app.py

import gradio as gr
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from transformers import pipeline
import plotly.express as px

# Initialize the Hugging Face model for expense categorization (use zero-shot classification)
expense_classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")

# Batch categorization function for efficiency
def categorize_transaction_batch(descriptions):
    candidate_labels = ["Groceries", "Entertainment", "Rent", "Utilities", "Dining", "Transportation", "Shopping", "Others"]
    return [expense_classifier(description, candidate_labels)["labels"][0] for description in descriptions]

# Function to process the uploaded CSV and generate visualizations
def process_expenses(file):
    # Read CSV data
    df = pd.read_csv(file.name)
    
    # Check if required columns are present
    if 'Date' not in df.columns or 'Description' not in df.columns or 'Amount' not in df.columns:
        return "CSV file should contain 'Date', 'Description', and 'Amount' columns."

    # Categorize the expenses (using batch processing to minimize model calls)
    df['Category'] = categorize_transaction_batch(df['Description'].tolist())

    # Create visualizations:
    # 1. Pie chart for Category-wise spending
    category_spending = df.groupby("Category")['Amount'].sum()
    fig1 = px.pie(category_spending, names=category_spending.index, values=category_spending.values, title="Category-wise Spending")
    
    # 2. Monthly spending trends (Line plot)
    df['Date'] = pd.to_datetime(df['Date'])
    df['Month'] = df['Date'].dt.to_period('M')
    monthly_spending = df.groupby('Month')['Amount'].sum()
    fig2 = px.line(monthly_spending, x=monthly_spending.index, y=monthly_spending.values, title="Monthly Spending Trends")
    
    # 3. Budget vs Actual Spending (Bar chart)
    category_list = df['Category'].unique()
    budget_dict = {category: 500 for category in category_list}  # Default budget is 500 for each category
    budget_spending = {category: [budget_dict[category], category_spending.get(category, 0)] for category in category_list}
    budget_df = pd.DataFrame(budget_spending, index=["Budget", "Actual"]).T
    fig3 = px.bar(budget_df, x=budget_df.index, y=["Budget", "Actual"], title="Budget vs Actual Spending")
    
    # 4. Suggested savings (only calculate if over budget)
    savings_tips = []
    for category, actual in category_spending.items():
        if actual > budget_dict.get(category, 500):
            savings_tips.append(f"- **{category}**: Over budget by ${actual - budget_dict.get(category, 500)}. Consider reducing this expense.")
    
    return df.head(), fig1, fig2, fig3, savings_tips

# Gradio interface definition
inputs = gr.File(label="Upload Expense CSV")
outputs = [
    gr.Dataframe(label="Categorized Expense Data"),
    gr.Plot(label="Category-wise Spending (Pie Chart)"),
    gr.Plot(label="Monthly Spending Trends (Line Chart)"),
    gr.Plot(label="Budget vs Actual Spending (Bar Chart)"),
    gr.Textbox(label="Savings Tips")
]

# Launch Gradio interface
gr.Interface(
    fn=process_expenses,
    inputs=inputs,
    outputs=outputs,
    live=True,
    title="Smart Expense Tracker",
    description="Upload your CSV of transactions, categorize them, and view insights like spending trends and budget analysis."
).launch()