mcpField
E
expense-report-generator

expense-report-generator

github.com/dkyazzentwatwaDocuments12471 installsUpdated 29w ago
Repository

About expense-report-generator

expense-report-generator creates PDF expense reports from CSV, JSON, or manually entered data for employees and project teams. It adds expense categorization, policy checks, approval status, and reimbursement tracking to report generation.

Key features

  1. 01Accepts CSV, JSON, and manual entries
  2. 02Categorizes expenses and calculates totals
  3. 03Links receipts to expense items
  4. 04Flags policy issues and missing receipts
  5. 05Tracks approval and reimbursement status
  6. 06Exports formatted PDF reports

Use cases

  1. 01Prepare an employee's monthly expense report
  2. 02Check expenses against configured limits
  3. 03Track project expenses and reimbursement status

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/dkyazzentwatwa/chatgpt-skills --skill expense-report-generator

After installing, type/expense-report-generatorin Claude Code to confirm it loaded.

Bundled filesscripts/expense_report.pyscripts/requirements.txt

Details

From the project README

Expense Report Generator

Create professional expense reports from receipt data with automatic categorization and totals.

Features

  • Multiple Input Formats: CSV, JSON, or manual entry
  • Auto-Categorization: Classify expenses by type
  • Receipt Tracking: Link receipts to expenses
  • Approval Workflow: Status tracking and approver info
  • Policy Compliance: Flag out-of-policy expenses
  • PDF Export: Professional formatted reports
  • Reimbursement Calculation: Track paid/unpaid amounts

Quick Start

from expense_report import ExpenseReportGenerator

report = ExpenseReportGenerator()

# Set report info
report.set_employee("John Doe", "EMP001", "Engineering")
report.set_period("2024-01-01", "2024-01-31")

# Add expenses
report.add_expense(
    date="2024-01-15",
    description="Client dinner",
    category="Meals",
    amount=125.50,
    receipt="receipt_001.jpg"
)

report.add_expense(
    date="2024-01-18",
    description="Uber to airport",
    category="Transportation",
    amount=45.00
)

# Generate report
report.generate_pdf("expense_report.pdf")

CLI Usage

# From CSV
python expense_report.py --input expenses.csv --employee "John Doe" --output report.pdf

# With date range
python expense_report.py --input data.csv --start 2024-01-01 --end 2024-01-31 -o report.pdf

# Set department and approver
python expense_report.py --input data.csv --employee "Jane Smith" --dept Sales \
    --approver "Bob Manager" -o report.pdf

# With policy limits
python expense_report.py --input data.csv --policy policy.json -o report.pdf

Input Format

CSV Format
date,description,category,amount,receipt,notes
2024-01-15,Client dinner at Restaurant,Meals,125.50,receipt_001.jpg,Met with ABC Corp
2024-01-16,Uber to client site,Transportation,32.00,,
2024-01-17,Office supplies,Supplies,45.99,receipt_002.jpg,
2024-01-18,Flight to NYC,Travel,450.00,flight_confirm.pdf,Project kickoff
JSON Format
{
  "employee": "John Doe",
  "employee_id": "EMP001",
  "department": "Engineering",
  "period": {"start": "2024-01-01", "end": "2024-01-31"},
  "expenses": [
    {
      "date": "2024-01-15",
      "description": "Client dinner",
      "category": "Meals",
      "amount": 125.50,
      "receipt": "receipt_001.jpg"
    }
  ]
}
Policy Configuration
{
  "limits": {
    "Meals": 75,
    "Transportation": 100,
    "Lodging": 250,
    "Supplies": 200
  },
  "requires_receipt": 25,
  "requires_approval": 500,
  "prohibited": ["Alcohol", "Personal items"]
}

API Reference

ExpenseReportGenerator Class
class ExpenseReportGenerator:
    def __init__(self)

    # Report Setup
    def set_employee(self, name: str, employee_id: str = None,
                    department: str = None) -> 'ExpenseReportGenerator'
    def set_period(self, start: str, end: str) -> 'ExpenseReportGenerator'
    def set_approver(self, name: str, title: str = None) -> 'ExpenseReportGenerator'
    def set_project(self, project_name: str, project_code: str = None) -> 'ExpenseReportGenerator'

    # Adding Expenses
    def add_expense(self, date: str, description: str, category: str,
                   amount: float, receipt: str = None, notes: str = None,
                   reimbursable: bool = True) -> 'ExpenseReportGenerator'
    def load_csv(self, filepath: str) -> 'ExpenseReportGenerator'
    def load_json(self, filepath: str) -> 'ExpenseReportGenerator'

    # Policy
    def set_policy(self, policy: Dict) -> 'ExpenseReportGenerator'
    def check_compliance(self) -> List[Dict]

    # Analysis
    def get_summary(self) -> Dict
    def by_category(self) -> Dict[str, float]
    def by_date(self) -> Dict[str, float]
    def get_total(self) -> float

    # Export
    def generate_pdf(self, output: str) -> str
    def generate_html(self, output: str) -> str
    def to_csv(self, output: str) -> str
    def to_json(self, output: str) -> str

Expense Categories

Standard categories:

  • Meals: Business meals and entertainment
  • Transportation: Taxi, rideshare, rental car, parking
  • Travel: Flights, trains, hotels
  • Lodging: Hotel, accommodation
  • Supplies: Office supplies, equipment
  • Communication: Phone, internet
  • Professional: Conferences, training, memberships
  • Other: Miscellaneous expenses

Report Summary

summary = report.get_summary()
# Returns:
# {
#     "employee": "John Doe",
#     "period": {"start": "2024-01-01", "end": "2024-01-31"},
#     "total_expenses": 1250.50,
#     "expense_count": 15,
#     "categories": {
#         "Meals": 325.00,
#         "Transportation": 180.50,
#         "Travel": 650.00,
#         "Supplies": 95.00
#     },
#     "reimbursable": 1150.50,
#     "non_reimbursable": 100.00,
#     "receipts_attached": 12,
#     "receipts_missing": 3
# }

Policy Compliance

# Set spending limits
report.set_policy({
    "limits": {
        "Meals": 75,      # Per transaction limit
        "Daily_meals": 100  # Daily limit
    },
    "requires_receipt": 25,  # Receipts required above this
    "requires_approval": 500  # Manager approval above this
})

# Check compliance
violations = report.check_compliance()
# Returns:
# [
#     {"expense_id": 3, "type": "over_limit", "category": "Meals",
#      "amount": 125.50, "limit": 75},
#     {"expense_id": 7, "type": "missing_receipt", "amount": 45.00}
# ]

Generated Report Contents

The PDF report includes:

  1. Header

    • Company logo (optional)
    • Report title and date range
    • Employee information
  2. Summary Section

    • Total amount
    • Category breakdown
    • Reimbursement status
  3. Expense Details Table

    • Date, description, category
    • Amount, receipt status
    • Notes
  4. Category Charts

    • Pie chart of spending by category
    • Daily spending bar chart
  5. Compliance Notes

    • Policy violations (if any)
    • Missing receipts
  6. Approval Section

    • Employee signature line
    • Approver signature line
    • Date fields

Example Workflows

Monthly Employee Report
report = ExpenseReportGenerator()
report.set_employee("Sarah Johnson", "EMP042", "Marketing")
report.set_period("2024-02-01", "2024-02-29")
report.set_approver("Mike Director", "VP Marketing")

# Load from tracking spreadsheet
report.load_csv("february_expenses.csv")

# Check policy
violations = report.check_compliance()
if violations:
    print(f"Warning: {len(violations)} policy violations")

# Generate report
report.generate_pdf("sarah_feb_expenses.pdf")
print(f"Total: ${report.get_total():,.2f}")
Project Expense Tracking
report = ExpenseReportGenerator()
report.set_employee("Project Team")
report.set_project("Website Redesign", "PRJ-2024-001")
report.set_period("2024-01-01", "2024-03-31")

# Add project expenses
report.add_expense("2024-01-15", "Design software license", "Software", 299.00)
report.add_expense("2024-02-01", "User testing incentives", "Research", 500.00)
report.add_expense("2024-02-20", "Stock photos", "Creative", 150.00)

summary = report.by_category()
print("Project Expenses by Category:")
for cat, amount in summary.items():
    print(f"  {cat}: ${amount:,.2f}")

Dependencies

  • pandas>=2.0.0
  • reportlab>=4.0.0
  • matplotlib>=3.7.0

Comments

Markdown supported