app.py

import logging
import psycopg2
from psycopg2.errors import UniqueViolation
import pandas as pd
import requests
import os
from dotenv import load_dotenv
import smtplib
from email.mime.text import MIMEText
from email.mime.multipart import MIMEMultipart

from fastapi import FastAPI, HTTPException, UploadFile, File, BackgroundTasks
from fastapi.responses import JSONResponse
from fastapi.middleware.cors import CORSMiddleware
from fastapi.encoders import jsonable_encoder
from pydantic import BaseModel
from typing import Optional

# Import our new database manager and the matching engine
from database import get_db_connection, init_db
from matching_engine import run_matching_engine
from diagnostics_engine import run_botanical_diagnosis

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("uzoagro-api")

load_dotenv()
PAYSTACK_SECRET_KEY = os.getenv("PAYSTACK_SECRET_KEY")

def verify_paystack_payment(reference: str) -> bool:
    """Securely asks Paystack if the transaction was actually successful."""
    url = f"https://api.paystack.co/transaction/verify/{reference}"
    headers = {
        "Authorization": f"Bearer {PAYSTACK_SECRET_KEY}"
    }
    try:
        response = requests.get(url, headers=headers)
        if response.status_code == 200:
            data = response.json()
            # Double-check that the transaction status is literally "success"
            if data.get("data", {}).get("status") == "success":
                return True
    except Exception as e:
        print(f"Paystack verification error: {e}")

    return False

app = FastAPI(title="UzoAgro AI Matching API", version="0.1.0")

# CORS config explicitly allowing your Vercel frontend (and everything else) to talk to this API
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Expanded 15-City Geographic Dictionary
CITIES = {
    # West / Mid-West
    "Lagos": (6.5244, 3.3792), "Ibadan": (7.3775, 3.9470), "Ilorin": (8.4799, 4.5418), "Akure": (7.2571, 5.2058),
    # North / Middle-Belt
    "Abuja": (9.0765, 7.3986), "Kano": (12.0022, 8.5920), "Kaduna": (10.5105, 7.4165), "Jos": (9.8965, 8.8583),
    "Sokoto": (13.0059, 5.2476),
    # East
    "Enugu": (6.4402, 7.4943), "Awka": (6.2100, 7.0700), "Onitsha": (6.1500, 6.7800), "Owerri": (5.4833, 7.0333),
    # South-South
    "Port Harcourt": (4.8156, 7.0498), "Uyo": (5.0380, 7.9098)
}


# --- Pydantic Data Models ---
class DiagnosisRequest(BaseModel):
    crop_type: str
    symptoms: str


class FarmerRequest(BaseModel):
    farmer_name: str
    pickup_city: str
    dropoff_city: str
    crop_type: str
    required_capacity: float
    requested_date: str


class UserSignup(BaseModel):
    role: str
    name: str
    phone: str
    nin: str
    primary_city: str
    pin: str
    email: Optional[str] = None


class UserLogin(BaseModel):
    phone: str
    pin: str


# --- API Endpoints ---

@app.on_event("startup")
def startup_load_data():
    """Ensure database is ready when the server starts."""
    try:
        init_db()
    except Exception as e:
        logger.exception("Failed to initialize database on startup")
        raise


@app.get("/")
def read_root():
    """Health check endpoint to prove the API is alive to Hugging Face."""
    return {"message": "UzoAgro API is running smoothly!"}


# ... (keep your existing imports and database connection setup) ...

# 1. THE MOCK KYC FUNCTION (Place this above your routes)
def verify_identity_mock(nin: str, name: str) -> bool:
    """
    Simulates a call to Prembly/Dojah.
    In the future, we replace this logic with the real API request.
    """
    # Simulate a small delay like a real server
    import time
    time.sleep(1)

    # Simple validation rule for testing:
    # If the NIN is exactly 11 digits, we assume it's valid.
    if len(nin) == 11 and nin.isdigit():
        return True
    return False


def send_welcome_email(user_email: str, user_name: str, role: str):
    """
    Sends a welcome email via SMTP.
    Skips execution if credentials are missing or email is invalid.
    """
    sender_email = os.getenv("EMAIL_SENDER")
    sender_password = os.getenv("EMAIL_PASSWORD")

    if not sender_email or not sender_password or not user_email:
        return

    msg = MIMEMultipart()
    msg['From'] = f"UzoAgro AI <{sender_email}>"
    msg['To'] = user_email
    msg['Subject'] = "Welcome to UzoAgro AI! 🌿"

    html_content = f"""
    <html>
        <body style="font-family: Arial, sans-serif; color: #143621;">
            <div style="padding: 20px; border: 1px solid #e2e8f0; border-radius: 10px;">
                <h2>Welcome, {user_name}!</h2>
                <p>You have successfully registered as a <strong>{role.capitalize()}</strong>.</p>
                <a href="https://uzoagro-ai.vercel.app/login.html" 
                   style="background-color: #D4ED6D; color: #143621; padding: 10px 20px; text-decoration: none; border-radius: 5px;">
                   Go to Dashboard
                </a>
            </div>
        </body>
    </html>
    """
    msg.attach(MIMEText(html_content, 'html'))


@app.post("/api/signup")
async def create_user(user_data: UserSignup, background_tasks: BackgroundTasks):
    """
    Creates a user account and triggers a background email task if an email is provided.
    """
    # Extract data securely from the Pydantic model
    role = user_data.role
    name = user_data.name
    phone = user_data.phone
    nin = user_data.nin
    primary_city = user_data.primary_city
    email = user_data.email
    pin = user_data.pin

    conn = get_db_connection()
    cursor = conn.cursor()

    try:
        # Check for duplicates
        cursor.execute("SELECT phone FROM users WHERE phone = %s", (phone,))
        if cursor.fetchone():
            raise HTTPException(status_code=400, detail="Phone number already registered.")

        # FIXED: Added the 7th %s placeholder to match the 7 columns
        cursor.execute(
            "INSERT INTO users (role, name, phone, nin, primary_city, email, pin) VALUES (%s, %s, %s, %s, %s, %s, %s)",
            (role, name, phone, nin, primary_city, email, pin)
        )

        # Insert driver logic if applicable
        if role == "driver":
            # Just a safety check: ensure allowed_crops is not null in your DB or handle it here
            cursor.execute(
                """INSERT INTO drivers (driver_id, driver_name, phone, current_city, home_base, capacity, allowed_crops)
                   VALUES (%s, %s, %s, %s, %s, %s, %s)""",
                (f"DRV-{phone[-4:]}", name, phone, primary_city, primary_city, 15.0, "All")
            )

        conn.commit()

        # Queue the email task
        if email and "@" in email:
            background_tasks.add_task(send_welcome_email, email, name, role)

        return {"status": "success", "message": "Account created successfully."}

    except Exception as e:
        conn.rollback()
        # This print helps you see the error in Hugging Face logs
        print(f"Signup Error: {str(e)}")
        raise HTTPException(status_code=500, detail=str(e))
    finally:
        cursor.close()
        conn.close()


# 3. The Fixed Login Route
@app.post("/api/login", response_class=JSONResponse)
def login_user(creds: UserLogin):
    """Verifies user credentials using PIN."""
    conn = get_db_connection()
    cursor = conn.cursor()

    try:
        # We select the columns we need
        cursor.execute(
            "SELECT role, name, phone, email, primary_city FROM users WHERE phone = %s AND pin = %s",
            (creds.phone, creds.pin)
        )
        user = cursor.fetchone()

        if user:
            # ROBUST FIX: Check if the database gave us a Dictionary or a Tuple
            if isinstance(user, dict):
                return JSONResponse(content={
                    "status": "success",
                    "user": {
                        "role": user['role'],
                        "name": user['name'],
                        "phone": user['phone'],
                        "email": user['email'],
                        "primary_city": user['primary_city']
                    }
                })
            else:
                # It's a Tuple (Access by Number: 0=role, 1=name, etc.)
                return JSONResponse(content={
                    "status": "success",
                    "user": {
                        "role": user[0],
                        "name": user[1],
                        "phone": user[2],
                        "email": user[3],
                        "primary_city": user[4]
                    }
                })
        else:
            raise HTTPException(status_code=401, detail="Invalid phone number or PIN.")

    except Exception as e:
        print(f"Login Error: {str(e)}")
        raise HTTPException(status_code=500, detail=f"Server Error: {str(e)}")
    finally:
        cursor.close()
        conn.close()


@app.post("/api/diagnose/image", response_class=JSONResponse)
async def analyze_crop_image(file: UploadFile = File(...)):
    """Receives a crop image from the frontend and passes it to the AI."""
    try:
        # Read the image file into memory
        image_bytes = await file.read()

        # Pass the raw image to your friend's wrapper
        results = run_botanical_diagnosis(image_bytes)

        return JSONResponse(content=results)
    except Exception as e:
        logger.exception("Failed to process uploaded image.")
        raise HTTPException(status_code=500, detail="Image processing failed.")


@app.post("/match/custom", response_class=JSONResponse)
def match_custom_request(req: FarmerRequest):
    """Saves the request to the DB and returns the top 3 AI matches."""
    if req.pickup_city not in CITIES or req.dropoff_city not in CITIES:
        raise HTTPException(status_code=400, detail="Invalid city selected")

    p_lat, p_lon = CITIES[req.pickup_city]
    d_lat, d_lon = CITIES[req.dropoff_city]

    # Generate a new unique request ID
    req_id = f"REQ-LIVE-{pd.Timestamp.now().strftime('%H%M%S')}"

    # 1. Save the new request permanently to the PostgreSQL database
    conn = get_db_connection()
    cursor = conn.cursor()

    # Crucial Fix: Changed SQLite's '?' to PostgreSQL's '%s'
    cursor.execute('''
                   INSERT INTO requests (request_id, sender_name, phone, pickup_city, pickup_lat, pickup_lon,
                                         dropoff_city, dropoff_lat, dropoff_lon, requested_date, required_capacity,
                                         crop_type)
                   VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s)
                   ''', (
                       req_id, req.farmer_name, "PENDING", req.pickup_city, p_lat, p_lon,
                       req.dropoff_city, d_lat, d_lon, req.requested_date, req.required_capacity, req.crop_type
                   ))
    conn.commit()

    # 2. Safely pull drivers OR inject live demo data if DB is empty
    cursor.execute("SELECT * FROM drivers")
    rows = cursor.fetchall()

    if not rows:
        # EMERGENCY DEMO DATA: Guarantees your pitch works perfectly on stage!
        drivers_df = pd.DataFrame([
            {"driver_id": "DRV-001", "driver_name": "Oluwaseun Logistics", "phone": "080111", "current_city": "Lagos",
             "current_lat": 6.5244, "current_lon": 3.3792, "home_base": "Kano", "capacity": 30.0,
             "allowed_crops": "Maize, Yam", "available_date": pd.Timestamp.now()},
            {"driver_id": "DRV-002", "driver_name": "Northern Freight", "phone": "080222", "current_city": "Ibadan",
             "current_lat": 7.3775, "current_lon": 3.9470, "home_base": "Kano", "capacity": 15.0,
             "allowed_crops": "Maize, Tomatoes", "available_date": pd.Timestamp.now()},
            {"driver_id": "DRV-003", "driver_name": "Emeka Transports", "phone": "080333", "current_city": "Abuja",
             "current_lat": 9.0765, "current_lon": 7.3986, "home_base": "Lagos", "capacity": 20.0,
             "allowed_crops": "Yam, Cassava", "available_date": pd.Timestamp.now()}
        ])
    else:
        # The bulletproof way to read PostgreSQL into Pandas without SQLAlchemy
        colnames = [desc[0] for desc in cursor.description]
        drivers_df = pd.DataFrame(rows, columns=colnames)
        drivers_df["available_date"] = pd.to_datetime(drivers_df["available_date"])

    conn.close()

    # 3. Create a DataFrame for this specific request to feed the AI Engine
    single_req_df = pd.DataFrame([{
        "request_id": req_id,
        "pickup_lat": p_lat,
        "pickup_lon": p_lon,
        "dropoff_lat": d_lat,
        "dropoff_lon": d_lon,
        "requested_date": pd.to_datetime(req.requested_date),
        "required_capacity": req.required_capacity,
        "crop_type": req.crop_type
    }])

    # 4. Run the matching intelligence
    try:
        matches_df = run_matching_engine(drivers_df, single_req_df, top_k=3)
        payload = jsonable_encoder(matches_df.to_dict(orient="records"))
        return JSONResponse(content=payload)
    except Exception as e:
        logger.exception("Error computing live matches")
        raise HTTPException(status_code=500, detail=str(e))