app.py

#Fast APi Packages
from fastapi import FastAPI, File, HTTPException, status
from fastapi.responses import JSONResponse
from pydantic import BaseModel
import json
from typing import List, Dict, Any
import pandas as pd
import numpy as np
from sklearn.metrics.pairwise import cosine_similarity
from scipy import sparse
from datetime import datetime
import warnings
import os
import logging
import requests
import io


warnings.filterwarnings('ignore')

# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

app = FastAPI()

# URL of the Excel file
EXCEL_URL = "https://huggingface.co/spaces/Vaibhav84/RecommendationAPI/resolve/main/DataSetSample.xlsx"

try:
    # Download the file from URL
    logger.info(f"Attempting to download Excel file from: {EXCEL_URL}")
    response = requests.get(EXCEL_URL)
    response.raise_for_status()  # Raises an HTTPError if the status is 4xx, 5xx
    
    # Read the Excel file from the downloaded content
    excel_content = io.BytesIO(response.content)
    purchase_history = pd.read_excel(excel_content, sheet_name='Transaction History', 
                                   parse_dates=['Purchase_Date'])
    logger.info("Successfully downloaded and loaded Excel file")
    
    # Process the data
    purchase_history['Customer_Id'] = purchase_history['Customer_Id'].astype(str)
    product_categories = purchase_history[['Product_Id', 'Category']].drop_duplicates().set_index('Product_Id')['Category'].to_dict()
    purchase_counts = purchase_history.groupby(['Customer_Id', 'Product_Id']).size().unstack(fill_value=0)
    sparse_purchase_counts = sparse.csr_matrix(purchase_counts)
    cosine_similarities = cosine_similarity(sparse_purchase_counts.T)
    
    logger.info("Data processing completed successfully")
    
except Exception as e:
    logger.error(f"Error downloading or processing data: {str(e)}")
    raise

def get_customer_items_and_recommendations(user_id: str, n: int = 5) -> tuple[List[Dict], List[Dict]]:
    """
    Get both purchased items and recommendations for a user
    """
    user_id = str(user_id)
    
    if user_id not in purchase_counts.index:
        return [], []
    
    purchased_items = list(purchase_counts.columns[purchase_counts.loc[user_id] > 0])
    
    purchased_items_info = []
    user_purchases = purchase_history[purchase_history['Customer_Id'] == user_id]
    
    for item in purchased_items:
        item_purchases = user_purchases[user_purchases['Product_Id'] == item]
        total_amount = float(item_purchases['Amount (In Dollars)'].sum())
        last_purchase = pd.to_datetime(item_purchases['Purchase_Date'].max())
        category = product_categories.get(item, 'Unknown')
        purchased_items_info.append({
            'product_id': item,
            'category': category,
            'total_amount': total_amount,
            'last_purchase': last_purchase.strftime('%Y-%m-%d')
        })
    
    user_idx = purchase_counts.index.get_loc(user_id)
    user_history = sparse_purchase_counts[user_idx].toarray().flatten()
    similarities = cosine_similarities.dot(user_history)
    purchased_indices = np.where(user_history > 0)[0]
    similarities[purchased_indices] = 0
    recommended_indices = np.argsort(similarities)[::-1][:n]
    recommended_items = list(purchase_counts.columns[recommended_indices])
    recommended_items = [item for item in recommended_items if item not in purchased_items]
    
    recommended_items_info = [
        {
            'product_id': item,
            'category': product_categories.get(item, 'Unknown')
        }
        for item in recommended_items
    ]

    return purchased_items_info, recommended_items_info

@app.get("/")
async def root():
    return {
        "message": "Welcome to the Recommendation API",
        "status": "running",
        "data_loaded": purchase_history is not None
    }

@app.get("/recommendations/{customer_id}")
async def get_recommendations(customer_id: str, n: int = 5):
    """
    Get recommendations for a customer
    
    Parameters:
    - customer_id: The ID of the customer
    - n: Number of recommendations to return (default: 5)
    
    Returns:
    - JSON object containing purchase history and recommendations
    """
    try:
        purchased_items, recommended_items = get_customer_items_and_recommendations(customer_id, n)
        
        return {
            "customer_id": customer_id,
            "purchase_history": purchased_items,
            "recommendations": recommended_items
        }
    except Exception as e:
        logger.error(f"Error processing request for customer {customer_id}: {str(e)}")
        raise HTTPException(status_code=404, detail=f"Error processing customer ID: {customer_id}. {str(e)}")

@app.get("/health")
async def health_check():
    """
    Health check endpoint that returns system information
    """
    return {
        "status": "healthy",
        "data_loaded": purchase_history is not None,
        "number_of_customers": len(purchase_counts.index) if purchase_history is not None else 0,
        "number_of_products": len(purchase_counts.columns) if purchase_history is not None else 0
    }

@app.post("/login")
async def login(customer_id: str, password: str):
    """
    Login endpoint to validate customer ID and password
    
    Parameters:
    - customer_id: The ID of the customer to validate
    - password: Password (first three chars of customer_id + "123")
    
    Returns:
    - JSON object containing login status and message
    """
    try:
        # Convert customer_id to string to match the format in purchase_history
        customer_id = str(customer_id)
        
        # Generate expected password (first three chars + "123")
        expected_password = f"{customer_id[:3]}123"
        
        # Check if customer exists and password matches
        if customer_id in purchase_history['Customer_Id'].unique():
            if password == expected_password:
                # Get customer's basic information
                customer_data = purchase_history[purchase_history['Customer_Id'] == customer_id]
                total_purchases = len(customer_data)
                total_spent = customer_data['Amount (In Dollars)'].sum()
                
                # Convert last purchase date to datetime if it's not already
                last_purchase = pd.to_datetime(customer_data['Purchase_Date'].max())
                last_purchase_str = last_purchase.strftime('%Y-%m-%d')
                
                return JSONResponse(
                    status_code=status.HTTP_200_OK,
                    content={
                        "status": "success",
                        "message": "Login successful",
                        "customer_id": customer_id,
                        "customer_stats": {
                            "total_purchases": total_purchases,
                            "total_spent": float(total_spent),
                            "last_purchase_date": last_purchase_str
                        }
                    }
                )
            else:
                return JSONResponse(
                    status_code=status.HTTP_401_UNAUTHORIZED,
                    content={
                        "status": "error",
                        "message": "Invalid password"
                    }
                )
        else:
            return JSONResponse(
                status_code=status.HTTP_401_UNAUTHORIZED,
                content={
                    "status": "error",
                    "message": "Invalid customer ID"
                }
            )
            
    except Exception as e:
        logger.error(f"Error during login for customer {customer_id}: {str(e)}")
        raise HTTPException(
            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
            detail=f"Error during login process: {str(e)}"
        )