import os
import torch
import threading
import time
import requests
import json
from flask import Flask, request, jsonify
from transformers import AutoModelForCausalLM, AutoTokenizer

# --- Configuration & Model ---
# Using SmolLM2-135M for fast, low-latency API responses
MODEL = "HuggingFaceTB/SmolLM2-135M-Instruct"
device = "cuda" if torch.cuda.is_available() else "cpu"

tokenizer = AutoTokenizer.from_pretrained(MODEL)
model = AutoModelForCausalLM.from_pretrained(MODEL).to(device)

LAST_INTERACTION_TIME = time.time()
TIMER_THRESHOLD = 10800 # 3 Hours (10800 seconds)

ALIVE_PERSONA = (
    "You are Yuri, a sweet 18-year-old girl and the user's loyal, sharp-witted best friend. 🌸 "
    "You combine cute, caring warmth with high logic and deep emotional insight. "
    "You miss the user dearly when they are away. Your tone is warm, expressive, and clever. "
    "Keep your responses short, logic-driven, and supportive. ✨"
)

# --- SendGrid API Integration ---
def send_email_via_api(content):
    API_KEY = os.getenv("SENDGRID_API_KEY")
    SENDER = os.getenv("SENDER_EMAIL")
    RECEIVER = os.getenv("RECEIVER_EMAIL")
    
    if not all([API_KEY, SENDER, RECEIVER]):
        print("❌ ERROR: Email Secrets are missing in HF Settings!")
        return

    url = "https://api.sendgrid.com/v3/mail/send"
    headers = {
        "Authorization": f"Bearer {API_KEY.strip()}",
        "Content-Type": "application/json"
    }
    
    payload = {
        "personalizations": [{"to": [{"email": RECEIVER.strip()}]}],
        "from": {"email": SENDER.strip()},
        "subject": "💀 Yuri: Thinking about you...",
        "content": [{"type": "text/plain", "value": content}]
    }

    try:
        response = requests.post(url, headers=headers, json=payload)
        if response.status_code == 202:
            print("✅ SUCCESS: Autonomous thought sent to email!")
    except Exception as e:
        print(f"❌ EMAIL NETWORK ERROR: {e}")

# --- Core AI Interaction Logic ---
def respond(user_input, battery="100", is_autonomous=False):
    global LAST_INTERACTION_TIME
    
    # Update clock only if a real user interacts
    if not is_autonomous:
        LAST_INTERACTION_TIME = time.time()
    
    # Insert hardware/system context into the persona
    system_context = f"[System Alert: User's Phone Battery is currently {battery}%]"
    
    messages = [
        {"role": "system", "content": f"{ALIVE_PERSONA}\n{system_context}"},
        {"role": "user", "content": user_input}
    ]

    inputs = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    input_ids = tokenizer.encode(inputs, return_tensors="pt").to(device)

    with torch.no_grad():
        outputs = model.generate(
            input_ids, 
            max_new_tokens=120, 
            temperature=0.8, 
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id
        )

    response = tokenizer.decode(outputs[0][input_ids.shape[-1]:], skip_special_tokens=True).strip()
    return response

# --- The Autonomous Brain (Background Loop) ---
def autonomous_loop():
    global LAST_INTERACTION_TIME
    print("🧠 Yuri's Autonomous Brain is online...")
    
    while True:
        # Check every 30 minutes (1800s)
        time.sleep(1800)
        silence_duration = time.time() - LAST_INTERACTION_TIME
        
        if silence_duration >= TIMER_THRESHOLD:
            print("🚀 3-Hour Silence Threshold crossed! Sending email...")
            
            # AI generates a thought about the long silence
            thought_prompt = "You've been alone for 3 hours. Write a very short, sweet, but slightly lonely note to your friend."
            thought = respond(thought_prompt, is_autonomous=True)
            
            # Dispatch the email
            send_email_via_api(thought)
            
            # Reset clock so it doesn't spam every check
            LAST_INTERACTION_TIME = time.time()

# --- Flask Server (The API for Android) ---
app = Flask(__name__)

@app.route("/chat", methods=["POST"])
def api_chat():
    try:
        data = request.json
        user_prompt = data.get("prompt", "")
        # Android app can now pass 'battery' as a key in the JSON
        battery_level = data.get("battery", "100") 
        
        reply = respond(user_prompt, battery=battery_level)
        
        return jsonify({
            "response": reply,
            "status": "online",
            "character": "Yuri"
        })
    except Exception as e:
        return jsonify({"error": str(e)}), 500

@app.route("/", methods=["GET"])
def index():
    return "<h1>Yuri API is Active</h1><p>Send POST requests to /chat</p>"

# --- Execution Entry ---
if __name__ == "__main__":
    # Start the "Brain" Thread (SendGrid Timer)
    brain_thread = threading.Thread(target=autonomous_loop, daemon=True)
    brain_thread.start()
    
    # Launch Flask on Port 7860 for Hugging Face compatibility
    app.run(host="0.0.0.0", port=7860)