OCR-to-LLM System

Overview

This project is an OCR (Optical Character Recognition) to LLM (Language Model) processing system. It extracts text from images or documents, processes the text using a pre-trained language model (LLAMA 3.1), and returns the processed text. The system is designed to be modular, scalable, and easily deployable on an on-site server.

Project Structure

project_root/
│
├── src/
│   ├── ocr/
│   │   ├── __init__.py
│   │   └── ocr_service.py
│   │
│   ├── llm/
│   │   ├── __init__.py
│   │   └── llama_service.py
│   │
│   ├── pipeline/
│   │   ├── __init__.py
│   │   └── pipeline_service.py
│   │
│   ├── webapp/
│   │   ├── __init__.py
│   │   └── main.py
│   │
│   ├── utils/
│   │   ├── __init__.py
│   │   ├── logger.py
│   │   └── config.py
│   │
│   └── __init__.py
│
├── data/
│   ├── input_images/
│   └── processed_text/
│
├── tests/
│   ├── __init__.py
│   ├── test_ocr_service.py
│   ├── test_llama_service.py
│   ├── test_pipeline_service.py
│   └── test_routes.py
│
├── docs/
│   ├── requirements.txt
│   ├── README.md
│   └── architecture_diagram.png
│
├── scripts/
│   ├── deploy.sh
│   └── start_dev.sh
│
├── .env
├── .gitignore
├── Dockerfile
└── docker-compose.yml

Requirements

Python Packages

• Python 3.8+
• FastAPI
• Uvicorn
• Gunicorn
• EasyOCR
• PaddleOCR
• Transformers
• Celery
• Redis
• SQLAlchemy
• Psycopg2-binary
• Docker
• Docker Compose

System Requirements

• Docker installed and running
• PostgreSQL database
• Redis for task queuing
• On-site server with sufficient resources for LLM processing

Setup Instructions

Step 1: Clone the Repository

git clone <repository-url>
cd ocr-to-llm

Step 2: Set Up the Python Environment

python3 -m venv venv
source venv/bin/activate
pip install -r docs/requirements.txt

Step 3: Set Up the Database

Ensure PostgreSQL is installed and running. Create a database and user:

CREATE DATABASE ocr_llm_db;
CREATE USER ocr_user WITH ENCRYPTED PASSWORD 'your_password';
GRANT ALL PRIVILEGES ON DATABASE ocr_llm_db TO ocr_user;

Step 4: Configure the Environment Variables

Create a .env file in the project root with the following contents:

DATABASE_URL=postgresql://ocr_user:your_password@localhost/ocr_llm_db
REDIS_URL=redis://localhost:6379/0

Step 5: Run the Application Locally

uvicorn src.webapp.main:app --reload

Step 6: Run Celery Worker

Start the Celery worker to handle background tasks:

celery -A src.pipeline.pipeline_service worker --loglevel=info

Step 7: Dockerize the Application

To build and run the application in Docker containers, use:

docker-compose up --build

Usage

• Access the application at http://localhost:8000.
• Upload images to extract text, which will be processed by the LLM.

Testing

Run unit tests with:

pytest

Contributing

Contributions are welcome! Please submit a pull request or open an issue to discuss any changes.

License

This project is licensed under the MIT License.