🧾 Invoice Extractor

Extract structured data from invoices and receipts using a local LLM — no cloud, no API keys, 100% private.

Architecture

┌──────────────────────────────────────────────┐
│         Invoice Input (text / file)          │
│                    │                          │
│            ┌──────▼───────┐                  │
│            │   Invoice    │                  │
│            │   Parser     │                  │
│            └──────┬───────┘                  │
│                   │                          │
│            ┌──────▼───────┐                  │
│            │   Ollama     │                  │
│            │  (Gemma 4)   │                  │
│            └──────┬───────┘                  │
│     ┌─────────────┼─────────────┐            │
│ ┌───▽────┐  ┌─────▽────┐  ┌────▽─────┐     │
│ │Vendor  │  │  Line    │  │Category  │     │
│ │Details │  │  Items   │  │Classifier│     │
│ └────────┘  └──────────┘  └──────────┘     │
│                   │                          │
│     ┌─────────────┼─────────────┐            │
│ ┌───▽────┐  ┌─────▽────┐  ┌────▽─────┐     │
│ │  JSON  │  │   CSV    │  │ Batch    │     │
│ │ Export │  │  Export  │  │ Report   │     │
│ └────────┘  └──────────┘  └──────────┘     │
└──────────────────────────────────────────────┘

Features

1. Structured Data Extraction — Parses vendor details, line items, totals, tax, and payment terms from raw invoice text
2. Automatic Categorization — Classifies line items into categories like Software, Hardware, Travel, and Consulting
3. Batch Processing — Process multiple invoices in one run with configurable batch sizes
4. Duplicate Detection — Identifies near-duplicate invoices using configurable similarity thresholds
5. Multi-Format Export — Output extracted data as JSON or CSV for downstream accounting systems
6. Rich CLI Interface — Beautiful terminal output with color-coded tables and progress indicators
7. Streamlit Web UI — Browser-based dashboard for uploading invoices and reviewing extracted data
8. FastAPI REST Endpoint — Programmatic API access for integrating into automated AP workflows
9. Docker Ready — Full Docker and Docker Compose support for containerized deployment
10. 100% Local & Private — All inference runs through Ollama locally; sensitive financial data never leaves your machine

Quick Start

Prerequisites

• Python 3.11 or higher
• Ollama installed and running
• Gemma 4 model pulled: ollama pull gemma4

Installation

git clone https://github.com/kennedyraju55/invoice-extractor.git
cd invoice-extractor
pip install -r requirements.txt

Running the Application

CLI:

python -m src.invoice_extractor.cli extract --file invoice.txt

Web UI:

streamlit run src/invoice_extractor/web_ui.py

Docker:

docker-compose up

Tech Stack

Technology	Purpose
Gemma 4 + Ollama	Local LLM inference for invoice parsing and categorization
Click + Rich	CLI framework with formatted tables and progress bars
Streamlit	Interactive web dashboard for invoice uploads

Project Structure

• src/invoice_extractor/ — Core application: parsing engine, categorizer, CLI, web UI, API
• common/ — Shared LLM client utilities for Ollama integration
• tests/ — Unit and integration test suite
• docs/ — Documentation and architecture diagrams
• examples/ — Sample invoices and expected extraction outputs

Author

Nrk Raju Guthikonda — GitHub: kennedyraju55 · Dev.to · LinkedIn