worker.py

import os
os.environ["PYTHONWARNINGS"] = "ignore"
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'

import warnings
import logging
import sys

warnings.filterwarnings("ignore")
logging.getLogger("asyncio").setLevel(logging.CRITICAL)
logging.getLogger("curl_cffi").setLevel(logging.CRITICAL) # ADD THIS LINE
import pandas as pd
pd.options.mode.chained_assignment = None

import threading
import time
import yfinance as yf
import numpy as np
from ddgs import DDGS # Updated to the silent package
from huggingface_hub import hf_hub_download
from llama_cpp import Llama
import storage
import json

TOP_N = 5
REGIME_FILTER_DAYS = 200
VOLATILITY_CAP = 0.60
PRICE_FLOOR = 10.0
MOMENTUM_LOOKBACK = 30 
CACHE_FILE = "fundamental_cache.json"

stop_event = threading.Event()
is_running = False

class MetanthropicBrain:
    def __init__(self):
        print("🧠 Booting Live AI (Qwen 1.5B FP16)...")
        try:
            model_path = hf_hub_download(
                repo_id="Qwen/Qwen2.5-1.5B-Instruct-GGUF",
                filename="qwen2.5-1.5b-instruct-fp16.gguf" 
            )
            cpu_threads = os.cpu_count() or 4
            self.llm = Llama(
                model_path=model_path,
                n_ctx=2048,      
                n_threads=cpu_threads,     
                verbose=False
            )
            print(f"✅ Live AI Online. Using {cpu_threads} CPU Cores.")
        except Exception as e:
            print(f"❌ AI Boot Failure: {e}")
            self.llm = None

    def analyze_news(self, ticker, news_context, math_score):
        if not self.llm: return 50, "AI Offline - Relied on Math"
        
        system_msg = """You are a ruthless Quantitative Analyst at a top hedge fund. 
        Your job is to read the latest news headlines for a stock and determine if there is a fundamental catalyst or a massive risk.
        Reply ONLY in this format:
        SCORE: [0 to 100]
        REASON: [One short sentence explaining why]"""
        
        user_msg = f"STOCK: {ticker}\nMATH MOMENTUM: {math_score}%\nLATEST GLOBAL NEWS: {news_context}"
        
        try:
            output = self.llm.create_chat_completion(
                messages=[{"role": "system", "content": system_msg}, {"role": "user", "content": user_msg}],
                max_tokens=60,
                temperature=0.3
            )
            response = output['choices'][0]['message']['content'].strip()
            
            score_line = [line for line in response.split('\n') if 'SCORE:' in line]
            reason_line = [line for line in response.split('\n') if 'REASON:' in line]
            
            score = int(score_line[0].replace('SCORE:', '').strip()) if score_line else 50
            reason = reason_line[0].replace('REASON:', '').strip() if reason_line else response
            
            return score, reason
        except:
            return 50, "AI Parsing Error"

class GlobalIntel:
    def get_news(self, ticker):
        try:
            clean_ticker = ticker.replace(".NS", "")
            query = f'"{clean_ticker}" stock market news india recent'
            
            with warnings.catch_warnings():
                warnings.simplefilter("ignore")
                with DDGS(timeout=10) as ddgs:  # ADD THE TIMEOUT HERE
                    results = list(ddgs.text(query, region='in-en', max_results=3))
            
            if not results: return "No major news recently. Operating purely on market mechanics."
            
            summary = " | ".join([r['title'] for r in results])
            return summary
        except:
            return "Internet search blocked or failed."

ai_brain = None

def get_live_recommendations():
    global ai_brain
    print("\n🦅 Initiating Live Alpha Scanner...")
    
    try:
        print("📊 Checking Global Market Regime...")
        nifty = yf.download("^NSEI", period="1y", progress=False)['Close']
        if isinstance(nifty, pd.DataFrame): nifty = nifty.squeeze()
        
        current_nifty = nifty.iloc[-1]
        nifty_200_sma = nifty.rolling(REGIME_FILTER_DAYS).mean().iloc[-1]
        
        if current_nifty < nifty_200_sma:
            print("🚨 CRASH DETECTED: Fleeing to Gold.")
            return [{
                "Ticker": "GOLD (GC=F)", "Score": 100, "Decision": "BUY (HEDGE)", 
                "Trend": "BEAR MARKET", "Key Metrics": f"Nifty < {nifty_200_sma:.0f}", 
                "AI Reasoning": "Macro regime is bearish. Capital Preservation Active."
            }]
    except:
        pass

    if not os.path.exists(CACHE_FILE):
        print("❌ Error: fundamental_cache.json not found. Rebuild it using the Data Tools in the UI.")
        return []
        
    with open(CACHE_FILE, 'r') as f:
        all_tickers = json.load(f)
    
    print(f"📡 Fetching live mathematical data for {len(all_tickers)} Elite stocks...")
    try:
        with warnings.catch_warnings():
            warnings.simplefilter("ignore")
            data = yf.download(all_tickers, period="3mo", progress=False)['Close']
            if isinstance(data.columns, pd.MultiIndex):
                try: data = data['Close']
                except: pass
                
            data = data.ffill().bfill().infer_objects()
        
        momentum = data.pct_change(MOMENTUM_LOOKBACK).iloc[-1]
        volatility = (data.pct_change(1).rolling(MOMENTUM_LOOKBACK).std() * np.sqrt(252)).iloc[-1]
        current_prices = data.iloc[-1]
        
        valid_stocks = momentum[
            (momentum > 0) & 
            (volatility < VOLATILITY_CAP) & 
            (current_prices > PRICE_FLOOR)
        ]
        
        math_winners = valid_stocks.sort_values(ascending=False).head(15)
        print(f"🧮 Math Engine found {len(math_winners)} raw candidates. Passing to AI for Macro vetting...")
        
    except Exception as e:
        print(f"❌ Error fetching math data: {e}")
        return []

    if ai_brain is None: ai_brain = MetanthropicBrain()
    intel = GlobalIntel()
    
    final_picks = []
    
    for ticker, mom in math_winners.items():
        print(f"   🔎 AI Investigating: {ticker} (+{mom*100:.1f}%)")
        news = intel.get_news(ticker)
        
        ai_score, ai_reason = ai_brain.analyze_news(ticker, news, round(mom*100, 1))
        blended_score = (mom * 100 * 0.5) + (ai_score * 0.5)
        
        final_picks.append({
            "Ticker": ticker,
            "Score": int(blended_score),
            "Decision": "BUY (APPROVED)" if ai_score >= 50 else "REJECTED BY AI",
            "Trend": f"+{mom*100:.1f}%",
            "Key Metrics": f"Price: ₹{current_prices[ticker]:.1f} | AI Confidence: {ai_score}/100",
            "AI Reasoning": ai_reason
        })

    approved_picks = [p for p in final_picks if p["Decision"] == "BUY (APPROVED)"]
    approved_picks.sort(key=lambda x: x['Score'], reverse=True)
    
    return approved_picks[:TOP_N]

def background_process():
    global is_running
    is_running = True
    print("🚀 LIVE PAPER TRADER STARTED")
    
    while not stop_event.is_set():
        picks = get_live_recommendations()
        
        if picks:
            storage.reset_database()
            for pick in picks:
                storage.save_result(pick)
                print(f"💎 LIVE PICK: {pick['Ticker']} -> AI Reason: {pick['AI Reasoning']}")
            
        for _ in range(21600):
            if stop_event.is_set(): break
            time.sleep(1)

def start_worker():
    global is_running
    if not is_running:
        stop_event.clear()
        t = threading.Thread(target=background_process, daemon=True)
        t.start()
        return "▶️ AI Paper Trader Active"
    return "⚠️ Already Running"

def stop_worker():
    stop_event.set()
    global is_running
    is_running = False
    return "🛑 Stopped"

def reset_worker():
    storage.reset_database()
    return "♻️ Database Wiped"