TaxHacker — Self-Hosted AI Accounting for Receipts and Invoices

Introduction

TaxHacker is a self-hosted, AI-powered accounting tool that automates the tedious work of processing receipts, invoices, and financial transactions. Upload a document and the LLM extracts vendor, amount, date, currency, and category, then organizes everything into a searchable, filterable ledger. It is built for freelancers, contractors, and small business owners who want to keep their financial data private and off third-party cloud services.

What TaxHacker Does

• Processes receipts and invoices via OCR and LLM analysis to extract structured financial data
• Categorizes expenses automatically using customizable prompts and category schemas
• Tracks income and expenses across multiple currencies with automatic exchange rate conversion
• Provides a filterable ledger view with date ranges, categories, and full-text search
• Exports data to CSV for import into spreadsheets or traditional accounting software

Architecture Overview

TaxHacker is a TypeScript application with a Next.js front-end and a Node.js back-end. Document processing works in two stages: first, OCR extracts text from uploaded images or PDFs; then the configured LLM (OpenAI, Google Gemini, or a local Ollama instance) analyzes the text using customizable prompts to produce structured JSON with amount, vendor, date, and category. Results are stored in a PostgreSQL database. The entire stack runs in Docker containers.

Self-Hosting & Configuration

• Deploy with Docker Compose; configure the LLM provider and API key in the .env file
• Supports OpenAI, Google Gemini, and local Ollama models for fully offline operation
• Customize category schemas and extraction prompts to match your accounting needs
• Set base currency and enable automatic exchange rate fetching for multi-currency transactions
• Back up the PostgreSQL database and uploaded document files for data safety

Key Features

• AI-powered receipt scanning that extracts vendor, amount, date, and category from photos and PDFs
• Customizable LLM prompts so you can tune extraction logic for your specific document types
• Multi-currency support with automatic conversion to your base currency
• Full data privacy when paired with a local Ollama model for zero-cloud processing
• Clean ledger interface with filtering, search, and CSV export for tax preparation

Comparison with Similar Tools

• Firefly III — comprehensive personal finance manager with budgeting and bank imports; TaxHacker focuses specifically on AI-driven document processing and expense extraction
• Actual Budget — local-first budgeting app; TaxHacker targets receipt and invoice processing rather than budget planning
• Invoice Ninja — invoicing and payment platform for sending invoices to clients; TaxHacker is for processing incoming receipts and tracking your own expenses
• Paperless-ngx — document management with OCR; TaxHacker adds AI-powered financial data extraction on top of document ingestion
• Spreadsheets — flexible but manual; TaxHacker automates the data entry step with AI extraction and categorization

FAQ

Q: Can I use TaxHacker completely offline? A: Yes. Configure Ollama as the LLM provider and run a local model. All OCR and AI processing happens on your machine with no data leaving your network.

Q: What document formats are supported? A: TaxHacker accepts JPEG, PNG, and PDF uploads. The OCR engine processes images and scanned documents, while the LLM handles text extraction from both.

Q: How accurate is the AI extraction? A: Accuracy depends on document quality and the LLM model used. GPT-4 class models achieve high accuracy on standard receipts. You can review and correct any extraction before saving.

Q: Does it replace proper accounting software? A: TaxHacker is a data capture and categorization tool. It exports CSV for import into accounting software like GnuCash, QuickBooks, or spreadsheets for final tax filing.

Sources

• https://github.com/vas3k/TaxHacker
• https://taxhacker.app