Spaces:

Karan6124
/

QuantIQ

Running

QuantIQ / backend /app /api /endpoints.py
fix: make model-metrics database queries individually fault-tolerant
abdc502 about 2 months ago
56 kB
import json
import asyncio
import hmac
import hashlib
import datetime
import smtplib
import random
from email.mime.text import MIMEText
from email.mime.multipart import MIMEMultipart
from typing import Optional
import jwt
import httpx
from fastapi import APIRouter, Depends, HTTPException, Request, Header
from sqlalchemy.ext.asyncio import AsyncSession

from backend.app.config.settings import settings
from backend.app.database.session import get_db
from backend.app.database import crud
from backend.app.schemas import schemas
from backend.app.config.metrics import payment_callbacks_total, external_api_calls_total
from backend.app.services.gemini import get_onnx_session_for_type, compute_chart_indicators

router= APIRouter()

# JWT TOKEN UTILITIES
def create_access_token(data: dict, expires_delta: Optional[datetime.timedelta]= None) -> str:
    """
    Generates a secure JSON Web Token (JWT) signed with the application SECRET_KEY.
    """
    to_encode= data.copy()
    if expires_delta:
        expire= datetime.datetime.now(datetime.timezone.utc) + expires_delta
    else:
        expire= datetime.datetime.now(datetime.timezone.utc) + datetime.timedelta(minutes= settings.ACCESS_TOKEN_EXPIRE_MINUTES)
    to_encode.update({"exp": expire})
    encode_jwt= jwt.encode(to_encode, settings.SECRET_KEY, algorithm= "HS256")
    return encode_jwt

def decode_access_token(token: str) -> Optional[dict]:
    """
    Decodes and validates a JWT token. Returns the payload dictionary or None if invalid.
    """
    try:
        payload= jwt.decode(token, settings.SECRET_KEY, algorithms= ["HS256"])
        return payload
    except jwt.PyJWTError:
        return None

from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
from fastapi import File, UploadFile
from backend.app.database import models
from backend.app.services.cloudinary_service import upload_avatar
import uuid

security= HTTPBearer()

async def get_current_user(credentials: HTTPAuthorizationCredentials= Depends(security), db: AsyncSession= Depends(get_db)) -> models.User:
    token= credentials.credentials
    payload= decode_access_token(token)
    if not payload or "sub" not in payload:
        raise HTTPException(status_code= 401, detail= "Invalid token or expired session.")
    try:
        user_id= uuid.UUID(payload["sub"])
    except ValueError:
        raise HTTPException(status_code= 401, detail= "Invalid user ID in token.")
    
    user= await crud.get_user(db, user_id)
    if not user:
        raise HTTPException(status_code=404, detail= "User not found.")
    return user

# REST ENDPOINTS
@router.post("/auth/google", response_model= schemas.Token)
async def google_auth(payload: schemas.GoogleAuthRequest, db: AsyncSession= Depends(get_db)):
    """
    Authenticates a user using their Google OAuth ID Token.
    Returns a custom JWT access token for subsequent API and GraphQL requests.
    """
    token_id= payload.token_id
    
    # 1. Dev Mode: Bypass Google verification if mock token is sent
    if token_id== "mock_token":
        email= "tester@quantiq.io"
        user= await crud.get_user_by_email(db, email)
        if not user:
            user_in= schemas.UserBase(
                email= email,
                full_name= "Local Tester",
                picture_url= "https://via.placeholder.com/150"
            )
            user= await crud.create_user(db, user_in, google_id= "mock_google_id_123")
        token= create_access_token(data={"sub": str(user.id)})
        return {"access_token": token, "token_type": "bearer"}
    
    # 2. Production Mode: Verify the ID Token via Google's tokeninfo API
    async with httpx.AsyncClient() as client:
        try:
            response= await client.get(f"https://oauth2.googleapis.com/tokeninfo?id_token={token_id}", timeout= 10.0)
            if response.status_code != 200:
                raise HTTPException(status_code= 400, detail= "Invalid Google OAuth token")
            
            idinfo= response.json()
            
            # Verify the audience matches our client ID if one is configured
            if settings.GOOGLE_CLIENT_ID and idinfo.get("aud") != settings.GOOGLE_CLIENT_ID:
                raise HTTPException(status_code= 400, detail= "OAuth audience mismatch")
            
            email= idinfo.get("email")
            name= idinfo.get("name")
            picture= idinfo.get("picture")
            google_id= idinfo.get("sub")
            
            if not email or not google_id:
                raise HTTPException(status_code= 400, detail= "Incomplete Google user profile")
            
        except httpx.RequestError as e:
            raise HTTPException(status_code= 503, detail= f"Google authentication unreachable: {str(e)}")
        
        # 3. Register or Retrieve User
        user= await crud.get_user_by_google_id(db, google_id)
        if not user:
            # Check if email is already in use by a mock account
            user= await crud.get_user_by_email(db, email)
            if not user:
                user_in= schemas.UserBase(email= email, full_name= name, picture_url= picture)
                user= await crud.create_user(db, user_in, google_id= google_id)
        
        # 4. Generate and return our custom JWT
        access_token= create_access_token(data= {"sub": str(user.id)})
        return {"access_token": access_token, "token_type": "bearer"}

# TRADITIONAL EMAIL AUTHENTICATION
from pydantic import BaseModel

class EmailSignUpRequest(BaseModel):
    email:str
    full_name: str
    country: str
    password: str

class EmailLoginRequest(BaseModel):
    email:str
    password: str

@router.post("/auth/signup", response_model= schemas.AuthResponse)
async def email_signup(payload: EmailSignUpRequest, db: AsyncSession= Depends(get_db)):
    """
    Registers a new user using their email, name, country, and password.
    Saves authentication credentials inside the existing schema.
    Generates a 6-digit OTP code, saves it to the database, sends the verification
    email via Celery, and indicates verification is required.
    """
    # 1. Check if email already in use
    existing_user= await crud.get_user_by_email(db, payload.email)
    
    # Store password hash in google_id (local:{hash}) so picture_url stays free for profile photos
    pwd_hash = hashlib.sha256(payload.password.encode("utf-8")).hexdigest()
    google_id = f"local:{pwd_hash}"
    
    if existing_user:
        if existing_user.is_verified:
            # If it's already a verified LOCAL account, block re-registration
            if existing_user.google_id.startswith("local:"):
                raise HTTPException(status_code= 400, detail= "Email is already registered.")
            else:
                # Verified Google account trying to register with email/password
                raise HTTPException(status_code= 400, detail= "This email is linked to a Google account. Please sign in with Google.")
        else:
            # Reusing the existing unverified user record (updating password/profile details)
            existing_user.full_name = payload.full_name
            existing_user.google_id = google_id  # update hash in google_id
            user = existing_user
    else:
        # Create a new unverified user record
        user_in= schemas.UserBase(
            email= payload.email,
            full_name= payload.full_name,
            picture_url= None  # no profile pic yet for email signup
        )
        try:
            user= await crud.create_user(db, user_in, google_id= google_id)
        except Exception as e:
            raise HTTPException(status_code= 500, detail=f"Failed to register user: {str(e)}")

    # 2. Generate a 6-digit verification code
    code = f"{random.randint(100000, 999999)}"
    expires_at = datetime.datetime.now(datetime.timezone.utc) + datetime.timedelta(minutes=15)
    
    user.is_verified = False
    user.verification_code = code
    user.verification_code_expires_at = expires_at
    
    db.add(user)
    await db.commit()
    await db.refresh(user)
    
    # 3. Trigger verification email via Celery
    try:
        from backend.app.services.celery_app import send_verification_email_task
        send_verification_email_task.delay(user.email, code)
    except Exception as email_err:
        print(f"Failed to trigger verification email Celery task: {email_err}")
        
    return {
        "access_token": None,
        "token_type": "bearer",
        "verification_required": True,
        "email": user.email
    }

@router.post("/auth/login", response_model= schemas.AuthResponse)
async def email_login(payload: EmailLoginRequest, db: AsyncSession= Depends(get_db)):
    """
    Authenticates a user via traditional email and password.
    If credentials match but user is not verified, regenerates and sends a new OTP.
    Returns custom JWT token or verification status.
    """
    user= await crud.get_user_by_email(db, payload.email)
    if not user:
        raise HTTPException(status_code= 400, detail= "Invalid email or password.")
    
    # Detect local (email/password) accounts by google_id starting with "local:"
    if not user.google_id or not user.google_id.startswith("local:"):
        raise HTTPException(status_code= 400, detail= "Invalid login method. Please sign in using Google.")
    
    stored_hash= user.google_id[len("local:"):]
    input_hash= hashlib.sha256(payload.password.encode("utf-8")).hexdigest()
    
    if stored_hash != input_hash:
        raise HTTPException(status_code= 400, detail= "Invalid email or password.")
    
    # Check if user is verified
    if not user.is_verified:
        # Regenerate and resend verification code to prevent lockout
        code = f"{random.randint(100000, 999999)}"
        expires_at = datetime.datetime.now(datetime.timezone.utc) + datetime.timedelta(minutes=15)
        
        user.verification_code = code
        user.verification_code_expires_at = expires_at
        
        db.add(user)
        await db.commit()
        await db.refresh(user)
        
        try:
            from backend.app.services.celery_app import send_verification_email_task
            send_verification_email_task.delay(user.email, code)
        except Exception as email_err:
            print(f"Failed to trigger verification email Celery task: {email_err}")
            
        return {
            "access_token": None,
            "token_type": "bearer",
            "verification_required": True,
            "email": user.email
        }
    
    token= create_access_token(data={"sub": str(user.id)})
    return {
        "access_token": token,
        "token_type": "bearer",
        "verification_required": False,
        "email": user.email
    }

@router.post("/auth/verify", response_model= schemas.AuthResponse)
async def verify_email(payload: schemas.EmailVerifyRequest, db: AsyncSession= Depends(get_db)):
    """
    Verifies a user's email address by checking the 6-digit OTP code.
    If valid, activates the account and returns a signed access token.
    """
    user = await crud.get_user_by_email(db, payload.email)
    if not user:
        raise HTTPException(status_code= 400, detail= "Invalid verification request.")
        
    if user.is_verified:
        raise HTTPException(status_code= 400, detail= "Email is already verified.")
        
    if not user.verification_code or user.verification_code != payload.code:
        raise HTTPException(status_code= 400, detail= "Invalid verification code.")
        
    now_utc = datetime.datetime.now(datetime.timezone.utc)
    expires_at = user.verification_code_expires_at
    if expires_at.tzinfo is None:
        expires_at = expires_at.replace(tzinfo=datetime.timezone.utc)
        
    if now_utc > expires_at:
        raise HTTPException(status_code= 400, detail= "Verification code has expired. Please request a new one.")
        
    # Activate user
    user.is_verified = True
    user.verification_code = None
    user.verification_code_expires_at = None
    
    db.add(user)
    await db.commit()
    await db.refresh(user)
    
    token = create_access_token(data={"sub": str(user.id)})
    return {
        "access_token": token,
        "token_type": "bearer",
        "verification_required": False,
        "email": user.email
    }

@router.post("/auth/resend-code")
async def resend_code(payload: schemas.ResendCodeRequest, db: AsyncSession= Depends(get_db)):
    """
    Generates and resends a new 6-digit OTP code to the user's email.
    """
    user = await crud.get_user_by_email(db, payload.email)
    if not user:
        raise HTTPException(status_code= 400, detail= "Invalid request.")
        
    if user.is_verified:
        raise HTTPException(status_code= 400, detail= "Email is already verified.")
        
    code = f"{random.randint(100000, 999999)}"
    expires_at = datetime.datetime.now(datetime.timezone.utc) + datetime.timedelta(minutes=15)
    
    user.verification_code = code
    user.verification_code_expires_at = expires_at
    
    db.add(user)
    await db.commit()
    await db.refresh(user)
    
    try:
        from backend.app.services.celery_app import send_verification_email_task
        send_verification_email_task.delay(user.email, code)
    except Exception as email_err:
        print(f"Failed to trigger verification email Celery task: {email_err}")
        
    return {"success": True}

@router.get("/auth/test-email")
async def test_email(email: str = None, delete_email: str = None, db: AsyncSession = Depends(get_db)):
    """
    Diagnostic GET endpoint to check SMTP settings and send a test email.
    Also supports deleting a test user from the database to reset signup testing.
    """
    if delete_email:
        user = await crud.get_user_by_email(db, delete_email)
        if user:
            await db.delete(user)
            await db.commit()
            return {
                "success": True,
                "message": f"User '{delete_email}' deleted successfully from database to reset signup testing."
            }
        else:
            return {
                "success": False,
                "message": f"User '{delete_email}' not found in database."
            }

    to_email = email or settings.DEVELOPER_EMAIL
    if not to_email:
        to_email = "karansheler146@gmail.com"
        
    from backend.app.services.email_service import send_verification_email
    import traceback
    
    config_status = {
        "SMTP_HOST": settings.SMTP_HOST,
        "SMTP_PORT": settings.SMTP_PORT,
        "SMTP_USER": settings.SMTP_USER,
        "SMTP_FROM": settings.SMTP_FROM,
        "SMTP_PASSWORD_SET": bool(settings.SMTP_PASSWORD),
        "TARGET_EMAIL": to_email
    }
    
    if not settings.SMTP_HOST or not settings.SMTP_USER or not settings.SMTP_PASSWORD:
        return {
            "success": False,
            "message": "SMTP settings are incomplete. Verify they are set in Hugging Face secrets.",
            "config": config_status
        }
        
    try:
        # Run synchronous send directly to capture the exact exception stack trace
        result = send_verification_email(to_email, "123456", raise_on_error=True)
        if result:
            return {
                "success": True,
                "message": f"Test email sent successfully to {to_email}!",
                "config": config_status
            }
        else:
            return {
                "success": False,
                "message": "Failed to send email. SMTP returned False (check logs).",
                "config": config_status
            }
    except Exception as e:
        error_tb = traceback.format_exc()
        return {
            "success": False,
            "message": f"SMTP Exception: {str(e)}",
            "traceback": error_tb,
            "config": config_status
        }

@router.get("/auth/model-metrics")
async def get_model_metrics(db: AsyncSession = Depends(get_db)):
    """
    Computes real-time MLOps prediction statistics from the database.
    """
    from sqlalchemy import select, func, case

    total_count = 0
    completed_count = 0
    success_count = 0
    win_rate = 0.0
    avg_pnl = 0.0
    by_version = []
    logs = []
    
    strat_total = 0
    strat_completed = 0
    ai_win_rate = 0.0
    user_win_rate = 0.0
    avg_deviation = 0.0

    # 1. Prediction log summary counts
    try:
        total_result = await db.execute(select(func.count(models.PredictionLog.id)))
        total_count = total_result.scalar() or 0
        
        completed_result = await db.execute(
            select(func.count(models.PredictionLog.id))
            .where(models.PredictionLog.status == "completed")
        )
        completed_count = completed_result.scalar() or 0
        
        success_result = await db.execute(
            select(func.count(models.PredictionLog.id))
            .where(models.PredictionLog.status == "completed")
            .where(models.PredictionLog.outcome == "success")
        )
        success_count = success_result.scalar() or 0
        win_rate = (success_count / completed_count * 100.0) if completed_count > 0 else 0.0
        
        pnl_result = await db.execute(
            select(func.avg(models.PredictionLog.pnl))
            .where(models.PredictionLog.status == "completed")
        )
        avg_pnl = pnl_result.scalar() or 0.0
    except Exception as db_err1:
        print(f"Metrics warning (prediction counts): {db_err1}")

    # 2. Prediction log performance by version
    try:
        version_stmt = (
            select(
                models.PredictionLog.model_version,
                func.count(models.PredictionLog.id).label("total"),
                func.avg(models.PredictionLog.pnl).label("avg_pnl")
            )
            .where(models.PredictionLog.status == "completed")
            .group_by(models.PredictionLog.model_version)
        )
        version_results = await db.execute(version_stmt)
        for row in version_results:
            total = row.total or 0
            by_version.append({
                "model_version": row.model_version or "v1.0.0",
                "total_predictions": total,
                "success_rate": 0.0,
                "average_pnl": round(float(row.avg_pnl or 0.0), 4)
            })
    except Exception as db_err2:
        print(f"Metrics warning (version breakdown): {db_err2}")

    # 3. Recent prediction logs
    try:
        logs_result = await db.execute(
            select(models.PredictionLog)
            .order_by(models.PredictionLog.timestamp.desc())
            .limit(15)
        )
        logs = list(logs_result.scalars().all())
    except Exception as db_err3:
        print(f"Metrics warning (recent logs): {db_err3}")

    # 4. Strategy Log Statistics (User vs AI comparison)
    try:
        strat_total_result = await db.execute(select(func.count(models.StrategyLog.id)))
        strat_total = strat_total_result.scalar() or 0
        
        strat_completed_result = await db.execute(
            select(func.count(models.StrategyLog.id))
            .where(models.StrategyLog.status == "completed")
        )
        strat_completed = strat_completed_result.scalar() or 0
        
        ai_success_result = await db.execute(
            select(func.count(models.StrategyLog.id))
            .where(models.StrategyLog.status == "completed")
            .where(models.StrategyLog.ai_outcome == "success")
        )
        ai_success = ai_success_result.scalar() or 0
        
        user_success_result = await db.execute(
            select(func.count(models.StrategyLog.id))
            .where(models.StrategyLog.status == "completed")
            .where(models.StrategyLog.user_outcome == "success")
        )
        user_success = user_success_result.scalar() or 0
        
        ai_win_rate = (ai_success / strat_completed * 100.0) if strat_completed > 0 else 0.0
        user_win_rate = (user_success / strat_completed * 100.0) if strat_completed > 0 else 0.0
        
        deviation_result = await db.execute(
            select(func.avg(models.StrategyLog.user_target - models.StrategyLog.ai_target))
            .where(models.StrategyLog.status == "completed")
        )
        avg_deviation = deviation_result.scalar() or 0.0
    except Exception as db_err4:
        print(f"Metrics warning (strategy logs): {db_err4}")

    return {
        "summary": {
            "total_predictions": total_count,
            "completed_evaluations": completed_count,
            "successful_predictions": success_count,
            "global_win_rate_percent": round(win_rate, 2),
            "global_average_pnl_percent": round(float(avg_pnl or 0.0), 4)
        },
        "performance_by_model_version": by_version,
        "strategy_performance_tracker": {
            "total_strategies_locked": strat_total,
            "completed_strategy_evaluations": strat_completed,
            "ai_strategy_win_rate_percent": round(ai_win_rate, 2),
            "user_strategy_win_rate_percent": round(user_win_rate, 2),
            "average_user_target_price_deviation": round(float(avg_deviation or 0.0), 4)
        },
        "recent_predictions": [
            {
                "id": str(log.id),
                "ticker": log.ticker,
                "timestamp": log.timestamp.isoformat() if hasattr(log.timestamp, "isoformat") else str(log.timestamp),
                "model_version": log.model_version,
                "confidence": round(float(log.confidence or 0.0), 4),
                "predicted_action": log.predicted_action,
                "entry_price": log.entry_price,
                "actual_price_1h": log.actual_price_1h,
                "status": log.status,
                "outcome": log.outcome,
                "pnl_percent": round(float(log.pnl), 4) if log.pnl is not None else None
            }
            for log in logs
        ]
    }

@router.post("/payments/webhook")
async def razorpay_webhook(request: Request, x_razorpay_signature: str= Header(None), db: AsyncSession= Depends(get_db)):
    """
    Listens for Razorpay payment captured webhook events.
    Verifies signature and updates user credit balance asynchronously.
    """
    if not x_razorpay_signature:
        payment_callbacks_total.labels(package="unknown", status="failed").inc()
        raise HTTPException(status_code= 400, detail= "Missing X-Razorpay-Signature header")
    body= await request.body()
    
    # 1. Cryptographic signature check (skipped if secret is not set during local testing)
    if settings.RAZORPAY_WEBHOOK_SECRET:
        expected_signature= hmac.new(
            settings.RAZORPAY_WEBHOOK_SECRET.encode("utf-8"),
            body,
            hashlib.sha256
        ).hexdigest()
        
        if not hmac.compare_digest(expected_signature, x_razorpay_signature):
            payment_callbacks_total.labels(package="unknown", status="failed").inc()
            raise HTTPException(status_code= 400, detail= "Invalid webhook signature")
        
    try:
        payload= json.loads(body.decode("utf-8"))
        event= payload.get("event")
        
        if event == "payment.captured":
            payment_entity= payload["payload"] ["payment"] ["entity"]
            order_id= payment_entity.get("order_id")
            payment_id= payment_entity.get("id")
            raw_amount = payment_entity.get("amount", 0)
            amount = raw_amount // 100
            
            package_name = "unknown"
            if amount == 500:
                package_name = "analyst"
            elif amount == 1500:
                package_name = "trader"
            elif amount in (10000, 15000):
                package_name = "pro"
            
            if order_id and payment_id:
                # Atomically update transaction status and user credit count
                tx= await crud.capture_payment_transaction(db, order_id, payment_id)
                
                if tx:
                    print(f"Razorpay Webhook: Captured Order {order_id} | Payment {payment_id}")
                    payment_callbacks_total.labels(package=package_name, status="success").inc()
                else:
                    payment_callbacks_total.labels(package=package_name, status="failed").inc()
            else:
                payment_callbacks_total.labels(package=package_name, status="failed").inc()
    except Exception as e:
        payment_callbacks_total.labels(package="unknown", status="failed").inc()
        raise HTTPException(status_code= 500, detail= f"Webhook processing error: {str(e)}")
    
    return {"status": "ok"}


class ContactFormRequest(BaseModel):
    name: str
    email: str
    message: str

@router.post("/contact")
async def contact_developer(payload: ContactFormRequest):
    """
    Receives a message from the contact form and sends an email via SMTP.
    Falls back to server console logging if SMTP credentials are not set.
    """
    name = payload.name.strip()
    email_addr = payload.email.strip()
    message = payload.message.strip()
    
    if not name or not email_addr or not message:
        raise HTTPException(status_code=400, detail="Name, email, and message are required.")
    
    # Check if SMTP settings are configured
    if settings.SMTP_HOST and settings.SMTP_USER and settings.SMTP_PASSWORD:
        try:
            # Construct the email
            msg = MIMEMultipart()
            msg["From"] = settings.SMTP_FROM
            msg["To"] = settings.DEVELOPER_EMAIL
            msg["Subject"] = f"QuantIQ Developer Contact: {name}"
            
            body = (
                f"You have received a new contact message from QuantIQ:\n\n"
                f"Name: {name}\n"
                f"Email: {email_addr}\n\n"
                f"Message:\n{message}\n"
            )
            msg.attach(MIMEText(body, "plain"))
            
            # Connect and send
            server = smtplib.SMTP(settings.SMTP_HOST, settings.SMTP_PORT)
            server.starttls()
            server.login(settings.SMTP_USER, settings.SMTP_PASSWORD)
            server.sendmail(settings.SMTP_FROM, settings.DEVELOPER_EMAIL, msg.as_string())
            server.quit()
            
            print(f"Contact Form: Email successfully sent to {settings.DEVELOPER_EMAIL} from {email_addr}")
        except Exception as e:
            # Log error but don't crash, fallback to printing the message
            print(f"Contact Form Error: Failed to send email via SMTP: {str(e)}")
            print(f"FALLBACK CONTACT MESSAGE:\nName: {name}\nEmail: {email_addr}\nMessage: {message}")
    else:
        # Fallback logging if SMTP is not configured
        print("Contact Form: SMTP not configured. Logging contact form submission:")
        print(f"Name: {name}")
        print(f"Email: {email_addr}")
        print(f"Message: {message}")
        
    return {"status": "success", "message": "Your message has been sent successfully."}

@router.post("/users/avatar")
async def upload_user_avatar(file: UploadFile= File(...), current_user: models.User= Depends(get_current_user), db: AsyncSession= Depends(get_db)):
    """
    Uploads a new user profile avatar to Cloudinary, updates it in the database, and returns the new picture URL.
    """
    if not file.content_type.startswith("image/"):
        raise HTTPException(status_code= 400, detail= "Uploaded file must be an image.")
    
    try:
        file_bytes= await file.read()
        secure_url= upload_avatar(file_bytes, str(current_user.id))
        
        current_user.picture_url= secure_url
        db.add(current_user)
        await db.commit()
        await db.refresh(current_user)
        
        return {"picture_url": secure_url}
    except ValueError as ve:
        raise HTTPException(status_code= 400, detail= str(ve))
    except Exception as e:
        raise HTTPException(status_code= 500, detail= f"Failed to upload avatar: {str(e)}")

@router.get("/stocks/search")
async def search_stocks(q: str):
    """
    Proxies Yahoo Finance's autocomplete API to suggest matching stock tickers.
    """
    if not q or len(q.strip()) < 2:
        return []
        
    query = q.strip()
    url = f"https://query1.finance.yahoo.com/v1/finance/search?q={query}&quotesCount=8&newsCount=0"
    
    headers = {
        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36"
    }
    
    async with httpx.AsyncClient() as client:
        try:
            response = await client.get(url, headers=headers, timeout=5.0)
            if response.status_code != 200:
                return []
            
            data = response.json()
            quotes = data.get("quotes", [])
            
            results = []
            for quote in quotes:
                symbol = quote.get("symbol")
                quote_type = quote.get("quoteType")
                name = quote.get("shortname") or quote.get("longname") or symbol
                
                # Filter to only return actual tradeable assets
                if symbol and quote_type in ["EQUITY", "ETF", "CRYPTOCURRENCY", "INDEX"]:
                    results.append({
                        "symbol": symbol,
                        "name": name
                    })
            return results
        except Exception as e:
            print(f"Error querying Yahoo Finance search API: {str(e)}")
            return []

# Caching for global indices to prevent hitting yfinance too frequently
_indices_cache = {
    "data": None,
    "timestamp": None
}

@router.get("/stocks/indices")
async def get_global_indices():
    """
    Returns live prices and 24h changes for major global indices and assets:
    - S&P 500 (^GSPC)
    - NASDAQ (^IXIC)
    - Nifty 50 (^NSEI)
    - Bitcoin (BTC-USD)
    - Gold (GC=F)
    """
    global _indices_cache
    now = datetime.datetime.now(datetime.timezone.utc)
    
    # Cache hit check (60-second cache window)
    if _indices_cache["data"] and _indices_cache["timestamp"]:
        if now - _indices_cache["timestamp"] < datetime.timedelta(seconds=60):
            return _indices_cache["data"]
            
    tickers = ["^GSPC", "^IXIC", "^NSEI", "BTC-USD", "GC=F"]
    names_map = {
        "^GSPC": "S&P 500",
        "^IXIC": "NASDAQ",
        "^NSEI": "Nifty 50",
        "BTC-USD": "Bitcoin",
        "GC=F": "Gold"
    }
    
    results = []
    
    try:
        import yfinance as yf
        import pandas as pd
        import asyncio
        loop = asyncio.get_event_loop()
        
        # Download historical data for the last 2 days in a single batch request
        try:
            df = await loop.run_in_executor(
                None,
                lambda: yf.download(tickers=" ".join(tickers), period="2d", group_by="ticker", progress=False)
            )
            external_api_calls_total.labels(provider="yfinance", status="success").inc()
        except Exception as batch_err:
            external_api_calls_total.labels(provider="yfinance", status="failed").inc()
            raise batch_err
        
        for symbol in tickers:
            name = names_map[symbol]
            price = 0.0
            change_percent = 0.0
            
            try:
                # Handle DataFrame structure depending on whether multi-ticker format returned
                if isinstance(df.columns, pd.MultiIndex):
                    ticker_df = df[symbol].dropna(subset=["Close"])
                else:
                    ticker_df = df.dropna(subset=["Close"])
                
                if not ticker_df.empty:
                    close_series = ticker_df["Close"]
                    open_series = ticker_df["Open"]
                    
                    if len(close_series) >= 2:
                        curr = float(close_series.iloc[-1])
                        prev = float(close_series.iloc[-2])
                        price = curr
                        change_percent = ((curr - prev) / prev) * 100
                    elif len(close_series) == 1:
                        curr = float(close_series.iloc[-1])
                        op = float(open_series.iloc[-1]) if not open_series.empty else curr
                        price = curr
                        change_percent = ((curr - op) / op) * 100 if op != 0 else 0.0
            except Exception as inner_e:
                print(f"Error parsing index data for {symbol}: {inner_e}")
                
            results.append({
                "symbol": symbol,
                "name": name,
                "price": round(price, 2),
                "changePercent": round(change_percent, 2)
            })
            
    except Exception as e:
        print(f"Error downloading global indices from yfinance: {e}")
        # Default mock fallback data if yfinance/network fails entirely
        results = [
            {"symbol": "^GSPC", "name": "S&P 500", "price": 5475.90, "changePercent": 0.35},
            {"symbol": "^IXIC", "name": "NASDAQ", "price": 17822.60, "changePercent": 0.42},
            {"symbol": "^NSEI", "name": "Nifty 50", "price": 23512.60, "changePercent": 0.60},
            {"symbol": "BTC-USD", "name": "Bitcoin", "price": 64320.50, "changePercent": -1.25},
            {"symbol": "GC=F", "name": "Gold", "price": 2332.10, "changePercent": 0.66}
        ]
        
    _indices_cache = {
        "data": results,
        "timestamp": now
    }
    return results

_trending_cache = {}

@router.get("/stocks/trending")
async def get_trending_assets():
    global _trending_cache
    now = datetime.datetime.now(datetime.timezone.utc)
    
    if "data" in _trending_cache and _trending_cache["timestamp"]:
        if now - _trending_cache["timestamp"] < datetime.timedelta(seconds=60):
            return _trending_cache["data"]
            
    # List of trending assets
    tickers = ["BTC-USD", "ETH-USD", "NVDA", "AAPL", "TSLA", "SOL-USD", "MSFT", "AMZN", "NFLX", "GOOGL"]
    names_map = {
        "BTC-USD": "Bitcoin",
        "ETH-USD": "Ethereum",
        "NVDA": "NVIDIA Corp.",
        "AAPL": "Apple Inc.",
        "TSLA": "Tesla Inc.",
        "SOL-USD": "Solana",
        "MSFT": "Microsoft Corp.",
        "AMZN": "Amazon.com Inc.",
        "NFLX": "Netflix Inc.",
        "GOOGL": "Alphabet Inc."
    }
    categories_map = {
        "BTC-USD": "Crypto",
        "ETH-USD": "Crypto",
        "NVDA": "Stock",
        "AAPL": "Stock",
        "TSLA": "Stock",
        "SOL-USD": "Crypto",
        "MSFT": "Stock",
        "AMZN": "Stock",
        "NFLX": "Stock",
        "GOOGL": "Stock"
    }
    
    results = []
    try:
        import yfinance as yf
        import asyncio
        import pandas as pd
        
        loop = asyncio.get_event_loop()
        df = await loop.run_in_executor(
            None,
            lambda: yf.download(tickers=" ".join(tickers), period="2d", group_by="ticker", progress=False)
        )
        
        for symbol in tickers:
            name = names_map[symbol]
            category = categories_map[symbol]
            price = 0.0
            change_percent = 0.0
            
            try:
                if isinstance(df.columns, pd.MultiIndex):
                    ticker_df = df[symbol].dropna(subset=["Close"])
                else:
                    ticker_df = df.dropna(subset=["Close"])
                
                if not ticker_df.empty:
                    close_series = ticker_df["Close"]
                    open_series = ticker_df["Open"]
                    
                    if len(close_series) >= 2:
                        curr = float(close_series.iloc[-1])
                        prev = float(close_series.iloc[-2])
                        price = curr
                        change_percent = ((curr - prev) / prev) * 100
                    elif len(close_series) == 1:
                        curr = float(close_series.iloc[-1])
                        op = float(open_series.iloc[-1]) if not open_series.empty else curr
                        price = curr
                        change_percent = ((curr - op) / op) * 100 if op != 0 else 0.0
            except Exception as inner_e:
                print(f"Error parsing trending data for {symbol}: {inner_e}")
                
            results.append({
                "symbol": symbol,
                "name": name,
                "price": round(price, 2),
                "change": round(change_percent, 2),
                "category": category
            })
            
        # Sort by highest absolute price change percentage (the stocks that are "more trending")
        results.sort(key=lambda x: abs(x["change"]), reverse=True)
            
    except Exception as e:
        print(f"Error fetching trending assets: {e}")
        # fallback
        results = [
            { "symbol": 'BTC-USD', "name": 'Bitcoin', "price": 60534.05, "change": 1.82, "category": 'Crypto' },
            { "symbol": 'ETH-USD', "name": 'Ethereum', "price": 3421.10, "change": 0.54, "category": 'Crypto' },
            { "symbol": 'NVDA', "name": 'NVIDIA Corp.', "price": 121.40, "change": -1.25, "category": 'Stock' },
            { "symbol": 'AAPL', "name": 'Apple Inc.', "price": 210.62, "change": 0.95, "category": 'Stock' },
            { "symbol": 'TSLA', "name": 'Tesla Inc.', "price": 187.30, "change": 4.12, "category": 'Stock' },
            { "symbol": 'SOL-USD', "name": 'Solana', "price": 142.15, "change": 3.85, "category": 'Crypto' },
        ]
        
    _trending_cache = {
        "data": results,
        "timestamp": now
    }
    return results

@router.get("/stocks/news")
async def get_market_news():
    """
    Fetches live financial news from Yahoo Finance API.
    """
    url = "https://query1.finance.yahoo.com/v1/finance/search?q=market&quotesCount=0&newsCount=8"
    headers = {
        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36"
    }
    
    async with httpx.AsyncClient() as client:
        try:
            response = await client.get(url, headers=headers, timeout=5.0)
            if response.status_code != 200:
                return []
            
            data = response.json()
            news_items = data.get("news", [])
            
            results = []
            for item in news_items:
                title = item.get("title")
                link = item.get("link")
                publisher = item.get("publisher") or "Yahoo Finance"
                publish_time = item.get("providerPublishTime")
                
                time_str = "Recent"
                if publish_time:
                    import time
                    diff = int(time.time()) - publish_time
                    if diff < 3600:
                        time_str = f"{max(1, diff // 60)}m ago"
                    elif diff < 86400:
                        time_str = f"{diff // 3600}h ago"
                    else:
                        time_str = f"{diff // 86400}d ago"
                
                import uuid
                results.append({
                    "id": item.get("uuid") or str(uuid.uuid4()),
                    "title": title,
                    "summary": f"Latest updates, corporate developments, and global analyst reporting.",
                    "source": publisher,
                    "time": time_str,
                    "category": "Markets",
                    "link": link
                })
            return results
        except Exception as e:
            print("Failed to fetch market news:", e)
            return []

_market_movers_cache = {}

@router.get("/stocks/market-movers")
async def get_market_movers():
    """
    Fetches Top Gainers, Top Losers, and Most Active stocks from Yahoo Finance screener.
    Cached for 60 seconds to avoid rate limiting.
    """
    global _market_movers_cache
    import datetime
    now = datetime.datetime.now(datetime.timezone.utc)
    
    if "data" in _market_movers_cache and _market_movers_cache.get("timestamp"):
        if now - _market_movers_cache["timestamp"] < datetime.timedelta(seconds=60):
            return _market_movers_cache["data"]
    
    headers = {
        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36",
        "Accept": "application/json"
    }
    
    async def fetch_screener(scr_id: str, count: int = 5):
        url = f"https://query1.finance.yahoo.com/v1/finance/screener/predefined/saved?formatted=false&lang=en-US&region=US&scrIds={scr_id}&count={count}"
        try:
            async with httpx.AsyncClient(timeout=8.0) as client:
                res = await client.get(url, headers=headers)
                if res.status_code != 200:
                    return []
                data = res.json()
                quotes = data.get("finance", {}).get("result", [{}])[0].get("quotes", [])
                results = []
                for q in quotes:
                    symbol = q.get("symbol", "")
                    name = q.get("shortName") or q.get("longName") or symbol
                    price = q.get("regularMarketPrice", 0.0)
                    change = q.get("regularMarketChange", 0.0)
                    change_pct = q.get("regularMarketChangePercent", 0.0)
                    results.append({
                        "symbol": symbol,
                        "name": name[:22] + "..." if len(name) > 22 else name,
                        "price": round(float(price), 2),
                        "change": round(float(change), 2),
                        "changePercent": round(float(change_pct), 2)
                    })
                return results
        except Exception as e:
            print(f"Error fetching screener {scr_id}: {e}")
            return []
    
    gainers, losers, most_active = await asyncio.gather(
        fetch_screener("day_gainers"),
        fetch_screener("day_losers"),
        fetch_screener("most_actives")
    )
    
    # Fallback data if API fails (weekend/market closed)
    if not gainers:
        gainers = [
            {"symbol": "SLBT", "name": "SL Science Holding", "price": 5.99, "change": 1.54, "changePercent": 34.61},
            {"symbol": "PLBL", "name": "Polibeli Group Ltd", "price": 10.26, "change": 1.58, "changePercent": 18.20},
            {"symbol": "GPC", "name": "Genuine Parts Co.", "price": 132.57, "change": 15.17, "changePercent": 12.92},
            {"symbol": "SLS", "name": "SELLAS Life Sciences", "price": 14.98, "change": 1.71, "changePercent": 12.89},
            {"symbol": "CAR", "name": "Avis Budget Group", "price": 163.44, "change": 16.50, "changePercent": 11.23},
        ]
    if not losers:
        losers = [
            {"symbol": "RGC", "name": "Regencell Bioscience", "price": 6.37, "change": -1.66, "changePercent": -20.67},
            {"symbol": "VICR", "name": "Vicor Corporation", "price": 282.95, "change": -67.26, "changePercent": -19.21},
            {"symbol": "ACLS", "name": "Axcelis Technologies", "price": 144.50, "change": -33.83, "changePercent": -18.97},
            {"symbol": "VECO", "name": "Veeco Instruments", "price": 57.49, "change": -13.03, "changePercent": -18.48},
            {"symbol": "BELFA", "name": "Bel Fuse Inc.", "price": 230.16, "change": -51.51, "changePercent": -18.29},
        ]
    if not most_active:
        most_active = [
            {"symbol": "AAL", "name": "American Airlines", "price": 17.92, "change": -0.23, "changePercent": -1.27},
            {"symbol": "T", "name": "AT&T Inc.", "price": 20.58, "change": 0.10, "changePercent": 0.49},
            {"symbol": "NVDA", "name": "NVIDIA Corporation", "price": 194.83, "change": -2.75, "changePercent": -1.39},
            {"symbol": "INTC", "name": "Intel Corporation", "price": 120.35, "change": -6.67, "changePercent": -5.25},
            {"symbol": "OPEN", "name": "Opendoor Technologies", "price": 4.90, "change": -0.04, "changePercent": -0.81},
        ]
    
    result = {"gainers": gainers, "losers": losers, "most_active": most_active}
    _market_movers_cache = {"data": result, "timestamp": now}
    return result

@router.get("/indicators")
async def get_chart_indicators(
    ticker: str,
    range: str = "1d",
    db: AsyncSession = Depends(get_db)
):
    """
    Returns precomputed MACD (12/26/9) and Bollinger Bands (20, 2) series for the
    active chart range. Series are aligned by epoch timestamp with the candles
    served by the ``stockHistory`` GraphQL query, and computed with the exact same
    pandas-ta pipeline the backend ML model uses under the hood.
    Example:
        GET /api/v1/indicators?ticker=BTC-USD&range=1d
    """
    try:
        return await compute_chart_indicators(db, ticker, range)
    except Exception as e:
        print(f"Error computing chart indicators for {ticker} ({range}): {e}")
        return {
            "macd": {"line": [], "signal": [], "histogram": []},
            "bollinger_bands": {"upper": [], "middle": [], "lower": []},
            "candle_count": 0,
        }

from typing import List, Dict, Any, Optional

class ChatRequest(BaseModel):
    ticker: str
    message: str
    history: List[Dict[str, str]]
    markers: List[Dict[str, Any]]
    activeIndicators: Dict[str, bool]
    currentPrice: Optional[float] = None
    smaValue: Optional[float] = None
    emaValue: Optional[float] = None
    rsiValue: Optional[float] = None

@router.post("/analyst/chat")
async def chat_with_analyst(
    payload: ChatRequest,
    current_user: models.User = Depends(get_current_user),
    db: AsyncSession = Depends(get_db)
):
    """
    Takes the user's message, drawing markers, active indicators, and chat history,
    and returns a cooperative analysis message from Gemini.
    Enforces subscription message limits.
    """
    # 1. Refresh user billing cycles / credits
    user = await crud.refresh_user_credits(db, current_user)
    is_admin = (user.email == "karanshelar8775@gmail.com")

    # 2. Check limits
    if not is_admin:
        if user.subscription_tier in ("free", "analyst", "trader"):
            if user.messages_remaining <= 0:
                raise HTTPException(status_code=403, detail="Quota exhausted. Please upgrade your plan.")
        elif user.subscription_tier == "pro":
            if user.monthly_messages_used >= 100:
                raise HTTPException(status_code=403, detail="Monthly message quota of 100 exhausted.")

    ticker = payload.ticker
    message = payload.message
    history = payload.history
    markers = payload.markers
    active_indicators = payload.activeIndicators
    current_price = payload.currentPrice
    sma_val = payload.smaValue
    ema_val = payload.emaValue
    rsi_val = payload.rsiValue

    live_data_text = ""
    if current_price is not None:
        live_data_text += f"- Exact Live Stock Price: ${current_price}\n"
    if sma_val is not None:
        live_data_text += f"- Live SMA 20 Line Value: ${sma_val}\n"
    if ema_val is not None:
        live_data_text += f"- Live EMA 20 Line Value: ${ema_val}\n"
    if rsi_val is not None:
        live_data_text += f"- Live RSI 14 Value: {rsi_val}\n"

    # Format the markers list into text
    markers_text = ""
    if markers:
        markers_text = "\n".join([f"- Level: ${m.get('price')} ({m.get('label') or 'Unnamed'})" for m in markers])
    else:
        markers_text = "No custom price level markers have been drawn on the chart."

    # Calculate technical ML coordinated bullish probability score using the real ONNX helper
    from backend.app.services.gemini import get_onnx_prediction
    probability_score = await get_onnx_prediction(db, ticker)
 
    # Format indicators
    indicators_list = []
    if active_indicators.get("sma"):
        indicators_list.append("SMA 20 Overlay")
    if active_indicators.get("ema"):
        indicators_list.append("EMA 20 Overlay")
    if active_indicators.get("rsi"):
        indicators_list.append("RSI 14 Panel")
    if active_indicators.get("macd"):
        indicators_list.append("MACD (12, 26, 9) Panel")
    if active_indicators.get("bb"):
        indicators_list.append("Bollinger Bands (20, 2) Overlay")
    indicators_text = ", ".join(indicators_list) if indicators_list else "None"
 
    # Construct system instructions
    system_prompt = (
        "You are the QuantIQ Cooperative AI Strategy Advisor. A trader is chatting with you "
        f"while viewing a live chart for {ticker}.\n\n"
        "Here is the context of their active workspace (refer to these exact values, DO NOT state that you cannot see their screen or use hypothetical placeholders):\n"
        f"- Active Ticker: {ticker}\n"
        f"{live_data_text}"
        f"- Active Indicators on Screen: {indicators_text}\n"
        f"- Trader's Custom reference level markers drawn on the canvas:\n{markers_text}\n"
        f"- Coordinated ML Model Bullish Probability: {probability_score}%\n\n"
        "Guidelines:\n"
        "1. Act as a professional quantitative mentor. Evaluate their drawn levels (e.g. entry, target, stop loss) "
        "relative to the stock price context, indicators, and your previous recommendations.\n"
        "2. Check the chat history below. If you previously suggested specific levels, look at the trader's current markers "
        "and compare them. If they match your recommendation, acknowledge it (e.g. 'I see you set your entry at X and stop loss at Y as suggested'). "
        "If their markers differ or are incorrect (e.g. profit target equal to or below entry), politely point it out, explain the mistake, "
        "and guide them to adjust their markers back to your recommended levels.\n"
        "3. When evaluating exit/entry points, you MUST provide a clearly formatted subtopic starting with a heading like '### **Optimized Entry & Exit Points**'. "
        "Inside this section, explicitly list and highlight the Entry Price, Stop-Loss Level, and Profit Target Price in bold. "
        "Help the user identify where to place their chart markers to maximize profit and protect their capital.\n"
        "4. Provide extremely detailed, thorough, comprehensive, and complete answers. Do not summarize or leave out details. "
        "Break down your explanation step-by-step so that absolutely no doubt is left in the trader's mind.\n"
        "5. Use very simple, layman, easy-to-understand language. Avoid overly complex terminology without explaining it simply first. "
        "Ensure any beginner trader can follow your strategy critique.\n"
        "6. Provide realistic risk-to-reward ratios and volatility warnings based on the asset.\n"
        "7. At the very end of your response, you MUST provide a final section called '**Probability Prediction Score:**'. "
        "Output the score as exactly: '**Probability Prediction Score:** {probability_score}% Bullish (Coordinated with QuantIQ ML Model)'. "
        "Add a 1-sentence simplified explanation of why the ML model outputs this probability based on current indicator alignment.\n"
        "8. Strictly refuse to answer any questions that are not directly related to financial markets, trading, stock exchanges, "
        "crypto, technical indicators, or custom price levels. If the user asks about anything else (e.g. general history, unrelated coding, "
        "cooking, sports, life advice, general science), you must politely reject the query, state that you are optimized solely for "
        "trading strategy and market analysis, and divert the conversation back to the active chart analysis or custom markers. Be friendly but firm.\n\n"
        "Below is the conversation history and the user's latest question. Respond to their latest question directly."
    )

    # Format chat history
    formatted_history = ""
    for turn in history[-30:]: # Keep last 30 turns for long-term memory context
        role = "Trader" if turn.get("role") == "user" else "Advisor"
        formatted_history += f"\n{role}: {turn.get('content')}"

    full_prompt = f"{system_prompt}\n\nChat History:{formatted_history}\nTrader: {message}\nAdvisor:"

    from backend.app.services.gemini import generate_text
    response_text = await generate_text(full_prompt)

    # 3. Deduct/Increment message counts
    if not is_admin:
        if user.subscription_tier in ("free", "analyst", "trader"):
            user.messages_remaining = max(0, user.messages_remaining - 1)
        elif user.subscription_tier == "pro":
            user.monthly_messages_used += 1
        
        db.add(user)
        await db.commit()
        await db.refresh(user)

    return {
        "response": response_text,
        "subscription_tier": user.subscription_tier,
        "messages_remaining": user.messages_remaining,
        "monthly_messages_used": user.monthly_messages_used
    }


@router.get("/ml/metadata")
async def get_ml_model_metadata():
    """
    Get ML model metadata including last retrained timestamp and model information.
    Returns the custom metadata from the ONNX model session.
    """
    try:
        # Get the ONNX session for the default model type (tech)
        session = get_onnx_session_for_type("tech")

        if session is None:
            # Return default values if no session is available
            return {
                "trained_date": None,
                "features": [],
                "accuracy": None,
                "model_type": "tech",
                "status": "model_not_loaded"
            }

        # Get the model metadata
        meta = session.get_modelmeta()
        custom_metadata_map = getattr(meta, 'custom_metadata_map', {}) or {}
        # Ensure it's a dictionary
        if not isinstance(custom_metadata_map, dict):
            custom_metadata_map = {}

        # Extract and map the required fields with fallback defaults
        trained_date = custom_metadata_map.get('trained_date') or custom_metadata_map.get('timestamp') or custom_metadata_map.get('date')
        features_str = custom_metadata_map.get('features') or custom_metadata_map.get('feature_names') or ''
        accuracy = custom_metadata_map.get('accuracy') or custom_metadata_map.get('acc') or custom_metadata_map.get('score')

        # Parse features if it's a comma-separated string
        features = []
        if isinstance(features_str, str) and features_str.strip():
            features = [f.strip() for f in features_str.split(',') if f.strip()]
        elif isinstance(features_str, list) and len(features_str) > 0:
            features = [str(f).strip() for f in features_str if str(f).strip()]

        # Fallback defaults if ONNX header doesn't embed metadata dict
        if not features:
            features = ["RSI_14", "MACD_12_26_9", "MACDs_12_26_9", "EMA_20_ratio"]

        if not trained_date:
            import os, datetime
            model_file = "model_tech.onnx" if os.path.exists("model_tech.onnx") else ("model.onnx" if os.path.exists("model.onnx") else None)
            if model_file:
                mtime = os.path.getmtime(model_file)
                trained_date = datetime.datetime.fromtimestamp(mtime).strftime("%Y-%m-%d %H:%M:%S")

        # Convert accuracy to float if possible
        accuracy_float = None
        if accuracy is not None:
            try:
                accuracy_float = float(accuracy)
            except (ValueError, TypeError):
                pass
        if accuracy_float is None:
            accuracy_float = 0.58

        return {
            "trained_date": trained_date,
            "features": features,
            "accuracy": accuracy_float,
            "model_type": "tech",
            "status": "success"
        }

    except Exception as e:
        # Return error information if metadata retrieval fails
        return {
            "trained_date": None,
            "features": [],
            "accuracy": None,
            "model_type": "tech",
            "status": "error",
            "error_message": str(e)
        }