app.py

import os
import gc
import asyncio
import pandas as pd
import numpy as np
import pickle
from contextlib import asynccontextmanager
from fastapi import FastAPI
from fastapi.responses import HTMLResponse
from fastapi.staticfiles import StaticFiles
from fastapi.templating import Jinja2Templates
from starlette.requests import Request
import uvicorn

# --- 🛡️ SHIELDS ---
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"

# --- 🩹 Keras Monkey Patch ---
import tensorflow as tf
import keras
import sys
sys.modules['tensorflow.keras'] = keras
sys.modules['tensorflow.keras.layers'] = keras.layers
sys.modules['tensorflow.keras.models'] = keras.models

from deepface import DeepFace

# Custom Modules
from core.config import LOG_FILE, MODELS, DETECTOR
from core.state import attendance_memory, KNOWN_VECTORS
from api.websocket import websocket_endpoint

# --- Registration Service (Internal) ---
def auto_register_faces():
    """Generates embeddings.pickle on the fly if missing using direct 1-to-1 mapping."""
    pickle_path = "core/embeddings.pickle"
    db_path = "faces_db"
    
    if os.path.exists(pickle_path):
        return 
        
    if not os.path.exists(db_path) or not os.listdir(db_path):
        print("⚠️ faces_db is empty.")
        return

    print("🚀 Auto-Registering Faces on Cloud CPU...")
    identity_map = {}
    image_files = [f for f in os.listdir(db_path) if f.lower().endswith(('.jpg', '.jpeg', '.png'))]
    
    for filename in image_files:
        name = os.path.splitext(filename)[0]
        img_path = os.path.join(db_path, filename)
        print(f"📸 Cloud Encoding: {name}...")
        
        try:
            identity_map[name] = {}
            for model in MODELS:
                emb_objs = DeepFace.represent(img_path=img_path, model_name=model, 
                                             enforce_detection=True, detector_backend=DETECTOR)
                if emb_objs:
                    vector = np.array(emb_objs[0]["embedding"])
                    identity_map[name][model] = vector / np.linalg.norm(vector)
            print(f"✅ Encoded: {name}")
        except Exception as e:
            print(f"❌ Failed {name}: {e}")

    with open(pickle_path, "wb") as f:
        pickle.dump(identity_map, f)
    print(f"✨ Mathematical Brain Created Successfully!")

@asynccontextmanager
async def lifespan(app: FastAPI):
    # --- STARTUP PHASE ---
    
    # 1. Run Auto-Registration if needed
    auto_register_faces()
    
    # 2. Load the Hard Pickle into RAM
    pickle_path = "core/embeddings.pickle"
    if os.path.exists(pickle_path):
        try:
            with open(pickle_path, "rb") as f:
                data = pickle.load(f)
                KNOWN_VECTORS.update(data)
            print(f"🧠 Mathematical Brain Loaded: Found {len(KNOWN_VECTORS)} identities.")
        except Exception as e:
            print(f"⚠️ Pickle Load Error: {e}")
    else:
        print("⚠️ WARNING: core/embeddings.pickle not found. Run register_faces.py first!")

    # 3. Load historical logs into RAM
    if os.path.exists(LOG_FILE):
        try:
            df = pd.read_csv(LOG_FILE)
            for _, row in df.iterrows():
                attendance_memory.insert(0, {"name": row['Name'], "time": row['Time']})
            print(f"✅ Loaded {len(attendance_memory)} records.")
        except: pass
    else:
        with open(LOG_FILE, "w") as f: f.write("Name,Time\n")
    
    yield
    # --- SHUTDOWN PHASE ---
    pass

# --- AI Warmup Sequence ---
print("🚀 Booting Dynamic Consensus AI Core...")
try:
    for model in MODELS:
        DeepFace.build_model(model)
    # Note: RetinaFace is a detector, not a recognition model, so we don't build it via build_model()
    print(f"🧠 {', '.join(MODELS)} Networks Loaded.")
    gc.collect() 
except Exception as e:
    print(f"⚠️ AI Boot Error: {e}")

# --- FastAPI Application ---
app = FastAPI(title="Enterprise Attendance AI", version="5.0.0", lifespan=lifespan)

app.mount("/static", StaticFiles(directory="static"), name="static")
templates = Jinja2Templates(directory="templates")

# Register the websocket route
app.add_api_websocket_route("/ws", websocket_endpoint)

@app.get("/", response_class=HTMLResponse)
async def read_root(request: Request):
    """Serves the main application dashboard."""
    return templates.TemplateResponse("index.html", {"request": request})

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)