🧾 AI-Powered Invoice Parser

A modular Python application that leverages Large Language Models (LLMs) like LLaMA 2 to extract structured data from PDF invoices. It features a user-friendly Streamlit web interface and supports custom prompt-based extraction, Pydantic validation, and multi-invoice parsing.

---

📁 Project Structure

Invoice-Parser-LLM/
├── app/ # Streamlit UI components
│ ├── main.py # App launcher
│ └── layout.py # Streamlit UI layout
│
├── core/ # Core logic
│ ├── parser.py # PDF parsing & text extraction
│ ├── prompt_templates.py # LLM prompt templates
│ └── validator.py # Data validation with Pydantic
│
├── models/ # LLM integration
│ └── llama_model.py # Load & run LLaMA from Hugging Face
│
├── data/sample_invoices/ # Sample invoice PDFs
├── outputs/extracted_data/ # Parsed JSON outputs
├── tests/test_parser.py # Unit tests
│
├── requirements.txt # Dependencies
├── .env # API keys / tokens
└── README.md # This file

---

🚀 Features

• 📄 PDF Invoice Upload – Upload invoices via a web interface.
• 💬 Custom Prompting – Use natural language to define extraction logic.
• 🧠 LLM Integration – Built with LLaMA 2 and Hugging Face transformers.
• 📦 Structured Output – Data returned in validated JSON format.
• ✅ Data Validation – Uses Pydantic to ensure schema correctness.
• 🧩 Modular Codebase – Scalable and easy to maintain.

---

⚙️ Setup & Installation

1. Clone the Repository

git clone https://github.com/your-username/Invoice-Parser-LLM.git

2. Create a Virtual Environment

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4.Authenticate with Hugging Face

Request Access: LLaMA-2 on Hugging Face

Login via CLI:

huggingface-cli login

▶️ How to Run the App Ensure you're in the project root and your virtual environment is active

python -m streamlit run app/main.py