trade_analysis/enhanced_api.py

# trade_analysis/enhanced_api.py

import os
from fastapi import FastAPI, Query, HTTPException
from pydantic import BaseModel
import httpx
from typing import Dict, Any, List
import pandas as pd
import numpy as np
import asyncio
from datetime import datetime
from pathlib import Path

# Import only modules that still exist
from .data import UnifiedDataProvider
from .indicators import enrich_with_indicators, identify_current_setup
from .enhanced_sentiment import EnhancedFinancialSentimentAnalyzer, analyze_momentum_sentiment
from .momentum_trading_engine import IntegratedMomentumEngine
from .enhanced_llm import EnhancedLLMEngine, generate_enhanced_llm_signal
from .tft_model import GapPredictionTFT
from .agent import TradingAgent, analyze_agent_performance

# Global dictionary to store TFT models
api_tft_models = {} 
trading_agent = None

def sanitize_for_json(data: any) -> any:
    """Recursively converts numpy and pandas types to JSON-serializable types."""
    if isinstance(data, dict):
        return {key: sanitize_for_json(value) for key, value in data.items()}
    elif isinstance(data, list):
        return [sanitize_for_json(item) for item in data]
    elif isinstance(data, np.bool_):
        return bool(data)
    elif isinstance(data, (np.integer, np.int64)):
        return int(data)
    elif isinstance(data, np.floating):
        return float(data)
    elif isinstance(data, pd.Timestamp):
        return data.isoformat()
    elif isinstance(data, (pd.Series, pd.Index, np.ndarray)):
        return data.tolist()
    return data

class EnhancedSignalResponse(BaseModel):
    """Enhanced response model with momentum and LLM analysis"""
    symbol: str
    signal: str
    confidence: float
    reasoning: str
    position_size: float
    status: str
    details: Dict[str, Any]
    
    # Enhanced fields
    momentum_analysis: Dict[str, Any] = {}
    llm_ensemble: Dict[str, Any] = {}
    options_strategy: Dict[str, Any] = {}
    timeframe_recommendation: str = "15m"
    expected_hold_time: str = "Unknown"

# Enhanced FastAPI App
app = FastAPI(
    title="Enhanced Intraday Momentum Engine", 
    version="2.0.0",
    description="SOTA Financial AI with multi-LLM ensemble and momentum analysis"
)

# Initialize enhanced components
data_provider = UnifiedDataProvider()
sentiment_analyzer = EnhancedFinancialSentimentAnalyzer()
momentum_engine = IntegratedMomentumEngine()
llm_engine = EnhancedLLMEngine()
tft_predictor = GapPredictionTFT(context_length=96, prediction_length=1)

@app.on_event("startup")
async def startup_event():
    """Initialize all AI models on startup and launch the agent."""
    print("🚀 Starting Enhanced Trading Engine...")

    # --- This new logic checks the environment before loading models ---
    from .deploy import DeploymentConfig
    config = DeploymentConfig.auto_detect()

    # Load sentiment models regardless of environment
    print("📊 Loading sentiment models...")
    sentiment_analyzer.initialize_models()

    # Only load LLMs if we are NOT on a CPU
    if config.device != "cpu":
        print("🧠 Loading LLM ensemble...")
        llm_engine.initialize_llm_models()
    else:
        print("🚫 CPU environment detected. Skipping LLM loading.")

    # Load TFT models
    print("🤖 Loading TFT models...")
    # (Your existing TFT model loading logic here, ensure it writes to /tmp if needed)
    symbols = ['QQQ', 'SPY', 'MSFT', 'TSLA', 'NVDA', 'META']
    for symbol in symbols:
        model_path = f"/tmp/tft_{symbol}_validated.pth" # Use /tmp for models
        tft_instance = GapPredictionTFT()
        # (The rest of your TFT loading logic...)
        api_tft_models[symbol] = tft_instance

    # Initialize and run the agent as a background task
    global trading_agent
    trading_agent = TradingAgent(api_url="http://localhost:7860")
    print("🤖 Launching Trading Agent as a background task...")
    asyncio.create_task(trading_agent.run())

    print("✅ Enhanced Trading Engine startup complete!")

@app.get("/")
def read_root():
    """Enhanced root endpoint with system info"""
    import torch
    
    gpu_info = "CPU only"
    if torch.cuda.is_available():
        gpu_name = torch.cuda.get_device_name(0)
        gpu_memory = torch.cuda.get_device_properties(0).total_memory / 1e9
        gpu_info = f"{gpu_name} ({gpu_memory:.1f} GB)"
    
    return {
        "status": "operational",
        "engine": "Enhanced Intraday Momentum Engine v2.0.0",
        "gpu_info": gpu_info,
        "features": [
            "Multi-LLM Ensemble Analysis",
            "Advanced Sentiment Analysis (10+ models)",
            "High-Frequency Momentum Engine", 
            "Options Strategy Generation",
            "TFT Gap Prediction",
            "Autonomous Trading Agent"
        ],
        "timestamp": datetime.now().isoformat()
    }

@app.post("/predict/enhanced/", response_model=EnhancedSignalResponse)
async def predict_enhanced_signal(
    symbol: str = Query(..., description="Stock symbol (e.g., QQQ, SPY)"),
    timeframe: str = Query("5m", description="Trading timeframe: 1m, 5m, 15m, 1h"),
    strategy_mode: str = Query("momentum", description="Strategy: momentum, scalp, gap, swing")
):
    """
    Enhanced prediction endpoint with full AI stack
    """
    try:
        start_time = datetime.now()
        
        # Fetch market data
        async with httpx.AsyncClient() as client:
            print(f"📈 Fetching data for {symbol}...")
            
            # Multi-timeframe OHLCV data
            ohlcv_data = await data_provider.fetch_multi_timeframe_stock_data(symbol)
            
            # News and social data - FIXED SYNTAX
            news_data, _ = await data_provider.fetch_news(symbol, client)
            reddit_data, _ = await data_provider.fetch_reddit_data(symbol)
            
            # Alternative data
            alt_data = data_provider.get_alternative_data(symbol)
        
        # Process dataframes
        news_df = pd.DataFrame(news_data) if news_data else pd.DataFrame()
        reddit_df = pd.DataFrame(reddit_data) if reddit_data else pd.DataFrame()
        
        # Technical analysis for each timeframe
        tech_setups = {}
        for tf, df in ohlcv_data.items():
            if not df.empty:
                enriched_df = enrich_with_indicators(df.copy(), tf)
                tech_setups[tf] = identify_current_setup(enriched_df, tf)
        
        print("🔄 Running AI analysis...")
        
        # 1. Enhanced Sentiment Analysis
        sentiment_analysis = await asyncio.get_event_loop().run_in_executor(
            None, 
            analyze_momentum_sentiment,
            news_df, reddit_df, symbol, timeframe
        )
        
        # 2. Momentum Analysis
        momentum_analysis = momentum_engine.generate_enhanced_signal(
            ohlcv_data, sentiment_analysis, alt_data
        )
        
        # 3. TFT Prediction
        daily_df = ohlcv_data.get("daily")
        tft_prediction = None
        tft_model = api_tft_models.get(symbol.upper())
        
        if daily_df is not None and len(daily_df) >= 96 and tft_model:
            if tft_model.is_trained:
                tft_prediction = tft_model.predict_gap_probability(daily_df)
                print(f"🚀 Using pretrained TFT model for {symbol}")
            else:
                print(f"🤖 Training TFT model for {symbol}...")
                tft_model.train(daily_df, epochs=20)
                tft_prediction = tft_model.predict_gap_probability(daily_df)
        else:
            if tft_model:
                tft_prediction = tft_model._default_prediction()
            else:
                temp_tft = GapPredictionTFT()
                tft_prediction = temp_tft._default_prediction()
        
        # 4. LLM Ensemble Analysis
        llm_analysis = {}
        try:
            llm_analysis = llm_engine.generate_enhanced_trading_signal(
                ohlcv_data, sentiment_analysis, momentum_analysis, alt_data
            )
        except Exception as e:
            print(f"LLM analysis failed: {e}")
            conditions = {
                "is_vix_high": alt_data.get('vix_level', 0) > 25,
                "is_15m_rsi_bullish": tech_setups.get("15m", {}).get('rsi', 50) > 65,
                "is_15m_rsi_bearish": tech_setups.get("15m", {}).get('rsi', 50) < 35,
                "is_15m_volume_spike": tech_setups.get("15m", {}).get('volume_spike', False),
                "is_hourly_trend_bullish": tech_setups.get("hourly", {}).get('direction') == 'up',
                "is_hourly_trend_bearish": tech_setups.get("hourly", {}).get('direction') == 'down'
            }
            llm_analysis = generate_enhanced_llm_signal(conditions)
        
        # 5. Master Signal Generation - FIXED FUNCTION NAME
        master_signal = _generate_master_signal(
            momentum_analysis, llm_analysis, sentiment_analysis, tft_prediction,
            timeframe, strategy_mode
        )
        
        # 6. Options Strategy - FIXED FUNCTION NAME
        options_strategy = _generate_options_strategy(
            master_signal, momentum_analysis, alt_data, timeframe, strategy_mode
        )
        
        # Calculate processing time
        processing_time = (datetime.now() - start_time).total_seconds()
        
        # Prepare response
        sanitized_details = sanitize_for_json({
            "tech_setups": tech_setups,
            "sentiment": sentiment_analysis,
            "alternative_data": alt_data,
            "tft_prediction": tft_prediction,
            "processing_time_seconds": processing_time,
            "data_sources": {
                "news_articles": len(news_df),
                "social_posts": len(reddit_df),
                "timeframes_analyzed": list(ohlcv_data.keys())
            }
        })
        
        return EnhancedSignalResponse(
            symbol=symbol,
            signal=master_signal["signal"],
            confidence=master_signal["confidence"],
            reasoning=master_signal["reasoning"],
            position_size=master_signal["position_size"],
            status="Success",
            details=sanitized_details,
            momentum_analysis=sanitize_for_json(momentum_analysis),
            llm_ensemble=sanitize_for_json(llm_analysis),
            options_strategy=sanitize_for_json(options_strategy),
            timeframe_recommendation=master_signal.get("timeframe", timeframe),
            expected_hold_time=master_signal.get("hold_time", "Unknown")
        )
        
    except Exception as e:
        import traceback
        traceback.print_exc()
        raise HTTPException(status_code=500, detail=f"Enhanced analysis failed: {e}")

@app.get("/agent/start")
async def start_agent():
    """Start the autonomous agent"""
    if trading_agent:
        asyncio.create_task(trading_agent.run())
        return {"status": "Agent started"}
    return {"status": "Agent not initialized"}

@app.get("/agent/stats")
async def get_agent_stats():
    """Get agent's performance stats"""
    if trading_agent:
        return trading_agent.get_stats()
    return {"error": "Agent not initialized"}

@app.get("/agent/positions")
async def get_agent_positions():
    """Get current positions"""
    if trading_agent:
        return {"positions": trading_agent.positions}
    return {"positions": {}}

@app.get("/agent/analyze")
async def analyze_agent():
    """Analyze agent's performance"""
    try:
        analyze_agent_performance()
        return {"status": "Analysis complete - check console output"}
    except Exception as e:
        return {"error": str(e)}

def _generate_master_signal(momentum_analysis: Dict, llm_analysis: Dict, 
                          sentiment_analysis: Dict, tft_prediction: Dict,
                          timeframe: str = "15m", strategy_mode: str = "momentum") -> Dict:
    """Generate master trading signal from all analyses - FIXED VERSION"""
    
    # Extract signals
    momentum_signal = momentum_analysis.get("signal", "HOLD")
    momentum_confidence = momentum_analysis.get("confidence", 50)
    
    # Use the actual momentum analysis results
    momentum_master = momentum_analysis.get("momentum_analysis", {}).get("master_signal", {})
    momentum_strategy = momentum_master.get("strategy", "WAIT")
    
    llm_signal = llm_analysis.get("signal", "HOLD")
    sentiment_composite = sentiment_analysis.get("composite_score", 0)
    tft_direction = tft_prediction.get("expected_direction", "FLAT") if tft_prediction else "FLAT"
    
    # TIMEFRAME-SPECIFIC THRESHOLDS
    timeframe_configs = {
        "1m": {
            "threshold": 0.2,
            "min_confidence": 70,
            "hold_time": "1-2 minutes",
            "position_multiplier": 0.5
        },
        "5m": {
            "threshold": 0.25,
            "min_confidence": 65,
            "hold_time": "2-5 minutes",
            "position_multiplier": 0.7
        },
        "15m": {
            "threshold": 0.3,
            "min_confidence": 60,
            "hold_time": "10-30 minutes",
            "position_multiplier": 1.0
        },
        "1h": {
            "threshold": 0.35,
            "min_confidence": 55,
            "hold_time": "30-60 minutes",
            "position_multiplier": 1.2
        }
    }
    
    config = timeframe_configs.get(timeframe, timeframe_configs["15m"])
    
    # STRATEGY MODE ADJUSTMENTS
    if strategy_mode == "scalp":
        config["threshold"] *= 0.8
        config["hold_time"] = "1-3 minutes"
    elif strategy_mode == "gap" and tft_prediction:
        if tft_direction != "FLAT" and tft_prediction.get("gap_probability", 50) > 70:
            config["min_confidence"] -= 10
    
    # Calculate weighted score
    if momentum_strategy in ["AGGRESSIVE_SCALP", "STANDARD_MOMENTUM"]:
        weighted_score = momentum_master.get("conviction", 0)
        weighted_confidence = momentum_confidence
    else:
        weights = {
            "momentum": 0.4,
            "llm": 0.25,
            "sentiment": 0.2,
            "tft": 0.15
        }
        
        signal_scores = {}
        signal_scores["momentum"] = 1.0 if momentum_signal == "CALLS" else -1.0 if momentum_signal == "PUTS" else 0.0
        signal_scores["llm"] = 1.0 if llm_signal == "CALLS" else -1.0 if llm_signal == "PUTS" else 0.0
        signal_scores["sentiment"] = np.clip(sentiment_composite, -1, 1)
        signal_scores["tft"] = 0.7 if tft_direction == "UP" else -0.7 if tft_direction == "DOWN" else 0.0
        
        weighted_score = sum(signal_scores[k] * weights[k] for k in weights)
        weighted_confidence = (momentum_confidence * 0.4 + 
                              llm_analysis.get("conviction", 50) * 0.3 +
                              (sentiment_analysis.get("confidence", "LOW") == "HIGH") * 80 * 0.3)
    
    # Generate final signal
    if weighted_score > config["threshold"] and weighted_confidence > config["min_confidence"]:
        final_signal = "CALLS"
        position_size = min(0.5, (weighted_confidence / 100) * config["position_multiplier"])
    elif weighted_score < -config["threshold"] and weighted_confidence > config["min_confidence"]:
        final_signal = "PUTS"
        position_size = min(0.5, (weighted_confidence / 100) * config["position_multiplier"])
    else:
        final_signal = "HOLD"
        position_size = 0.0
        config["hold_time"] = "Wait for better setup"
    
    # Build reasoning
    reasoning = []
    reasoning.append(f"{strategy_mode.upper()} {timeframe}: {final_signal}")
    reasoning.append(f"Confidence: {weighted_confidence:.0f}%")
    
    if momentum_strategy != "WAIT":
        reasoning.append(f"Momentum: {momentum_strategy}")
    if abs(sentiment_composite) > 0.3:
        reasoning.append(f"Sentiment: {'Bullish' if sentiment_composite > 0 else 'Bearish'}")
    if tft_prediction and tft_prediction.get("gap_probability", 50) > 70:
        reasoning.append(f"Gap probability: {tft_prediction['gap_probability']:.0f}%")
    
    return {
        "signal": final_signal,
        "confidence": int(weighted_confidence),
        "reasoning": ". ".join(reasoning),
        "position_size": position_size,
        "timeframe": timeframe,
        "hold_time": config["hold_time"],
        "weighted_score": weighted_score,
        "strategy_mode": strategy_mode,
        "momentum_strategy": momentum_strategy
    }

def _generate_options_strategy(master_signal: Dict, momentum_analysis: Dict, 
                               alt_data: Dict, timeframe: str = "15m", 
                               strategy_mode: str = "momentum") -> Dict:
    """Generate options strategy with timeframe awareness"""
    
    signal = master_signal["signal"]
    confidence = master_signal["confidence"]
    vix_level = alt_data.get("vix_level", 20)
    
    if signal == "HOLD":
        return {
            "strategy": "WAIT",
            "reasoning": "No clear directional bias",
            "contracts": [],
            "risk_management": "Wait for better setup"
        }
    
    # TIMEFRAME-SPECIFIC STRATEGIES
    if timeframe in ["1m", "5m"] and strategy_mode == "scalp":
        strategy = {
            "strategy": "0DTE_SCALP",
            "reasoning": f"{timeframe} scalp: {signal} with {confidence}% confidence",
            "contracts": [
                {
                    "type": "CALL" if signal == "CALLS" else "PUT",
                    "strike": "ATM",
                    "quantity": min(int(confidence / 8), 15),
                    "dte": 0,
                    "target_profit": 20,
                    "stop_loss": 10
                }
            ],
            "max_hold_time": f"{timeframe} bars (max 5 minutes)",
            "risk_management": "Ultra-tight stops, quick exits"
        }
        
    elif timeframe == "15m" and confidence > 70:
        strategy = {
            "strategy": "MOMENTUM_15M",
            "reasoning": f"15-minute momentum {signal} play, {confidence}% confidence",
            "contracts": [
                {
                    "type": "CALL" if signal == "CALLS" else "PUT",  
                    "strike": "1% ITM",
                    "quantity": min(int(confidence / 12), 8),
                    "dte": 1,
                    "target_profit": 40,
                    "stop_loss": 20
                }
            ],
            "max_hold_time": "30 minutes",
            "risk_management": "Standard momentum stops"
        }
        
    elif timeframe == "1h":
        strategy = {
            "strategy": "HOURLY_SWING",
            "reasoning": f"Hourly swing {signal}, {confidence}% confidence",
            "contracts": [
                {
                    "type": "CALL_SPREAD" if signal == "CALLS" else "PUT_SPREAD",
                    "long_strike": "ATM",
                    "short_strike": "3% OTM",
                    "quantity": min(int(confidence / 15), 5),
                    "dte": 3,
                    "target_profit": 35,
                    "stop_loss": 25
                }
            ],
            "max_hold_time": "2-4 hours",
            "risk_management": "Defined risk spreads"
        }
        
    else:
        strategy = {
            "strategy": "CONSERVATIVE",
            "reasoning": f"Lower conviction {signal}, using conservative approach",
            "contracts": [
                {
                    "type": "CALL_SPREAD" if signal == "CALLS" else "PUT_SPREAD",
                    "long_strike": "ATM",
                    "short_strike": "5% OTM",
                    "quantity": 3,
                    "dte": 7,
                    "target_profit": 25,
                    "stop_loss": 20
                }
            ],
            "max_hold_time": "End of day",
            "risk_management": "Limited risk, defined reward"
        }
    
    # VIX adjustments
    if vix_level > 30:
        strategy["reasoning"] += f". High VIX ({vix_level}) - reduced size"
        for contract in strategy["contracts"]:
            contract["quantity"] = max(1, contract["quantity"] // 2)
    
    return strategy

@app.post("/backtest/enhanced/")
async def enhanced_backtest(
    symbol: str = Query(..., description="Stock symbol"),
    start_date: str = Query(..., description="Start date (YYYY-MM-DD)"),
    end_date: str = Query(..., description="End date (YYYY-MM-DD)"),
    strategy_mode: str = Query("momentum", description="Strategy mode"),
    initial_capital: float = Query(100000, description="Initial capital")
):
    """Enhanced backtesting with momentum strategies"""
    try:
        return {
            "status": "success",
            "message": "Enhanced backtesting ready",
            "features": [
                "Multi-timeframe momentum analysis",
                "LLM ensemble signal validation", 
                "Options strategy backtesting",
                "Risk-adjusted performance metrics",
                "Slippage and commission modeling"
            ]
        }
    except Exception as e:
        return {"status": "error", "message": str(e)}

@app.get("/health/detailed")
async def detailed_health_check():
    """Detailed system health check"""
    import torch
    
    health_status = {
        "timestamp": datetime.now().isoformat(),
        "overall_status": "healthy",
        "components": {}
    }
    
    # Check GPU
    if torch.cuda.is_available():
        gpu_memory_used = torch.cuda.memory_allocated(0) / 1e9
        gpu_memory_total = torch.cuda.get_device_properties(0).total_memory / 1e9
        health_status["components"]["gpu"] = {
            "status": "available",
            "device": torch.cuda.get_device_name(0),
            "memory_used_gb": gpu_memory_used,
            "memory_total_gb": gpu_memory_total,
            "utilization": f"{gpu_memory_used/gpu_memory_total*100:.1f}%"
        }
    else:
        health_status["components"]["gpu"] = {"status": "not_available"}
    
    # Check model status
    health_status["components"]["sentiment_models"] = {
        "loaded": len(sentiment_analyzer.models),
        "status": "ready" if sentiment_analyzer.models else "not_loaded"
    }
    
    health_status["components"]["llm_models"] = {
        "loaded": len(llm_engine.models),
        "status": "ready" if llm_engine.models else "not_loaded"
    }
    
    health_status["components"]["tft_model"] = {
        "status": "trained" if tft_predictor.is_trained else "not_trained"
    }
    
    health_status["components"]["agent"] = {
        "status": "initialized" if trading_agent else "not_initialized"
    }
    
    return health_status

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)