIS MESSAGE RECEIVED!")
                        logger.info(f"   📡 Channel: {message['channel']}")
                        logger.info(f"   📥 Raw data: {message['data']}")
                        logger.info(f"   📏 Data size: {len(str(message['data']))} bytes")
                        
                        try:
                            lpr_data = json.loads(message['data'])
                            logger.info("✅ JSON parsing successful")
                            logger.info("🏁 Starting LPR processing...")
                            process_license_result(lpr_data)
                            logger.info("✅ LPR processing completed")
                        except json.JSONDecodeError as e:
                            logger.error("❌ JSON PARSING FAILED!")
                            logger.error(f"   Error: {e}")
                            logger.error(f"   Raw data: {message['data']}")
                            logger.error(f"   Data type: {type(message['data'])}")
                        except Exception as e:
                            logger.error("❌ LPR PROCESSING ERROR!")
                            logger.error(f"   Error: {e}")
                            import traceback
                            logger.error(f"   Traceback: {traceback.format_exc()}")
                    elif message['type'] == 'subscribe':
                        logger.info(f"📡 LPR listener subscribed to channel: {message['channel']}")
                        logger.info("🎧 Ready to receive license plate results...")
                    elif message['type'] == 'unsubscribe':
                        logger.warning(f"📡 LPR listener unsubscribed from channel: {message['channel']}")
                    else:
                        logger.debug(f"📡 Redis message type: {message['type']}")
                        
                except Exception as e:
                    logger.error(f"❌ Error in LPR message processing loop: {e}")
                    break
                    
        except redis.exceptions.ConnectionError as e:
            logger.error(f"❌ Redis connection lost in LPR listener: {e}")
            time.sleep(5)  # Wait before reconnecting
        except Exception as e:
            logger.error(f"❌ Unexpected error in LPR listener: {e}")
            time.sleep(10)
    
    logger.warning("🛑 LPR listener thread stopped")

def cleanup_expired_sessions():
    """Remove sessions older than TTL (10 minutes)"""
    try:
        current_time = time.time()
        ttl_seconds = 600  # 10 minutes
        
        expired_sessions = [
            session_id for session_id, timestamp in detection_timestamps.items()
            if current_time - timestamp > ttl_seconds
        ]
        
        if expired_sessions:
            logger.info(f"🧹 Cleaning up {len(expired_sessions)} expired sessions")
            
            for session_id in expired_sessions:
                session_detections.pop(session_id, None)
                camera_id = session_to_camera.pop(session_id, None)
                detection_timestamps.pop(session_id, None)
                logger.debug(f"Cleaned up expired session: {session_id} (camera: {camera_id})")
                
        else:
            logger.debug(f"🧹 No expired sessions to clean up ({len(detection_timestamps)} active)")
            
    except Exception as e:
        logger.error(f"❌ Error in session cleanup: {e}")

def cleanup_timer():
    """Background thread for periodic session cleanup"""
    logger.info("⏰ Starting session cleanup timer thread...")
    
    while True:
        try:
            time.sleep(120)  # Run cleanup every 2 minutes
            cleanup_expired_sessions()
        except Exception as e:
            logger.error(f"❌ Error in cleanup timer: {e}")
            time.sleep(120)

def start_lpr_integration():
    """Start LPR integration threads"""
    global lpr_listener_thread, cleanup_timer_thread
    
    # Start LPR listener thread
    lpr_listener_thread = threading.Thread(target=license_results_listener, daemon=True, name="LPR-Listener")
    lpr_listener_thread.start()
    logger.info("✅ LPR listener thread started")
    
    # Start cleanup timer thread  
    cleanup_timer_thread = threading.Thread(target=cleanup_timer, daemon=True, name="Session-Cleanup")
    cleanup_timer_thread.start()
    logger.info("✅ Session cleanup timer thread started")

WORKER_TIMEOUT_MS = 10000
logger.debug(f"Heartbeat interval set to {HEARTBEAT_INTERVAL} seconds")

# Locks for thread-safe operations
streams_lock = threading.Lock()
models_lock = threading.Lock()
logger.debug("Initialized thread locks")


# Add helper to fetch snapshot image from HTTP/HTTPS URL
def fetch_snapshot(url: str):
    try:
        from requests.auth import HTTPBasicAuth, HTTPDigestAuth
        import requests.adapters
        import urllib3
        
        # Parse URL to extract credentials
        parsed = urlparse(url)
        
        # Prepare headers - some cameras require User-Agent and specific headers
        headers = {
            'User-Agent': 'Mozilla/5.0 (compatible; DetectorWorker/1.0)',
            'Accept': 'image/jpeg,image/*,*/*',
            'Connection': 'close',
            'Cache-Control': 'no-cache'
        }
        
        # Create a session with custom adapter for better connection handling
        session = requests.Session()
        adapter = requests.adapters.HTTPAdapter(
            pool_connections=1,
            pool_maxsize=1,
            max_retries=urllib3.util.retry.Retry(
                total=2,
                backoff_factor=0.1,
                status_forcelist=[500, 502, 503, 504]
            )
        )
        session.mount('http://', adapter)
        session.mount('https://', adapter)
        
        # Reconstruct URL without credentials
        clean_url = f"{parsed.scheme}://{parsed.hostname}"
        if parsed.port:
            clean_url += f":{parsed.port}"
        clean_url += parsed.path
        if parsed.query:
            clean_url += f"?{parsed.query}"
        
        auth = None
        response = None
        
        if parsed.username and parsed.password:
            # Try HTTP Digest authentication first (common for IP cameras)
            try:
                auth = HTTPDigestAuth(parsed.username, parsed.password)
                response = session.get(clean_url, auth=auth, headers=headers, timeout=(5, 15), stream=True)
                if response.status_code == 200:
                    logger.debug(f"Successfully authenticated using HTTP Digest for {clean_url}")
                elif response.status_code == 401:
                    # If Digest fails, try Basic auth
                    logger.debug(f"HTTP Digest failed, trying Basic auth for {clean_url}")
                    auth = HTTPBasicAuth(parsed.username, parsed.password)
                    response = session.get(clean_url, auth=auth, headers=headers, timeout=(5, 15), stream=True)
                    if response.status_code == 200:
                        logger.debug(f"Successfully authenticated using HTTP Basic for {clean_url}")
            except Exception as auth_error:
                logger.debug(f"Authentication setup error: {auth_error}")
                # Fallback to original URL with embedded credentials
                response = session.get(url, headers=headers, timeout=(5, 15), stream=True)
        else:
            # No credentials in URL, make request as-is
            response = session.get(url, headers=headers, timeout=(5, 15), stream=True)
        
        if response and response.status_code == 200:
            # Read content with size limit to prevent memory issues
            content = b''
            max_size = 10 * 1024 * 1024  # 10MB limit
            for chunk in response.iter_content(chunk_size=8192):
                content += chunk
                if len(content) > max_size:
                    logger.error(f"Snapshot too large (>{max_size} bytes) from {clean_url}")
                    return None
            
            # Convert response content to numpy array
            nparr = np.frombuffer(content, np.uint8)
            # Decode image
            frame = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
            if frame is not None:
                logger.debug(f"Successfully fetched snapshot from {clean_url}, shape: {frame.shape}, size: {len(content)} bytes")
                return frame
            else:
                logger.error(f"Failed to decode image from snapshot URL: {clean_url} (content size: {len(content)} bytes)")
                return None
        elif response:
            logger.error(f"Failed to fetch snapshot (status code {response.status_code}): {clean_url}")
            # Log response headers and first part of content for debugging
            logger.debug(f"Response headers: {dict(response.headers)}")
            if len(response.content) < 1000:
                logger.debug(f"Response content: {response.content[:500]}")
            return None
        else:
            logger.error(f"No response received from snapshot URL: {clean_url}")
            return None
    except requests.exceptions.Timeout as e:
        logger.error(f"Timeout fetching snapshot from {url}: {str(e)}")
        return None
    except requests.exceptions.ConnectionError as e:
        logger.error(f"Connection error fetching snapshot from {url}: {str(e)}")
        return None
    except Exception as e:
        logger.error(f"Exception fetching snapshot from {url}: {str(e)}", exc_info=True)
        return None

# Helper to get crop coordinates from stream
def get_crop_coords(stream):
    return {
        "cropX1": stream.get("cropX1"),
        "cropY1": stream.get("cropY1"),
        "cropX2": stream.get("cropX2"),
        "cropY2": stream.get("cropY2")
    }

# Camera state management functions
def set_camera_connected(camera_id, connected=True, error_msg=None):
    """Set camera connection state and track error information"""
    current_time = time.time()
    
    if camera_id not in camera_states:
        camera_states[camera_id] = {
            "connected": True,
            "last_error": None,
            "last_error_time": None,
            "consecutive_failures": 0,
            "disconnection_notified": False
        }
    
    state = camera_states[camera_id]
    was_connected = state["connected"]
    
    if connected:
        state["connected"] = True
        state["consecutive_failures"] = 0
        state["disconnection_notified"] = False
        if not was_connected:
            logger.info(f"📶 CAMERA RECONNECTED: {camera_id}")
    else:
        state["connected"] = False
        state["last_error"] = error_msg
        state["last_error_time"] = current_time
        state["consecutive_failures"] += 1
        
        # Distinguish between temporary and permanent disconnection
        is_permanent = state["consecutive_failures"] >= 3
        
        if was_connected and is_permanent:
            logger.error(f"📵 CAMERA DISCONNECTED: {camera_id} - {error_msg} (consecutive failures: {state['consecutive_failures']})")
            logger.info(f"🚨 CAMERA ERROR DETECTED - Will send detection: null to reset backend session for {camera_id}")

def is_camera_connected(camera_id):
    """Check if camera is currently connected"""
    return camera_states.get(camera_id, {}).get("connected", True)

def should_notify_disconnection(camera_id):
    """Check if we should notify backend about disconnection"""
    state = camera_states.get(camera_id, {})
    is_disconnected = not state.get("connected", True)
    not_yet_notified = not state.get("disconnection_notified", False)
    has_enough_failures = state.get("consecutive_failures", 0) >= 3
    
    return is_disconnected and not_yet_notified and has_enough_failures

def mark_disconnection_notified(camera_id):
    """Mark that we've notified backend about this disconnection"""
    if camera_id in camera_states:
        camera_states[camera_id]["disconnection_notified"] = True
        logger.debug(f"Marked disconnection notification sent for camera {camera_id}")

def get_or_init_session_pipeline_state(camera_id):
    """Get or initialize session pipeline state for a camera"""
    if camera_id not in session_pipeline_states:
        session_pipeline_states[camera_id] = {
            "mode": "validation_detecting",  # "validation_detecting", "send_detections", "waiting_for_session_id", "full_pipeline", "lightweight", "car_gone_waiting"
            "session_id_received": False,
            "full_pipeline_completed": False,
            "absence_counter": 0,
            "validation_counter": 0,         # Counter for validation phase
            "validation_threshold": 4,       # Default validation threshold
            "max_absence_frames": 3,
            "yolo_inference_enabled": True,  # Controls whether to run YOLO inference
            "cached_detection_dict": None,   # Cached detection dict for lightweight mode
            "stable_track_id": None,         # The stable track ID we're monitoring
            "validated_detection": None,     # Stored detection result from validation phase for full_pipeline reuse
            "progression_stage": None        # Tracks current progression stage (welcome, car_wait_staff, car_fueling, car_waitpayment)
        }
    return session_pipeline_states[camera_id]

def update_session_pipeline_mode(camera_id, new_mode, session_id=None):
    """Update session pipeline mode and related state"""
    state = get_or_init_session_pipeline_state(camera_id)
    old_mode = state["mode"]
    state["mode"] = new_mode
    
    # Reset counters based on mode transition
    if new_mode == "validation_detecting":
        # Transitioning to validation mode - reset both counters for fresh start
        old_validation_counter = state.get("validation_counter", 0)
        old_absence_counter = state.get("absence_counter", 0)
        state["validation_counter"] = 0
        state["absence_counter"] = 0
        if old_validation_counter > 0 or old_absence_counter > 0:
            logger.info(f"🧹 Camera {camera_id}: VALIDATION MODE RESET - validation_counter: {old_validation_counter}→0, absence_counter: {old_absence_counter}→0")
    
    if session_id:
        state["session_id_received"] = True
        state["absence_counter"] = 0  # Reset absence counter when session starts
    
    logger.info(f"📊 Camera {camera_id}: Pipeline mode changed from '{old_mode}' to '{new_mode}'")
    return state

####################################################
# REST API endpoint for image retrieval
####################################################
@app.get("/lpr/debug")
async def get_lpr_debug_info():
    """Debug endpoint to inspect LPR integration state"""
    try:
        return {
            "status": "success",
            "lpr_integration_started": lpr_integration_started,
            "redis_connected": redis_client_global is not None and redis_client_global.ping() if redis_client_global else False,
            "active_sessions": len(session_detections),
            "session_details": {
                session_id: {
                    "camera_id": session_to_camera.get(session_id, "unknown"),
                    "timestamp": detection_timestamps.get(session_id, 0),
                    "age_seconds": time.time() - detection_timestamps.get(session_id, time.time()),
                    "has_license": session_detections[session_id].get('data', {}).get('detection', {}).get('licensePlateText') is not None
                }
                for session_id in session_detections.keys()
            },
            "thread_status": {
                "lpr_listener_alive": lpr_listener_thread.is_alive() if lpr_listener_thread else False,
                "cleanup_timer_alive": cleanup_timer_thread.is_alive() if cleanup_timer_thread else False,
                "model_registry": get_registry_status(),
                "mpta_manager": get_mpta_manager_status()
            },
            "cached_detections_by_camera": list(cached_detections.keys())
        }
    except Exception as e:
        return {
            "status": "error",
            "error": str(e),
            "lpr_integration_started": lpr_integration_started
        }

@app.get("/camera/{camera_id}/image")
async def get_camera_image(camera_id: str):
    """
    Get the current frame from a camera as JPEG image
    """
    try:
        # URL decode the camera_id to handle encoded characters like %3B for semicolon
        from urllib.parse import unquote
        original_camera_id = camera_id
        camera_id = unquote(camera_id)
        logger.debug(f"REST API request: original='{original_camera_id}', decoded='{camera_id}'")
        
        with streams_lock:
            if camera_id not in streams:
                logger.warning(f"Camera ID '{camera_id}' not found in streams. Current streams: {list(streams.keys())}")
                raise HTTPException(status_code=404, detail=f"Camera {camera_id} not found or not active")
            
            # Check if we have a cached frame for this camera
            if camera_id not in latest_frames:
                logger.warning(f"No cached frame available for camera '{camera_id}'.")
                raise HTTPException(status_code=404, detail=f"No frame available for camera {camera_id}")
            
            frame = latest_frames[camera_id]
            logger.debug(f"Retrieved cached frame for camera '{camera_id}', frame shape: {frame.shape}")
        # Encode frame as JPEG
        success, buffer_img = cv2.imencode('.jpg', frame, [cv2.IMWRITE_JPEG_QUALITY, 85])
        if not success:
            raise HTTPException(status_code=500, detail="Failed to encode image as JPEG")
        
        # Return image as binary response
        return Response(content=buffer_img.tobytes(), media_type="image/jpeg")
        
    except HTTPException:
        raise
    except Exception as e:
        logger.error(f"Error retrieving image for camera {camera_id}: {str(e)}", exc_info=True)
        raise HTTPException(status_code=500, detail=f"Internal server error: {str(e)}")

####################################################
# Detection and frame processing functions
####################################################
@app.websocket("/")
async def detect(websocket: WebSocket):
    logger.info("WebSocket connection accepted")
    persistent_data_dict = {}

    async def handle_detection(camera_id, stream, frame, websocket, model_tree, persistent_data):
        try:
            # Check camera connection state first - handle disconnection immediately
            if should_notify_disconnection(camera_id):
                logger.error(f"🚨 CAMERA DISCONNECTION DETECTED: {camera_id} - sending immediate detection: null")
                
                # Clear cached detections and occupancy state
                cached_detections.pop(camera_id, None)
                frame_skip_flags.pop(camera_id, None)
                cached_full_pipeline_results.pop(camera_id, None)  # Clear cached pipeline results
                session_pipeline_states.pop(camera_id, None)  # Reset session pipeline state
                
                # Reset pipeline state immediately
                from siwatsystem.pympta import reset_tracking_state
                model_id = stream.get("modelId", "unknown")
                reset_tracking_state(camera_id, model_id, "camera disconnected")
                
                # Send immediate detection: null to backend
                detection_data = {
                    "type": "imageDetection",
                    "subscriptionIdentifier": stream["subscriptionIdentifier"],
                    "timestamp": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
                    "data": {
                        "detection": None,  # null detection for disconnection
                        "modelId": stream["modelId"],
                        "modelName": stream["modelName"]
                    }
                }
                
                try:
                    ws_logger.info(f"TX -> {json.dumps(detection_data, separators=(',', ':'))}")
                    await websocket.send_json(detection_data)
                except RuntimeError as e:
                    if "websocket.close" in str(e):
                        logger.warning(f"WebSocket connection closed - cannot send disconnection signal for camera {camera_id}")
                        return persistent_data
                    else:
                        raise
                mark_disconnection_notified(camera_id)
                logger.info(f"📡 SENT DISCONNECTION SIGNAL - detection: null for camera {camera_id}, backend should clear session")
                
                return persistent_data
            
            # Apply crop if specified
            cropped_frame = frame
            if all(coord is not None for coord in [stream.get("cropX1"), stream.get("cropY1"), stream.get("cropX2"), stream.get("cropY2")]):
                cropX1, cropY1, cropX2, cropY2 = stream["cropX1"], stream["cropY1"], stream["cropX2"], stream["cropY2"]
                cropped_frame = frame[cropY1:cropY2, cropX1:cropX2]
                logger.debug(f"Applied crop coordinates ({cropX1}, {cropY1}, {cropX2}, {cropY2}) to frame for camera {camera_id}")
            
            logger.debug(f"Processing frame for camera {camera_id} with model {stream['modelId']}")
            start_time = time.time()
            
            # Extract display identifier for pipeline context
            subscription_parts = stream["subscriptionIdentifier"].split(';')
            display_identifier = subscription_parts[0] if subscription_parts else None
            
            # Get backend session ID if available
            backend_session_id = session_ids.get(display_identifier)
            
            # Get or initialize session pipeline state
            pipeline_state = get_or_init_session_pipeline_state(camera_id)
            current_mode = pipeline_state["mode"]
            
            logger.debug(f"🔍 SESSIONID LOOKUP: display='{display_identifier}', session_id={repr(backend_session_id)}, mode='{current_mode}'")
            logger.debug(f"🔍 Available session_ids: {session_ids}")
            logger.debug(f"🔍 VALIDATED_DETECTION TRACE: {pipeline_state.get('validated_detection')}")
            
            # ═══ SESSION ID-BASED PROCESSING MODE ═══
            if not backend_session_id:
                # No session ID - handle different modes appropriately
                if current_mode == "lightweight":
                    # Check if we're in car_waitpayment stage - if so, don't reset immediately
                    current_progression = pipeline_state.get("progression_stage")
                    if current_progression == "car_waitpayment":
                        # Stay in lightweight mode - let absence counter + sessionId null logic handle reset
                        logger.debug(f"🔍 Camera {camera_id}: No session ID but in car_waitpayment - staying in lightweight mode for dual reset condition")
                    else:
                        # Not in car_waitpayment - reset immediately (situation 1)
                        update_session_pipeline_mode(camera_id, "validation_detecting")
                        current_mode = "validation_detecting"
                        logger.debug(f"🔍 Camera {camera_id}: No session ID - reset to validation_detecting (not in car_waitpayment)")
                elif current_mode not in ["validation_detecting", "send_detections", "waiting_for_session_id"]:
                    # Other modes - reset to validation_detecting
                    update_session_pipeline_mode(camera_id, "validation_detecting")
                    current_mode = "validation_detecting"
                    logger.debug(f"🔍 Camera {camera_id}: No session ID - reset to validation_detecting from {current_mode}")
                else:
                    logger.debug(f"🔍 Camera {camera_id}: No session ID - staying in {current_mode} mode")
            else:
                # Session ID available - switch to full pipeline mode
                if current_mode in ["send_detections", "waiting_for_session_id"]:
                    # Session ID just arrived - switch to full pipeline mode
                    update_session_pipeline_mode(camera_id, "full_pipeline", backend_session_id)
                    current_mode = "full_pipeline"
                    logger.info(f"🔥 Camera {camera_id}: Session ID received ({backend_session_id}) - switching to FULL PIPELINE mode")
            
            # Create context for pipeline execution
            pipeline_context = {
                "camera_id": camera_id,
                "display_id": display_identifier,
                "backend_session_id": backend_session_id,
                "current_mode": current_mode  # Pass current mode to pipeline
            }
            
            start_time = time.time()
            detection_result = None
            
            if current_mode == "validation_detecting":
                # ═══ TRACK VALIDATION MODE ═══
                # Run tracking-based validation with track ID stability
                logger.debug(f"🔍 Camera {camera_id}: In validation_detecting mode - running track-based validation")
                
                # Get tracking configuration from model_tree
                tracking_config = model_tree.get("tracking", {})
                tracking_enabled = tracking_config.get("enabled", True)
                stability_threshold = tracking_config.get("stabilityThreshold", 4)
                
                # Default to "none" - only proceed after track validation
                detection_result = {"class": "none", "confidence": 1.0, "bbox": [0, 0, 0, 0]}
                
                if tracking_enabled:
                    # Run full tracking detection to get track IDs
                    from siwatsystem.pympta import run_detection_with_tracking
                    all_detections, regions_dict, track_validation_result = run_detection_with_tracking(cropped_frame, model_tree, pipeline_context)
                    
                    if track_validation_result.get("validation_complete", False):
                        # Track validation completed - we have stable track IDs
                        stable_tracks = track_validation_result.get("stable_tracks", [])
                        logger.info(f"🎯 Camera {camera_id}: TRACK VALIDATION COMPLETED - stable tracks: {stable_tracks}")
                        
                        # Switch to send_detections mode
                        update_session_pipeline_mode(camera_id, "send_detections")
                        
                        # Send the best detection with stable track
                        if all_detections:
                            # Find detection with stable track ID
                            stable_detection = None
                            for detection in all_detections:
                                if detection.get("id") in stable_tracks:
                                    stable_detection = detection
                                    break
                            
                            if stable_detection:
                                detection_result = {
                                    "class": stable_detection.get("class", "car"),
                                    "confidence": stable_detection.get("confidence", 0.0),
                                    "bbox": stable_detection.get("bbox", [0, 0, 0, 0]),
                                    "track_id": stable_detection.get("id")
                                }
                                
                                # Store validated detection for full_pipeline mode to reuse
                                pipeline_state["validated_detection"] = detection_result.copy()
                                logger.debug(f"🔍 Camera {camera_id}: VALIDATION DEBUG - storing detection_result = {detection_result}")
                                logger.debug(f"🔍 Camera {camera_id}: VALIDATION DEBUG - pipeline_state after storing = {pipeline_state.get('validated_detection')}")
                                logger.info(f"🚗 Camera {camera_id}: SENDING STABLE DETECTION - track ID {detection_result['track_id']}")
                                logger.info(f"💾 Camera {camera_id}: STORED VALIDATED DETECTION for full_pipeline reuse")
                            else:
                                logger.warning(f"⚠️ Camera {camera_id}: Stable tracks found but no matching detection")
                    else:
                        # Track validation still in progress
                        stable_tracks = track_validation_result.get("stable_tracks", [])
                        current_tracks = track_validation_result.get("current_tracks", [])
                        
                        if current_tracks:
                            track_id = current_tracks[0] if current_tracks else "None"
                            stable_status = "STABLE" if stable_tracks else "validating"
                            logger.info(f"🔍 Camera {camera_id}: TRACK VALIDATION - car track_id {track_id} ({stable_status}, need {stability_threshold} consecutive frames)")
                        else:
                            logger.debug(f"👻 Camera {camera_id}: No car detected")
                        
                        logger.debug(f"📤 Camera {camera_id}: Sending 'none' (track validation in progress)")
                else:
                    # Tracking disabled - fall back to basic detection validation
                    logger.debug(f"🔍 Camera {camera_id}: Tracking disabled - using basic detection validation")
                    from siwatsystem.pympta import run_lightweight_detection
                    basic_detection = run_lightweight_detection(cropped_frame, model_tree)
                    
                    if basic_detection and basic_detection.get("car_detected"):
                        best_detection = basic_detection.get("best_detection")
                        
                        # Increment validation counter for basic detection
                        pipeline_state["validation_counter"] += 1
                        current_count = pipeline_state["validation_counter"]
                        threshold = pipeline_state["validation_threshold"]
                        
                        if current_count >= threshold:
                            update_session_pipeline_mode(camera_id, "send_detections")
                            detection_result = {
                                "class": best_detection.get("class", "car"),
                                "confidence": best_detection.get("confidence", 0.0),
                                "bbox": best_detection.get("bbox", [0, 0, 0, 0])
                            }
                            
                            # Store validated detection for full_pipeline mode to reuse
                            pipeline_state["validated_detection"] = detection_result.copy()
                            logger.debug(f"🔍 Camera {camera_id}: BASIC VALIDATION DEBUG - storing detection_result = {detection_result}")
                            logger.info(f"💾 Camera {camera_id}: STORED BASIC VALIDATED DETECTION for full_pipeline reuse")
                            logger.info(f"🎯 Camera {camera_id}: BASIC VALIDATION COMPLETED after {current_count} frames")
                        else:
                            logger.info(f"📊 Camera {camera_id}: Basic validation progress {current_count}/{threshold}")
                    else:
                        # Reset validation counter
                        if pipeline_state["validation_counter"] > 0:
                            pipeline_state["validation_counter"] = 0
                            logger.info(f"🔄 Camera {camera_id}: Reset validation counter (no detection)")
                    
            elif current_mode == "send_detections":
                # ═══ SEND DETECTIONS MODE ═══
                # Validation completed - now send detection_dict for car detections, detection: null for no car
                logger.debug(f"📤 Camera {camera_id}: In send_detections mode - sending detection_dict for cars")
                from siwatsystem.pympta import run_lightweight_detection
                basic_detection = run_lightweight_detection(cropped_frame, model_tree)
                
                if basic_detection and basic_detection.get("car_detected"):
                    # Car detected - send detection_dict
                    best_detection = basic_detection.get("best_detection")
                    detection_result = {
                        "class": best_detection.get("class", "car"),
                        "confidence": best_detection.get("confidence", 0.0),
                        "bbox": best_detection.get("bbox", [0, 0, 0, 0])
                    }
                    logger.info(f"🚗 Camera {camera_id}: SENDING DETECTION_DICT - {detection_result['class']} (conf={detection_result['confidence']:.3f}) - backend should generate session ID")
                else:
                    # No car detected - send "none"
                    detection_result = {"class": "none", "confidence": 1.0, "bbox": [0, 0, 0, 0]}
                    logger.debug(f"👻 Camera {camera_id}: No car detected - sending 'none'")
                    
            elif current_mode == "waiting_for_session_id":
                # ═══ WAITING FOR SESSION ID MODE ═══
                # Stop processing snapshots, wait for session ID
                logger.debug(f"⏳ Camera {camera_id}: In waiting_for_session_id mode - not processing snapshots")
                return persistent_data  # Don't process or send anything
                
            elif current_mode == "full_pipeline":
                # ═══ FULL PIPELINE MODE ═══
                logger.info(f"🔥 Camera {camera_id}: Running FULL PIPELINE (classification branches + Redis + PostgreSQL)")
                
                # Use validated detection from validation phase instead of detecting again
                validated_detection = pipeline_state.get("validated_detection")
                logger.debug(f"🔍 Camera {camera_id}: FULL_PIPELINE DEBUG - validated_detection = {validated_detection}")
                logger.debug(f"🔍 Camera {camera_id}: FULL_PIPELINE DEBUG - pipeline_state keys = {list(pipeline_state.keys())}")
                if validated_detection:
                    logger.info(f"🔄 Camera {camera_id}: Using validated detection for full pipeline: track_id={validated_detection.get('track_id')}")
                    detection_result = run_pipeline(cropped_frame, model_tree, context=pipeline_context, validated_detection=validated_detection)
                    # Clear the validated detection after using it
                    pipeline_state["validated_detection"] = None
                else:
                    logger.warning(f"⚠️ Camera {camera_id}: No validated detection found for full pipeline - this shouldn't happen")
                    detection_result = run_pipeline(cropped_frame, model_tree, context=pipeline_context)
                
                if detection_result and isinstance(detection_result, dict):
                    # Cache the full pipeline result
                    cached_full_pipeline_results[camera_id] = {
                        "result": detection_result.copy(),
                        "timestamp": time.time()
                    }
                    
                    # Note: Will cache detection_dict after branch processing completes
                    
                    # Store the stable track ID for lightweight monitoring
                    track_id = detection_result.get("track_id") or detection_result.get("id")
                    if track_id is not None:
                        pipeline_state["stable_track_id"] = track_id
                        logger.info(f"💾 Camera {camera_id}: Cached stable track_id={track_id}")
                    else:
                        logger.warning(f"⚠️ Camera {camera_id}: No track_id found in detection_result: {detection_result.keys()}")
                    
                    # Ensure we have a cached detection dict for lightweight mode
                    if not pipeline_state.get("cached_detection_dict"):
                        # Create fallback cached detection dict if branch processing didn't populate it
                        fallback_detection = {
                            "carModel": None,
                            "carBrand": None, 
                            "carYear": None,
                            "bodyType": None,
                            "licensePlateText": None,
                            "licensePlateConfidence": None
                        }
                        pipeline_state["cached_detection_dict"] = fallback_detection
                        logger.warning(f"⚠️ Camera {camera_id}: Created fallback cached detection dict (branch processing may have failed)")
                    
                    # Switch to lightweight mode
                    update_session_pipeline_mode(camera_id, "lightweight")
                    logger.info(f"✅ Camera {camera_id}: Full pipeline completed - switching to LIGHTWEIGHT mode")
                
            elif current_mode == "lightweight":
                # ═══ SIMPLIFIED LIGHTWEIGHT MODE ═══ 
                # Send cached detection dict + check for 2 consecutive empty frames to reset
                
                stable_track_id = pipeline_state.get("stable_track_id")
                cached_detection_dict = pipeline_state.get("cached_detection_dict")
                
                logger.debug(f"🪶 Camera {camera_id}: LIGHTWEIGHT MODE - stable_track_id={stable_track_id}")
                
                if not pipeline_state.get("yolo_inference_enabled", True):
                    # YOLO inference disabled during car_fueling - continue sending cached detection dict
                    logger.debug(f"🛑 Camera {camera_id}: YOLO inference disabled during car_fueling - continue sending cached detection dict")
                    if cached_detection_dict:
                        detection_result = cached_detection_dict  # Continue sending cached data
                        logger.info(f"⛽ Camera {camera_id}: YOLO disabled during car_fueling but sending cached detection dict")
                    else:
                        logger.warning(f"⚠️ Camera {camera_id}: YOLO disabled but no cached detection dict available")
                        detection_result = None
                else:
                    # Run lightweight YOLO inference to check car presence for reset logic (no tracking validation needed)
                    from siwatsystem.pympta import run_lightweight_detection
                    basic_detection = run_lightweight_detection(cropped_frame, model_tree)
                    
                    any_car_detected = basic_detection and basic_detection.get("car_detected", False)
                    logger.debug(f"🔍 Camera {camera_id}: LIGHTWEIGHT - simple car presence check: {any_car_detected}")
                    
                    if any_car_detected:
                        # Car detected - reset absence counter, continue sending cached detection dict
                        pipeline_state["absence_counter"] = 0  # Reset absence since cars are present
                        
                        if cached_detection_dict:
                            detection_result = cached_detection_dict  # Always send cached data
                            logger.info(f"💾 Camera {camera_id}: LIGHTWEIGHT - car detected, sending cached detection dict")
                        else:
                            logger.warning(f"⚠️ Camera {camera_id}: LIGHTWEIGHT - car detected but no cached detection dict available")
                            detection_result = None
                    else:
                        # No car detected - increment absence counter
                        pipeline_state["absence_counter"] += 1
                        absence_count = pipeline_state["absence_counter"]
                        max_absence = 3  # Need 3 consecutive empty frames
                        
                        logger.info(f"👻 Camera {camera_id}: LIGHTWEIGHT - no car detected (absence {absence_count}/{max_absence})")
                        
                        # Check if we should reset: Need BOTH 3 consecutive absence frames AND sessionId: null
                        current_progression = pipeline_state.get("progression_stage")
                        should_check_session_null = current_progression == "car_waitpayment"
                        
                        if absence_count >= max_absence:
                            if should_check_session_null:
                                # In car_waitpayment stage - require BOTH conditions
                                if backend_session_id is None:
                                    # Both conditions met: 3 absence frames + sessionId: null
                                    logger.info(f"🔄 Camera {camera_id}: DUAL RESET CONDITIONS MET - {max_absence} consecutive absence frames + sessionId: null")
                                    
                                    # Clear all state and prepare for next car
                                    cached_full_pipeline_results.pop(camera_id, None)
                                    pipeline_state["cached_detection_dict"] = None
                                    pipeline_state["stable_track_id"] = None
                                    pipeline_state["validated_detection"] = None
                                    pipeline_state["progression_stage"] = None
                                    old_absence_counter = pipeline_state["absence_counter"]
                                    old_validation_counter = pipeline_state.get("validation_counter", 0)
                                    pipeline_state["absence_counter"] = 0
                                    pipeline_state["validation_counter"] = 0
                                    pipeline_state["yolo_inference_enabled"] = True
                                    
                                    logger.info(f"🧹 Camera {camera_id}: DUAL RESET - absence_counter: {old_absence_counter}→0, validation_counter: {old_validation_counter}→0, progression_stage: {current_progression}→None")
                                    
                                    # Clear stability tracking data for this camera
                                    from siwatsystem.pympta import reset_camera_stability_tracking
                                    reset_camera_stability_tracking(camera_id, model_tree.get("modelId", "unknown"))
                                    
                                    # Switch back to validation phase
                                    update_session_pipeline_mode(camera_id, "validation_detecting")
                                    logger.info(f"✅ Camera {camera_id}: DUAL RESET TO VALIDATION COMPLETE - ready for new car")
                                    
                                    # Now in validation mode - send what YOLO detection finds (will be null since no car)
                                    detection_result = {"class": "none", "confidence": 1.0, "bbox": [0, 0, 0, 0]}
                                else:
                                    # Only absence frames met, but sessionId is not null - continue sending cached detection
                                    logger.info(f"⏳ Camera {camera_id}: {max_absence} absence frames reached but sessionId={backend_session_id} (not null) - continuing with cached detection")
                                    if cached_detection_dict:
                                        detection_result = cached_detection_dict
                                    else:
                                        logger.warning(f"⚠️ Camera {camera_id}: No cached detection dict available")
                                        detection_result = None
                            else:
                                # Not in car_waitpayment - use original simple reset condition (situation 1)
                                logger.info(f"🔄 Camera {camera_id}: SIMPLE RESET CONDITION MET - {max_absence} consecutive empty frames (not in car_waitpayment)")
                                
                                # Clear all state and prepare for next car
                                cached_full_pipeline_results.pop(camera_id, None)
                                pipeline_state["cached_detection_dict"] = None
                                pipeline_state["stable_track_id"] = None
                                pipeline_state["validated_detection"] = None
                                pipeline_state["progression_stage"] = None
                                old_absence_counter = pipeline_state["absence_counter"]
                                old_validation_counter = pipeline_state.get("validation_counter", 0)
                                pipeline_state["absence_counter"] = 0
                                pipeline_state["validation_counter"] = 0
                                pipeline_state["yolo_inference_enabled"] = True
                                
                                logger.info(f"🧹 Camera {camera_id}: SIMPLE RESET - absence_counter: {old_absence_counter}→0, validation_counter: {old_validation_counter}→0")
                                
                                # Clear stability tracking data for this camera
                                from siwatsystem.pympta import reset_camera_stability_tracking
                                reset_camera_stability_tracking(camera_id, model_tree.get("modelId", "unknown"))
                                
                                # Switch back to validation phase
                                update_session_pipeline_mode(camera_id, "validation_detecting")
                                logger.info(f"✅ Camera {camera_id}: SIMPLE RESET TO VALIDATION COMPLETE - ready for new car")
                                
                                # Now in validation mode - send what YOLO detection finds (will be null since no car)
                                detection_result = {"class": "none", "confidence": 1.0, "bbox": [0, 0, 0, 0]}
                        else:
                            # Still within absence threshold - continue sending cached detection dict
                            if cached_detection_dict:
                                detection_result = cached_detection_dict  # Send cached data
                                logger.info(f"⏳ Camera {camera_id}: LIGHTWEIGHT - no car but absence<{max_absence}, still sending cached detection dict")
                            else:
                                logger.warning(f"⚠️ Camera {camera_id}: LIGHTWEIGHT - no cached detection dict available")
                                detection_result = None
            
            elif current_mode == "car_gone_waiting":
                # ═══ CAR GONE WAITING STATE ═══ 
                # Car is gone (both conditions met), YOLO inference disabled, waiting for new session
                
                logger.debug(f"🛑 Camera {camera_id}: CAR GONE WAITING - YOLO inference stopped")
                
                # Check if backend has started a new session (indicates new car scenario)
                if backend_session_id is not None:
                    # Backend started new session - re-enable YOLO and reset to validation
                    pipeline_state["yolo_inference_enabled"] = True
                    pipeline_state["absence_counter"] = 0
                    pipeline_state["stable_track_id"] = None
                    pipeline_state["cached_detection_dict"] = None
                    pipeline_state["validated_detection"] = None
                    
                    # Clear stability tracking data for this camera
                    from siwatsystem.pympta import reset_camera_stability_tracking
                    reset_camera_stability_tracking(camera_id, model_tree.get("modelId", "unknown"))
                    
                    update_session_pipeline_mode(camera_id, "validation_detecting")
                    logger.info(f"🔄 Camera {camera_id}: New session detected (id={backend_session_id}) - re-enabling YOLO inference")
                    logger.info(f"✅ Camera {camera_id}: Reset to validation mode - cleared all tracking, ready for new car detection")
                    
                    # Don't run detection this frame - let next frame start fresh
                    detection_result = {"class": "none", "confidence": 1.0, "bbox": [0, 0, 0, 0]}
                else:
                    # Still waiting - no sessionId, no detection to send
                    logger.debug(f"🛑 Camera {camera_id}: Car gone waiting - no YOLO inference, no data sent")
                    detection_result = None
            
            process_time = (time.time() - start_time) * 1000
            logger.debug(f"Detection for camera {camera_id} completed in {process_time:.2f}ms (mode: {current_mode})")
            
            # Skip processing if no detection result (blocked by session gating)
            if detection_result is None:
                logger.debug(f"No detection result to process for camera {camera_id}")
                return persistent_data
                
            # Log the raw detection result for debugging
            logger.debug(f"Raw detection result for camera {camera_id}:\n{json.dumps(detection_result, indent=2, default=str)}")
            
            # Extract session_id from pipeline result (always use backend sessionId)
            session_id = backend_session_id
            logger.debug(f"Using backend session_id: {session_id}")
            
            
            # Process detection result based on current mode
            if current_mode == "validation_detecting":
                # ═══ VALIDATION DETECTING MODE ═══
                # Always send detection: null during validation phase
                detection_dict = None
                logger.debug(f"🔍 SENDING 'NONE' - validation_detecting mode for camera {camera_id}")
                
            elif current_mode == "send_detections":
                # ═══ SEND DETECTIONS MODE ═══
                if detection_result.get("class") == "none":
                    # No car detected - send detection: null
                    detection_dict = None
                    logger.debug(f"📤 SENDING 'NONE' - send_detections mode (no car) for camera {camera_id}")
                else:
                    # Car detected in send_detections mode - ALWAYS send empty dict to trigger backend sessionId
                    # Purpose: Tell backend "car is here, please create sessionId"
                    detection_dict = {}
                    logger.info(f"📤 SENDING EMPTY DETECTION_DICT - send_detections mode, requesting backend to create sessionId (conf={detection_result.get('confidence', 0):.3f}) for camera {camera_id}")
                    
                    if backend_session_id:
                        logger.debug(f"🔄 Camera {camera_id}: Note - sessionId {backend_session_id} exists but still in send_detections mode (transition pending)")
                    
            elif current_mode == "lightweight":
                # ═══ SIMPLIFIED LIGHTWEIGHT MODE DETECTION PROCESSING ═══
                if detection_result.get("class") == "none":
                    # No car detected - this happens when resetting to validation
                    detection_dict = None  # Send detection: null
                    logger.info(f"🚫 LIGHTWEIGHT - no car detected, sending detection=null")
                elif isinstance(detection_result, dict) and ("carBrand" in detection_result or "carModel" in detection_result):
                    # Check if we're waiting for dual reset condition
                    current_progression = pipeline_state.get("progression_stage")
                    if current_progression == "car_waitpayment" and backend_session_id is None:
                        # In car_waitpayment + sessionId: null - STOP sending cached detection to prevent new session creation
                        detection_dict = None
                        logger.info(f"🛑 LIGHTWEIGHT - in car_waitpayment with sessionId: null, NOT sending cached detection (waiting for dual reset)")
                    else:
                        # Normal lightweight mode - send cached detection dict
                        detection_dict = detection_result
                        logger.info(f"💾 LIGHTWEIGHT - sending cached detection dict")
                else:
                    logger.warning(f"⚠️ LIGHTWEIGHT - unexpected detection_result type: {type(detection_result)}")
                    detection_dict = None
                        
            elif detection_result.get("class") == "none":
                # Other modes - send null to clear session
                detection_dict = None
                logger.info(f"📤 SENDING 'NONE' (detection: null) - Car absent, expecting backend to clear session for camera {camera_id}")
            elif detection_result and "carBrand" in detection_result:
                # Handle cached detection dict format (fallback for compatibility)
                detection_dict = detection_result
                logger.info(f"💾 Camera {camera_id}: LIGHTWEIGHT MODE - using detection_result as detection_dict:")
                logger.info(f"💾 Camera {camera_id}: - detection_dict: {detection_dict}")
            else:
                # Valid detection - convert to backend format (will be populated by branch processing)
                detection_dict = {
                    "carModel": None,
                    "carBrand": None,
                    "carYear": None,
                    "bodyType": None,
                    "licensePlateText": None,
                    "licensePlateConfidence": None
                }
            
            # Extract and process branch results from parallel classification (only for valid detections, skip cached mode)
            if detection_result.get("class") != "none" and "branch_results" in detection_result and not detection_result.get("cached_mode", False):
                def process_branch_results(branch_results, depth=0):
                    """Recursively process branch results including nested branches."""
                    if not isinstance(branch_results, dict):
                        return
                    
                    indent = "  " * depth
                    for branch_id, branch_data in branch_results.items():
                        if isinstance(branch_data, dict):
                            logger.debug(f"{indent}Processing branch {branch_id}: {branch_data}")
                            
                            # Map common classification fields to backend-expected names
                            if "brand" in branch_data:
                                detection_dict["carBrand"] = branch_data["brand"]
                                logger.debug(f"{indent}Mapped carBrand: {branch_data['brand']}")
                            if "body_type" in branch_data:
                                detection_dict["bodyType"] = branch_data["body_type"]
                                logger.debug(f"{indent}Mapped bodyType: {branch_data['body_type']}")
                            if "class" in branch_data:
                                class_name = branch_data["class"]
                                
                                # Map based on branch/model type
                                if "brand" in branch_id.lower():
                                    detection_dict["carBrand"] = class_name
                                    logger.debug(f"{indent}Mapped carBrand from class: {class_name}")
                                elif "bodytype" in branch_id.lower() or "body" in branch_id.lower():
                                    detection_dict["bodyType"] = class_name
                                    logger.debug(f"{indent}Mapped bodyType from class: {class_name}")
                            
                            # Process nested branch results recursively
                            if "branch_results" in branch_data:
                                logger.debug(f"{indent}Processing nested branches in {branch_id}")
                                process_branch_results(branch_data["branch_results"], depth + 1)

                branch_results = detection_result.get("branch_results", {})
                if branch_results:
                    logger.debug(f"Processing branch results: {branch_results}")
                    process_branch_results(branch_results)
                    logger.info(f"Detection payload after branch processing: {detection_dict}")
                    
                    # Cache the detection_dict for lightweight mode (after branch processing completes)
                    if current_mode == "full_pipeline":
                        pipeline_state = get_or_init_session_pipeline_state(camera_id)
                        pipeline_state["cached_detection_dict"] = detection_dict.copy()
                        logger.info(f"💾 Camera {camera_id}: CACHED DETECTION DICT after branch processing: {detection_dict}")
                        
                else:
                    logger.debug("No branch results found in detection result")
            
            detection_data = {
                "type": "imageDetection",
                "subscriptionIdentifier": stream["subscriptionIdentifier"],
                "timestamp": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
                # "timestamp": time.strftime("%Y-%m-%dT%H:%M:%S", time.gmtime()) + f".{int(time.time() * 1000) % 1000:03d}Z",
                "data": {
                    "detection": detection_dict,
                    "modelId": stream["modelId"],
                    "modelName": stream["modelName"]
                }
            }
            
            # SessionId should NEVER be sent from worker to backend - it's uni-directional (backend -> worker only)
            # Backend manages sessionIds independently based on detection content
            logger.debug(f"TX message prepared (no sessionId) - detection_dict type: {type(detection_dict)}")
            
            # Log detection details for different modes
            if current_mode == "lightweight":
                if detection_result and detection_result.get("class") == "none":
                    logger.info(f"🚫 Camera {camera_id}: LIGHTWEIGHT - No car detected (resetting to validation)")
                elif isinstance(detection_result, dict) and ("carBrand" in detection_result or "carModel" in detection_result):
                    logger.info(f"💾 Camera {camera_id}: LIGHTWEIGHT - Sending cached detection data")
                else:
                    logger.info(f"🪶 Camera {camera_id}: LIGHTWEIGHT - Processing detection")
            elif detection_result and "class" in detection_result and detection_result.get("class") != "none":
                confidence = detection_result.get("confidence", 0.0)
                logger.info(f"🚗 Camera {camera_id}: Detected {detection_result['class']} with confidence {confidence:.2f} using model {stream['modelName']}")
            
            # Send detection data to backend (session gating handled above in processing logic)
            logger.debug(f"📤 SENDING TO BACKEND for camera {camera_id}: {json.dumps(detection_data, indent=2)}")
            try:
                ws_logger.info(f"TX -> {json.dumps(detection_data, separators=(',', ':'))}")
                await websocket.send_json(detection_data)
                logger.debug(f"Sent detection data to client for camera {camera_id}")
                
                # Cache the detection data for potential resubscriptions (only if not null detection)
                if detection_dict is not None and detection_result.get("class") != "none":
                    cached_detections[camera_id] = detection_data.copy()
                    logger.debug(f"Cached detection for camera {camera_id}: {detection_dict}")
                    
                    # Enhanced caching: Store by session_id for LPR integration
                    session_id = detection_data.get('sessionId')
                    if session_id:
                        session_id_str = str(session_id)
                        session_detections[session_id_str] = detection_data.copy()
                        session_to_camera[session_id_str] = camera_id
                        detection_timestamps[session_id_str] = time.time()
                        logger.debug(f"🔑 Cached detection for LPR by session_id {session_id_str}: {camera_id}")
                else:
                    # Don't cache null/none detections - let them reset properly
                    cached_detections.pop(camera_id, None)
                    logger.debug(f"Not caching null/none detection for camera {camera_id}")
                
            except RuntimeError as e:
                if "websocket.close" in str(e):
                    logger.warning(f"WebSocket connection closed - cannot send detection data for camera {camera_id}")
                    return persistent_data
                else:
                    raise
            
            # Log status after sending (no sessionId sent to backend)
            if detection_dict is None:
                logger.info(f"📡 SENT 'none' detection - backend should clear session for camera {camera_id}")
            elif detection_dict == {}:
                logger.info(f"📡 SENT empty detection - backend should create sessionId for camera {camera_id}")
            else:
                logger.info(f"📡 SENT detection data - backend manages sessionId independently for camera {camera_id}")
            return persistent_data
        except Exception as e:
            logger.error(f"Error in handle_detection for camera {camera_id}: {str(e)}", exc_info=True)
            return persistent_data

    def frame_reader(camera_id, cap, buffer, stop_event):
        retries = 0
        logger.info(f"Starting frame reader thread for camera {camera_id}")
        frame_count = 0
        last_log_time = time.time()
        
        try:
            # Log initial camera status and properties
            if cap.isOpened():
                width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
                height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
                fps = cap.get(cv2.CAP_PROP_FPS)
                logger.info(f"Camera {camera_id} opened successfully with resolution {width}x{height}, FPS: {fps}")
                set_camera_connected(camera_id, True)
            else:
                logger.error(f"Camera {camera_id} failed to open initially")
                set_camera_connected(camera_id, False, "Failed to open camera initially")
            
            while not stop_event.is_set():
                try:
                    if not cap.isOpened():
                        logger.error(f"Camera {camera_id} is not open before trying to read")
                        # Attempt to reopen
                        cap = cv2.VideoCapture(streams[camera_id]["rtsp_url"])
                        time.sleep(reconnect_interval)
                        continue
                    
                    logger.debug(f"Attempting to read frame from camera {camera_id}")
                    ret, frame = cap.read()
                    
                    if not ret:
                        error_msg = f"Connection lost for camera: {camera_id}, retry {retries+1}/{max_retries}"
                        logger.warning(error_msg)
                        set_camera_connected(camera_id, False, error_msg)
                        cap.release()
                        time.sleep(reconnect_interval)
                        retries += 1
                        if retries > max_retries and max_retries != -1:
                            logger.error(f"Max retries reached for camera: {camera_id}, stopping frame reader")
                            set_camera_connected(camera_id, False, "Max retries reached")
                            break
                        # Re-open
                        logger.info(f"Attempting to reopen RTSP stream for camera: {camera_id}")
                        cap = cv2.VideoCapture(streams[camera_id]["rtsp_url"])
                        if not cap.isOpened():
                            logger.error(f"Failed to reopen RTSP stream for camera: {camera_id}")
                            set_camera_connected(camera_id, False, "Failed to reopen RTSP stream")
                            continue
                        logger.info(f"Successfully reopened RTSP stream for camera: {camera_id}")
                        set_camera_connected(camera_id, True)
                        continue
                    
                    # Successfully read a frame
                    frame_count += 1
                    current_time = time.time()
                    # Log frame stats every 5 seconds
                    if current_time - last_log_time > 5:
                        logger.info(f"Camera {camera_id}: Read {frame_count} frames in the last {current_time - last_log_time:.1f} seconds")
                        frame_count = 0
                        last_log_time = current_time
                    
                    logger.debug(f"Successfully read frame from camera {camera_id}, shape: {frame.shape}")
                    retries = 0
                    set_camera_connected(camera_id, True)  # Mark as connected on successful frame read
                    
                    # Overwrite old frame if buffer is full
                    if not buffer.empty():
                        try:
                            buffer.get_nowait()
                            logger.debug(f"[frame_reader] Removed old frame from buffer for camera {camera_id}")
                        except queue.Empty:
                            pass
                    buffer.put(frame)
                    logger.debug(f"[frame_reader] Added new frame to buffer for camera {camera_id}. Buffer size: {buffer.qsize()}")
                    
                    # Short sleep to avoid CPU overuse
                    time.sleep(0.01)
                    
                except cv2.error as e:
                    error_msg = f"OpenCV error for camera {camera_id}: {e}"
                    logger.error(error_msg, exc_info=True)
                    set_camera_connected(camera_id, False, error_msg)
                    cap.release()
                    time.sleep(reconnect_interval)
                    retries += 1
                    if retries > max_retries and max_retries != -1:
                        logger.error(f"Max retries reached after OpenCV error for camera {camera_id}")
                        set_camera_connected(camera_id, False, "Max retries reached after OpenCV error")
                        break
                    logger.info(f"Attempting to reopen RTSP stream after OpenCV error for camera: {camera_id}")
                    cap = cv2.VideoCapture(streams[camera_id]["rtsp_url"])
                    if not cap.isOpened():
                        logger.error(f"Failed to reopen RTSP stream for camera {camera_id} after OpenCV error")
                        set_camera_connected(camera_id, False, "Failed to reopen after OpenCV error")
                        continue
                    logger.info(f"Successfully reopened RTSP stream after OpenCV error for camera: {camera_id}")
                    set_camera_connected(camera_id, True)
                except Exception as e:
                    error_msg = f"Unexpected error for camera {camera_id}: {str(e)}"
                    logger.error(error_msg, exc_info=True)
                    set_camera_connected(camera_id, False, error_msg)
                    cap.release()
                    break
        except Exception as e:
            logger.error(f"Error in frame_reader thread for camera {camera_id}: {str(e)}", exc_info=True)
        finally:
            logger.info(f"Frame reader thread for camera {camera_id} is exiting")
            if cap and cap.isOpened():
                cap.release()

    def snapshot_reader(camera_id, snapshot_url, snapshot_interval, buffer, stop_event):
        """Frame reader that fetches snapshots from HTTP/HTTPS URL at specified intervals"""
        retries = 0
        consecutive_failures = 0  # Track consecutive failures for backoff
        logger.info(f"Starting snapshot reader thread for camera {camera_id} from {snapshot_url}")
        frame_count = 0
        last_log_time = time.time()
        
        # Initialize camera state
        set_camera_connected(camera_id, True)
        
        try:
            interval_seconds = snapshot_interval / 1000.0  # Convert milliseconds to seconds
            logger.info(f"Snapshot interval for camera {camera_id}: {interval_seconds}s")
            
            while not stop_event.is_set():
                try:
                    start_time = time.time()
                    frame = fetch_snapshot(snapshot_url)
                    
                    if frame is None:
                        consecutive_failures += 1
                        error_msg = f"Failed to fetch snapshot for camera: {camera_id}, consecutive failures: {consecutive_failures}"
                        logger.warning(error_msg)
                        set_camera_connected(camera_id, False, error_msg)
                        retries += 1
                        
                        # Check network connectivity with a simple ping-like test
                        if consecutive_failures % 5 == 1:  # Every 5th failure, test connectivity
                            try:
                                test_response = requests.get(snapshot_url, timeout=(2, 5), stream=False)
                                logger.info(f"Camera {camera_id}: Connectivity test result: {test_response.status_code}")
                            except Exception as test_error:
                                logger.warning(f"Camera {camera_id}: Connectivity test failed: {test_error}")
                        
                        if retries > max_retries and max_retries != -1:
                            logger.error(f"Max retries reached for snapshot camera: {camera_id}, stopping reader")
                            set_camera_connected(camera_id, False, "Max retries reached for snapshot camera")
                            break
                        
                        # Exponential backoff based on consecutive failures
                        backoff_delay = min(30, max(1, min(2 ** min(consecutive_failures - 1, 6), interval_seconds * 2)))  # Start with 1s, max 30s
                        logger.debug(f"Camera {camera_id}: Backing off for {backoff_delay:.1f}s (consecutive failures: {consecutive_failures})")
                        if stop_event.wait(backoff_delay):  # Use wait with timeout instead of sleep
                            break  # Exit if stop_event is set during backoff
                        continue
                    
                    # Successfully fetched a frame - reset consecutive failures
                    consecutive_failures = 0  # Reset backoff on success
                    frame_count += 1
                    current_time = time.time()
                    # Log frame stats every 5 seconds
                    if current_time - last_log_time > 5:
                        logger.info(f"Camera {camera_id}: Fetched {frame_count} snapshots in the last {current_time - last_log_time:.1f} seconds")
                        frame_count = 0
                        last_log_time = current_time
                    
                    logger.debug(f"Successfully fetched snapshot from camera {camera_id}, shape: {frame.shape}")
                    retries = 0
                    set_camera_connected(camera_id, True)  # Mark as connected on successful snapshot
                    
                    # Overwrite old frame if buffer is full
                    if not buffer.empty():
                        try:
                            buffer.get_nowait()
                            logger.debug(f"[snapshot_reader] Removed old snapshot from buffer for camera {camera_id}")
                        except queue.Empty:
                            pass
                    buffer.put(frame)
                    logger.debug(f"[snapshot_reader] Added new snapshot to buffer for camera {camera_id}. Buffer size: {buffer.qsize()}")
                    
                    # Wait for the specified interval
                    elapsed = time.time() - start_time
                    sleep_time = max(interval_seconds - elapsed, 0)
                    if sleep_time > 0:
                        time.sleep(sleep_time)
                
                except Exception as e:
                    consecutive_failures += 1
                    error_msg = f"Unexpected error fetching snapshot for camera {camera_id}: {str(e)}"
                    logger.error(error_msg, exc_info=True)
                    set_camera_connected(camera_id, False, error_msg)
                    retries += 1
                    if retries > max_retries and max_retries != -1:
                        logger.error(f"Max retries reached after error for snapshot camera {camera_id}")
                        set_camera_connected(camera_id, False, "Max retries reached after error")
                        break
                    
                    # Exponential backoff for exceptions too
                    backoff_delay = min(30, max(1, min(2 ** min(consecutive_failures - 1, 6), interval_seconds * 2)))  # Start with 1s, max 30s
                    logger.debug(f"Camera {camera_id}: Exception backoff for {backoff_delay:.1f}s (consecutive failures: {consecutive_failures})")
                    if stop_event.wait(backoff_delay):  # Use wait with timeout instead of sleep
                        break  # Exit if stop_event is set during backoff
        except Exception as e:
            logger.error(f"Error in snapshot_reader thread for camera {camera_id}: {str(e)}", exc_info=True)
        finally:
            logger.info(f"Snapshot reader thread for camera {camera_id} is exiting")

    async def reconcile_subscriptions(desired_subscriptions, websocket):
        """
        Declarative reconciliation: Compare desired vs current subscriptions and make changes
        """
        logger.info(f"Reconciling subscriptions: {len(desired_subscriptions)} desired")
        
        with streams_lock:
            # Get current subscriptions
            current_subscription_ids = set(streams.keys())
            desired_subscription_ids = set(sub["subscriptionIdentifier"] for sub in desired_subscriptions)
            
            # Find what to add and remove
            to_add = desired_subscription_ids - current_subscription_ids
            to_remove = current_subscription_ids - desired_subscription_ids
            to_check_for_changes = current_subscription_ids & desired_subscription_ids
            
            logger.info(f"Reconciliation: {len(to_add)} to add, {len(to_remove)} to remove, {len(to_check_for_changes)} to check for changes")
            
            # Remove subscriptions that are no longer wanted
            for subscription_id in to_remove:
                await unsubscribe_internal(subscription_id)
                
            # Check existing subscriptions for parameter changes
            for subscription_id in to_check_for_changes:
                desired_sub = next(sub for sub in desired_subscriptions if sub["subscriptionIdentifier"] == subscription_id)
                current_stream = streams[subscription_id]
                
                # Check if parameters changed
                if has_subscription_changed(desired_sub, current_stream):
                    logger.info(f"Parameters changed for {subscription_id}, resubscribing")
                    logger.debug(f"Parameter comparison for {subscription_id}:")
                    logger.debug(f"  rtspUrl: '{desired_sub.get('rtspUrl')}' vs '{current_stream.get('rtsp_url')}'")
                    logger.debug(f"  snapshotUrl: '{desired_sub.get('snapshotUrl')}' vs '{current_stream.get('snapshot_url')}'")
                    logger.debug(f"  modelUrl: '{extract_model_file_identifier(desired_sub.get('modelUrl'))}' vs '{extract_model_file_identifier(current_stream.get('modelUrl'))}'")
                    logger.debug(f"  modelId: {desired_sub.get('modelId')} vs {current_stream.get('modelId')}")
                    
                    # Preserve detection state for resubscription 
                    cached_detection = cached_detections.get(subscription_id)
                    logger.debug(f"Preserving detection state for resubscription: {cached_detection is not None}")
                    
                    await unsubscribe_internal(subscription_id, preserve_detection=True)
                    await subscribe_internal(desired_sub, websocket, cached_detection=cached_detection)
                    
            # Add new subscriptions
            for subscription_id in to_add:
                desired_sub = next(sub for sub in desired_subscriptions if sub["subscriptionIdentifier"] == subscription_id)
                await subscribe_internal(desired_sub, websocket)

    def extract_model_file_identifier(model_url):
        """Extract the core model file identifier from S3 URLs, ignoring timestamp parameters"""
        if not model_url:
            return None
        
        # For S3 URLs, extract just the path portion before query parameters
        try:
            from urllib.parse import urlparse
            parsed = urlparse(model_url)
            # Return the path which contains the actual model file identifier
            # e.g. "/adsist-cms-staging/models/bangchak_poc-1756312318569.mpta"
            return parsed.path
        except Exception as e:
            logger.warning(f"Failed to parse model URL {model_url}: {e}")
            return model_url

    def has_subscription_changed(desired_sub, current_stream):
        """Check if subscription parameters have changed"""
        # Smart model URL comparison - ignore timestamp changes in signed URLs
        desired_model_id = extract_model_file_identifier(desired_sub.get("modelUrl"))
        current_model_id = extract_model_file_identifier(current_stream.get("modelUrl"))
        
        return (
            desired_sub.get("rtspUrl") != current_stream.get("rtsp_url") or
            desired_sub.get("snapshotUrl") != current_stream.get("snapshot_url") or  
            desired_sub.get("snapshotInterval") != current_stream.get("snapshot_interval") or
            desired_sub.get("cropX1") != current_stream.get("cropX1") or
            desired_sub.get("cropY1") != current_stream.get("cropY1") or
            desired_sub.get("cropX2") != current_stream.get("cropX2") or
            desired_sub.get("cropY2") != current_stream.get("cropY2") or
            desired_sub.get("modelId") != current_stream.get("modelId") or
            desired_sub.get("modelName") != current_stream.get("modelName") or
            desired_model_id != current_model_id
        )

    async def subscribe_internal(subscription, websocket, cached_detection=None):
        """Internal subscription logic extracted from original subscribe handler"""
        subscriptionIdentifier = subscription.get("subscriptionIdentifier")
        rtsp_url = subscription.get("rtspUrl") 
        snapshot_url = subscription.get("snapshotUrl")
        snapshot_interval = subscription.get("snapshotInterval")
        model_url = subscription.get("modelUrl")
        modelId = subscription.get("modelId")
        modelName = subscription.get("modelName")
        cropX1 = subscription.get("cropX1")
        cropY1 = subscription.get("cropY1")
        cropX2 = subscription.get("cropX2")
        cropY2 = subscription.get("cropY2")

        # Extract camera_id from subscriptionIdentifier
        parts = subscriptionIdentifier.split(';')
        if len(parts) != 2:
            logger.error(f"Invalid subscriptionIdentifier format: {subscriptionIdentifier}")
            return
        
        display_identifier, camera_identifier = parts
        camera_id = subscriptionIdentifier

        # Load model if needed using shared MPTA manager
        if model_url:
            with models_lock:
                if (camera_id not in models) or (modelId not in models[camera_id]):
                    logger.info(f"Getting shared MPTA for camera {camera_id}, modelId {modelId}")
                    
                    # Use shared MPTA manager for optimized downloads
                    mpta_result = get_or_download_mpta(modelId, model_url, camera_id)
                    if not mpta_result:
                        logger.error(f"Failed to get/download MPTA for modelId {modelId}")
                        return
                    
                    shared_extraction_path, local_mpta_file = mpta_result
                    
                    # Load pipeline from local MPTA file
                    model_tree = load_pipeline_from_zip(local_mpta_file, shared_extraction_path)
                    if model_tree is None:
                        logger.error(f"Failed to load model {modelId} from shared MPTA")
                        return
                        
                    if camera_id not in models:
                        models[camera_id] = {}
                    models[camera_id][modelId] = model_tree
                    
                    # Start LPR integration threads after first model is loaded (only once)
                    global lpr_integration_started
                    if not lpr_integration_started and hasattr(model_tree, 'get') and model_tree.get('redis_client'):
                        try:
                            start_lpr_integration()
                            lpr_integration_started = True
                            logger.info("🚀 LPR integration started after first model load")
                        except Exception as e:
                            logger.error(f"❌ Failed to start LPR integration: {e}")

        # Create stream (same logic as original)
        if camera_id and (rtsp_url or snapshot_url) and len(streams) < max_streams:
            camera_url = snapshot_url if snapshot_url else rtsp_url
            
            # Check if we already have a stream for this camera URL
            shared_stream = camera_streams.get(camera_url)
            
            if shared_stream:
                # Reuse existing stream
                buffer = shared_stream["buffer"]
                stop_event = shared_stream["stop_event"] 
                thread = shared_stream["thread"]
                mode = shared_stream["mode"]
                shared_stream["ref_count"] = shared_stream.get("ref_count", 0) + 1
            else:
                # Create new stream
                buffer = queue.Queue(maxsize=1)
                stop_event = threading.Event()
                
                if snapshot_url and snapshot_interval:
                    thread = threading.Thread(target=snapshot_reader, args=(camera_id, snapshot_url, snapshot_interval, buffer, stop_event))
                    thread.daemon = True
                    thread.start()
                    mode = "snapshot"
                    shared_stream = {
                        "buffer": buffer, "thread": thread, "stop_event": stop_event,
                        "mode": mode, "url": snapshot_url, "snapshot_interval": snapshot_interval, "ref_count": 1
                    }
                    camera_streams[camera_url] = shared_stream
                elif rtsp_url:
                    cap = cv2.VideoCapture(rtsp_url)
                    if not cap.isOpened():
                        logger.error(f"Failed to open RTSP stream for camera {camera_id}")
                        return
                    thread = threading.Thread(target=frame_reader, args=(camera_id, cap, buffer, stop_event))
                    thread.daemon = True
                    thread.start()
                    mode = "rtsp"
                    shared_stream = {
                        "buffer": buffer, "thread": thread, "stop_event": stop_event,
                        "mode": mode, "url": rtsp_url, "cap": cap, "ref_count": 1
                    }
                    camera_streams[camera_url] = shared_stream
                else:
                    logger.error(f"No valid URL provided for camera {camera_id}")
                    return

            # Create stream info
            stream_info = {
                "buffer": buffer, "thread": thread, "stop_event": stop_event,
                "modelId": modelId, "modelName": modelName, "subscriptionIdentifier": subscriptionIdentifier,
                "cropX1": cropX1, "cropY1": cropY1, "cropX2": cropX2, "cropY2": cropY2,
                "mode": mode, "camera_url": camera_url, "modelUrl": model_url,
                # Always store both URLs for comparison consistency
                "rtsp_url": rtsp_url,
                "snapshot_url": snapshot_url,
                "snapshot_interval": snapshot_interval
            }
            
            if mode == "rtsp":
                stream_info["cap"] = shared_stream["cap"]
            
            streams[camera_id] = stream_info
            subscription_to_camera[camera_id] = camera_url
            logger.info(f"Subscribed to camera {camera_id}")
            
            # Send initial detection to backend - use cached if available, otherwise "none"
            if cached_detection:
                # Restore cached detection with updated timestamp (RESUBSCRIPTION STATUS UPDATE)
                initial_detection_data = cached_detection.copy()
                initial_detection_data["timestamp"] = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
                logger.info(f"📡 RESUBSCRIPTION: Restoring cached detection for camera {camera_id}")
                logger.debug(f"📡 RESUBSCRIPTION: Cached detection has sessionId: {initial_detection_data.get('sessionId', 'None')}")
            else:
                # Send "none" detection for new subscriptions
                initial_detection_data = {
                    "type": "imageDetection",
                    "subscriptionIdentifier": subscriptionIdentifier,
                    "timestamp": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
                    "data": {
                        "detection": None,
                        "modelId": modelId,
                        "modelName": modelName
                    }
                }
                logger.info(f"📡 NEW SUBSCRIPTION: Sending initial 'none' detection for camera {camera_id}")
            
            ws_logger.info(f"TX -> {json.dumps(initial_detection_data, separators=(',', ':'))}")
            await websocket.send_json(initial_detection_data)
            logger.debug(f"Initial detection data sent (resubscription={cached_detection is not None}): {initial_detection_data}")
            
            # This cached detection was just a one-time status update for resubscription
            # Normal frame processing will continue independently

    async def unsubscribe_internal(subscription_id, preserve_detection=False):
        """Internal unsubscription logic"""
        if subscription_id in streams:
            stream = streams.pop(subscription_id)
            camera_url = subscription_to_camera.pop(subscription_id, None)
            
            # Clean up model references for this camera
            with models_lock:
                if subscription_id in models:
                    camera_models = models[subscription_id]
                    for model_id, model_tree in camera_models.items():
                        logger.info(f"🧹 Cleaning up model references for camera {subscription_id}, modelId {model_id}")
                        # Release model registry references
                        cleanup_pipeline_node(model_tree)
                        # Release MPTA manager reference
                        release_mpta(model_id, subscription_id)
                    del models[subscription_id]
            
            if camera_url and camera_url in camera_streams:
                shared_stream = camera_streams[camera_url]
                shared_stream["ref_count"] -= 1
                
                if shared_stream["ref_count"] <= 0:
                    shared_stream["stop_event"].set()
                    shared_stream["thread"].join()
                    if "cap" in shared_stream:
                        shared_stream["cap"].release()
                    del camera_streams[camera_url]
            
            latest_frames.pop(subscription_id, None)
            if not preserve_detection:
                cached_detections.pop(subscription_id, None)  # Clear cached detection only if not preserving
            frame_skip_flags.pop(subscription_id, None)   # Clear frame skip flag
            camera_states.pop(subscription_id, None)     # Clear camera state
            cached_full_pipeline_results.pop(subscription_id, None)  # Clear cached pipeline results
            session_pipeline_states.pop(subscription_id, None)  # Clear session pipeline state
            cleanup_camera_stability(subscription_id)
            logger.info(f"Unsubscribed from camera {subscription_id} (preserve_detection={preserve_detection})")

    async def process_streams():
        logger.info("Started processing streams")
        try:
            while True:
                start_time = time.time()
                with streams_lock:
                    current_streams = list(streams.items())
                    if current_streams:
                        logger.debug(f"Processing {len(current_streams)} active streams")
                    else:
                        logger.debug("No active streams to process")
                
                for camera_id, stream in current_streams:
                    buffer = stream["buffer"]
                    if buffer.empty():
                        logger.debug(f"Frame buffer is empty for camera {camera_id}")
                        continue
                    
                    logger.debug(f"Got frame from buffer for camera {camera_id}")
                    frame = buffer.get()
                    
                    # Cache the frame for REST API access
                    latest_frames[camera_id] = frame.copy()
                    logger.debug(f"Cached frame for REST API access for camera {camera_id}")
                    
                    with models_lock:
                        model_tree = models.get(camera_id, {}).get(stream["modelId"])
                        if not model_tree:
                            logger.warning(f"Model not found for camera {camera_id}, modelId {stream['modelId']}")
                            continue
                        logger.debug(f"Found model tree for camera {camera_id}, modelId {stream['modelId']}")
                    
                    key = (camera_id, stream["modelId"])
                    persistent_data = persistent_data_dict.get(key, {})
                    logger.debug(f"Starting detection for camera {camera_id} with modelId {stream['modelId']}")
                    updated_persistent_data = await handle_detection(
                        camera_id, stream, frame, websocket, model_tree, persistent_data
                    )
                    persistent_data_dict[key] = updated_persistent_data
                
                elapsed_time = (time.time() - start_time) * 1000  # ms
                sleep_time = max(poll_interval - elapsed_time, 0)
                logger.debug(f"Frame processing cycle: {elapsed_time:.2f}ms, sleeping for: {sleep_time:.2f}ms")
                await asyncio.sleep(sleep_time / 1000.0)
        except asyncio.CancelledError:
            logger.info("Stream processing task cancelled")
        except Exception as e:
            logger.error(f"Error in process_streams: {str(e)}", exc_info=True)

    async def send_heartbeat():
        while True:
            try:
                cpu_usage = psutil.cpu_percent()
                memory_usage = psutil.virtual_memory().percent
                if torch.cuda.is_available():
                    gpu_usage = torch.cuda.utilization() if hasattr(torch.cuda, 'utilization') else None
                    gpu_memory_usage = torch.cuda.memory_reserved() / (1024 ** 2)
                else:
                    gpu_usage = None
                    gpu_memory_usage = None

                camera_connections = [
                    {
                        "subscriptionIdentifier": stream["subscriptionIdentifier"],
                        "modelId": stream["modelId"],
                        "modelName": stream["modelName"],
                        "online": True,
                        # Include all subscription parameters for proper change detection
                        "rtspUrl": stream.get("rtsp_url"),
                        "snapshotUrl": stream.get("snapshot_url"),
                        "snapshotInterval": stream.get("snapshot_interval"),
                        **{k: v for k, v in get_crop_coords(stream).items() if v is not None}
                    }
                    for camera_id, stream in streams.items()
                ]

                state_report = {
                    "type": "stateReport",
                    "cpuUsage": cpu_usage,
                    "memoryUsage": memory_usage,
                    "gpuUsage": gpu_usage,
                    "gpuMemoryUsage": gpu_memory_usage,
                    "cameraConnections": camera_connections
                }
                await websocket.send_text(json.dumps(state_report))
                logger.debug(f"Sent stateReport as heartbeat: CPU {cpu_usage:.1f}%, Memory {memory_usage:.1f}%, {len(camera_connections)} active cameras")
                await asyncio.sleep(HEARTBEAT_INTERVAL)
            except Exception as e:
                logger.error(f"Error sending stateReport heartbeat: {e}")
                break

    async def on_message():
        while True:
            try:
                msg = await websocket.receive_text()
                ws_logger.info(f"RX <- {msg}")
                logger.debug(f"Received message: {msg}")
                data = json.loads(msg)
                msg_type = data.get("type")

                if msg_type == "setSubscriptionList":
                    # Declarative approach: Backend sends list of subscriptions this worker should have
                    desired_subscriptions = data.get("subscriptions", [])
                    logger.info(f"Received subscription list with {len(desired_subscriptions)} subscriptions")
                    
                    await reconcile_subscriptions(desired_subscriptions, websocket)
                    
                elif msg_type == "subscribe":
                    # Legacy support - convert single subscription to list
                    payload = data.get("payload", {})
                    await reconcile_subscriptions([payload], websocket)
                    
                elif msg_type == "unsubscribe":
                    # Legacy support - remove subscription
                    payload = data.get("payload", {})
                    subscriptionIdentifier = payload.get("subscriptionIdentifier")
                    # Remove from current subscriptions and reconcile
                    current_subs = []
                    with streams_lock:
                        for camera_id, stream in streams.items():
                            if stream["subscriptionIdentifier"] != subscriptionIdentifier:
                                # Convert stream back to subscription format
                                current_subs.append({
                                    "subscriptionIdentifier": stream["subscriptionIdentifier"],
                                    "rtspUrl": stream.get("rtsp_url"),
                                    "snapshotUrl": stream.get("snapshot_url"), 
                                    "snapshotInterval": stream.get("snapshot_interval"),
                                    "modelId": stream["modelId"],
                                    "modelName": stream["modelName"],
                                    "modelUrl": stream.get("modelUrl", ""),
                                    "cropX1": stream.get("cropX1"),
                                    "cropY1": stream.get("cropY1"),
                                    "cropX2": stream.get("cropX2"),
                                    "cropY2": stream.get("cropY2")
                                })
                    await reconcile_subscriptions(current_subs, websocket)
                    
                elif msg_type == "unsubscribe":
                    payload = data.get("payload", {})
                    subscriptionIdentifier = payload.get("subscriptionIdentifier")
                    camera_id = subscriptionIdentifier
                    with streams_lock:
                        if camera_id and camera_id in streams:
                            stream = streams.pop(camera_id)
                            camera_url = subscription_to_camera.pop(camera_id, None)
                            
                            if camera_url and camera_url in camera_streams:
                                shared_stream = camera_streams[camera_url]
                                shared_stream["ref_count"] -= 1
                                
                                # If no more references, stop the shared stream
                                if shared_stream["ref_count"] <= 0:
                                    logger.info(f"Stopping shared stream for camera URL: {camera_url}")
                                    shared_stream["stop_event"].set()
                                    shared_stream["thread"].join()
                                    if "cap" in shared_stream:
                                        shared_stream["cap"].release()
                                    del camera_streams[camera_url]
                                else:
                                    logger.info(f"Shared stream for {camera_url} still has {shared_stream['ref_count']} references")
                            
                            # Clean up cached frame and stability tracking
                            latest_frames.pop(camera_id, None)
                            cached_detections.pop(camera_id, None)  # Clear cached detection
                            frame_skip_flags.pop(camera_id, None)   # Clear frame skip flag
                            camera_states.pop(camera_id, None)     # Clear camera state
                            cleanup_camera_stability(camera_id)
                            logger.info(f"Unsubscribed from camera {camera_id}")
                            # Note: Keep models in memory for potential reuse
                elif msg_type == "requestState":
                    cpu_usage = psutil.cpu_percent()
                    memory_usage = psutil.virtual_memory().percent
                    if torch.cuda.is_available():
                        gpu_usage = torch.cuda.utilization() if hasattr(torch.cuda, 'utilization') else None
                        gpu_memory_usage = torch.cuda.memory_reserved() / (1024 ** 2)
                    else:
                        gpu_usage = None
                        gpu_memory_usage = None

                    camera_connections = [
                        {
                            "subscriptionIdentifier": stream["subscriptionIdentifier"],
                            "modelId": stream["modelId"],
                            "modelName": stream["modelName"],
                            "online": True,
                            # Include all subscription parameters for proper change detection
                            "rtspUrl": stream.get("rtsp_url"),
                            "snapshotUrl": stream.get("snapshot_url"),
                            "snapshotInterval": stream.get("snapshot_interval"),
                            **{k: v for k, v in get_crop_coords(stream).items() if v is not None}
                        }
                        for camera_id, stream in streams.items()
                    ]

                    state_report = {
                        "type": "stateReport",
                        "cpuUsage": cpu_usage,
                        "memoryUsage": memory_usage,
                        "gpuUsage": gpu_usage,
                        "gpuMemoryUsage": gpu_memory_usage,
                        "cameraConnections": camera_connections
                    }
                    await websocket.send_text(json.dumps(state_report))
                
                elif msg_type == "setSessionId":
                    payload = data.get("payload", {})
                    display_identifier = payload.get("displayIdentifier")
                    session_id = payload.get("sessionId")
                    
                    # Debug sessionId value types and contents
                    session_id_type = type(session_id).__name__
                    if session_id is None:
                        logger.info(f"🆔 BACKEND SESSIONID RECEIVED: displayId={display_identifier}, sessionId=None (type: {session_id_type})")
                        logger.info(f"🔄 BACKEND WANTS TO CLEAR SESSION for display {display_identifier}")
                    elif session_id == "null":
                        logger.info(f"🆔 BACKEND SESSIONID RECEIVED: displayId={display_identifier}, sessionId='null' (type: {session_id_type})")
                        logger.info(f"🔄 BACKEND SENT STRING 'null' for display {display_identifier}")
                    elif session_id == "":
                        logger.info(f"🆔 BACKEND SESSIONID RECEIVED: displayId={display_identifier}, sessionId='' (empty string, type: {session_id_type})")
                        logger.info(f"🔄 BACKEND SENT EMPTY STRING for display {display_identifier}")
                    else:
                        logger.info(f"🆔 BACKEND SESSIONID RECEIVED: displayId={display_identifier}, sessionId='{session_id}' (type: {session_id_type}, length: {len(str(session_id))})")
                        logger.info(f"🔄 BACKEND CREATED/UPDATED SESSION: {session_id} for display {display_identifier}")
                    
                    logger.debug(f"Full setSessionId payload: {payload}")
                    logger.debug(f"WebSocket message raw data: {json.dumps(data, indent=2)}")
                    logger.debug(f"Current active cameras: {list(streams.keys())}")
                    
                    if display_identifier:
                        # Store session ID for this display
                        if session_id is None or session_id == "null" or session_id == "":
                            old_session_id = session_ids.get(display_identifier)
                            session_ids.pop(display_identifier, None)
                            
                            if session_id is None:
                                logger.info(f"🚫 BACKEND ENDED SESSION: Cleared session ID for display {display_identifier} (was: {old_session_id}) - received None")
                            elif session_id == "null":
                                logger.info(f"🚫 BACKEND ENDED SESSION: Cleared session ID for display {display_identifier} (was: {old_session_id}) - received string 'null'")
                            elif session_id == "":
                                logger.info(f"🚫 BACKEND ENDED SESSION: Cleared session ID for display {display_identifier} (was: {old_session_id}) - received empty string")
                            
                            logger.debug(f"Session IDs after clearing: {session_ids}")
                            
                            # Reset tracking state for all cameras associated with this display
                            with streams_lock:
                                affected_cameras = []
                                for camera_id, stream in streams.items():
                                    if stream["subscriptionIdentifier"].startswith(display_identifier + ";"):
                                        affected_cameras.append(camera_id)
                                        # Import here to avoid circular import
                                        from siwatsystem.pympta import reset_tracking_state
                                        model_id = stream.get("modelId", "unknown")
                                        reset_tracking_state(camera_id, model_id, "backend session ended")
                                        
                                        
                                        logger.info(f"Reset tracking for camera {camera_id} (display: {display_identifier})")
                                logger.debug(f"Reset tracking for {len(affected_cameras)} cameras: {affected_cameras}")
                        else:
                            old_session_id = session_ids.get(display_identifier)
                            session_ids[display_identifier] = session_id
                            logger.info(f"✅ BACKEND SESSION STARTED: Set session ID {session_id} for display {display_identifier} (previous: {old_session_id})")
                            logger.debug(f"Session IDs after update: {session_ids}")
                            logger.debug(f"🎯 CMS Backend created sessionId {session_id} after receiving detection data")
                            
                            # 🔑 LPR Integration: Retroactively cache the last detection by this new session_id
                            session_id_str = str(session_id)
                            logger.info(f"🔑 LPR: Attempting to retroactively cache detection for session_id {session_id_str}")
                            
                            # Find cameras associated with this display
                            display_cameras = []
                            with streams_lock:
                                for camera_id, stream in streams.items():
                                    if stream["subscriptionIdentifier"].startswith(display_identifier + ";"):
                                        display_cameras.append(camera_id)
                            
                            logger.debug(f"🔍 Found {len(display_cameras)} cameras for display {display_identifier}: {display_cameras}")
                            
                            # Cache the most recent detection for each camera by the new session_id
                            cached_count = 0
                            for camera_id in display_cameras:
                                if camera_id in cached_detections:
                                    detection_data = cached_detections[camera_id].copy()
                                    
                                    # Add sessionId to the detection data
                                    detection_data['sessionId'] = session_id
                                    
                                    # Cache by session_id for LPR lookup
                                    session_detections[session_id_str] = detection_data
                                    session_to_camera[session_id_str] = camera_id
                                    detection_timestamps[session_id_str] = time.time()
                                    cached_count += 1
                                    
                                    logger.info(f"✅ LPR: Cached detection for session_id {session_id_str} -> camera {camera_id}")
                                    logger.debug(f"🔍 Detection data: {detection_data.get('data', {}).get('detection', {})}")
                                else:
                                    logger.debug(f"⚠️ No cached detection available for camera {camera_id}")
                            
                            if cached_count > 0:
                                logger.info(f"🎉 LPR: Successfully cached {cached_count} detection(s) for session_id {session_id_str}")
                                logger.info(f"📊 Total LPR sessions now cached: {len(session_detections)}")
                            else:
                                logger.warning(f"⚠️ LPR: No detections could be cached for session_id {session_id_str}")
                                logger.warning(f"   Display cameras: {display_cameras}")
                                logger.warning(f"   Available cached detections: {list(cached_detections.keys())}")
                            
                            # Clear waiting state for cameras associated with this display
                            with streams_lock:
                                affected_cameras = []
                                for camera_id, stream in streams.items():
                                    if stream["subscriptionIdentifier"].startswith(display_identifier + ";"):
                                        affected_cameras.append(camera_id)
                                        from siwatsystem.pympta import get_camera_stability_data
                                        model_id = stream.get("modelId", "unknown")
                                        stability_data = get_camera_stability_data(camera_id, model_id)
                                        session_state = stability_data["session_state"]
                                        if session_state.get("waiting_for_backend_session", False):
                                            session_state["waiting_for_backend_session"] = False
                                            session_state["wait_start_time"] = 0.0
                                            logger.info(f"🚀 PIPELINE UNBLOCKED: Backend sessionId {session_id} received - camera {camera_id} can proceed with database operations")
                                        logger.debug(f"📋 Camera {camera_id}: SessionId {session_id} now available for future database operations")
                                logger.debug(f"Updated session state for {len(affected_cameras)} cameras: {affected_cameras}")
                    else:
                        logger.warning(f"🚨 Invalid setSessionId message: missing displayIdentifier in payload")
                
                elif msg_type == "patchSession":
                    session_id = data.get("sessionId")
                    patch_data = data.get("data", {})
                    
                    # For now, just acknowledge the patch - actual implementation depends on backend requirements
                    response = {
                        "type": "patchSessionResult",
                        "payload": {
                            "sessionId": session_id,
                            "success": True,
                            "message": "Session patch acknowledged"
                        }
                    }
                    ws_logger.info(f"TX -> {json.dumps(response, separators=(',', ':'))}")
                    await websocket.send_json(response)
                    logger.info(f"Acknowledged patch for session {session_id}")
                
                elif msg_type == "setProgressionStage":
                    payload = data.get("payload", {})
                    display_identifier = payload.get("displayIdentifier")
                    progression_stage = payload.get("progressionStage")
                    
                    logger.info(f"🏁 PROGRESSION STAGE RECEIVED: displayId={display_identifier}, stage={progression_stage}")
                    
                    if display_identifier:
                        # Find all cameras associated with this display
                        with streams_lock:
                            affected_cameras = []
                            for camera_id, stream in streams.items():
                                if stream["subscriptionIdentifier"].startswith(display_identifier + ";"):
                                    affected_cameras.append(camera_id)
                            
                            logger.debug(f"🎯 Found {len(affected_cameras)} cameras for display {display_identifier}: {affected_cameras}")
                            
                            # Handle different progression stages
                            for camera_id in affected_cameras:
                                pipeline_state = get_or_init_session_pipeline_state(camera_id)
                                current_mode = pipeline_state.get("mode", "validation_detecting")
                                
                                if progression_stage == "car_fueling":
                                    # Situation 2: Stop YOLO inference, continue sending cached detection dict
                                    if current_mode == "lightweight":
                                        pipeline_state["yolo_inference_enabled"] = False
                                        pipeline_state["progression_stage"] = "car_fueling"
                                        logger.info(f"⏸️ Camera {camera_id}: YOLO inference DISABLED for car_fueling stage (still sending cached detection dict)")
                                    else:
                                        logger.debug(f"📊 Camera {camera_id}: car_fueling received but not in lightweight mode (mode: {current_mode})")
                                        
                                elif progression_stage == "car_waitpayment":
                                    # Resume YOLO inference for absence counter
                                    pipeline_state["yolo_inference_enabled"] = True
                                    pipeline_state["progression_stage"] = "car_waitpayment"
                                    logger.info(f"▶️ Camera {camera_id}: YOLO inference RE-ENABLED for car_waitpayment stage")
                                    
                                elif progression_stage == "welcome":
                                    # Ignore welcome messages during car_waitpayment as per requirement
                                    current_progression = pipeline_state.get("progression_stage")
                                    if current_progression == "car_waitpayment":
                                        logger.info(f"🚫 Camera {camera_id}: IGNORING welcome stage (currently in car_waitpayment)")
                                    else:
                                        pipeline_state["progression_stage"] = "welcome"
                                        logger.info(f"🎉 Camera {camera_id}: Progression stage set to welcome")
                                        
                                elif progression_stage in ["car_wait_staff"]:
                                    pipeline_state["progression_stage"] = progression_stage
                                    logger.info(f"📋 Camera {camera_id}: Progression stage set to {progression_stage}")
                    else:
                        logger.warning(f"🚨 Invalid setProgressionStage message: missing displayIdentifier in payload")
                
                else:
                    logger.error(f"Unknown message type: {msg_type}")
            except json.JSONDecodeError:
                logger.error("Received invalid JSON message")
            except (WebSocketDisconnect, ConnectionClosedError) as e:
                logger.warning(f"WebSocket disconnected: {e}")
                break
            except Exception as e:
                logger.error(f"Error handling message: {e}")
                break
    try:
        await websocket.accept()
        stream_task = asyncio.create_task(process_streams())
        heartbeat_task = asyncio.create_task(send_heartbeat())
        message_task = asyncio.create_task(on_message())
        await asyncio.gather(heartbeat_task, message_task)
    except Exception as e:
        logger.error(f"Error in detect websocket: {e}")
    finally:
        stream_task.cancel()
        await stream_task
        with streams_lock:
            # Clean up shared camera streams
            for camera_url, shared_stream in camera_streams.items():
                shared_stream["stop_event"].set()
                shared_stream["thread"].join()
                if "cap" in shared_stream:
                    shared_stream["cap"].release()
                while not shared_stream["buffer"].empty():
                    try:
                        shared_stream["buffer"].get_nowait()
                    except queue.Empty:
                        pass
                logger.info(f"Released shared camera stream for {camera_url}")
            
            streams.clear()
            camera_streams.clear()
            subscription_to_camera.clear()
        with models_lock:
            # Clean up all model references before clearing models dict
            for camera_id, camera_models in models.items():
                for model_id, model_tree in camera_models.items():
                    logger.info(f"🧹 Shutdown cleanup: Releasing model {model_id} for camera {camera_id}")
                    # Release model registry references
                    cleanup_pipeline_node(model_tree)
                    # Release MPTA manager reference
                    release_mpta(model_id, camera_id)
            models.clear()
            
        # Clean up the entire model registry and MPTA manager
        # logger.info("🏭 Performing final model registry cleanup...")
        # cleanup_registry()
        # logger.info("🏭 Performing final MPTA manager cleanup...")
        # cleanup_mpta_manager()
        
        latest_frames.clear()
        cached_detections.clear()
        frame_skip_flags.clear()
        camera_states.clear()
        cached_full_pipeline_results.clear()
        session_pipeline_states.clear()
        session_ids.clear()
        # Clean up LPR integration caches
        session_detections.clear()
        session_to_camera.clear()
        detection_timestamps.clear()
        logger.info("WebSocket connection closed")