OpenCustomsGPT

OpenCustomsGPT is an advanced open-source tool designed to interact with Sydonia (Customs Management System) data through AI-powered agents. It enables customs officials and analysts to extract, analyze, and visualize customs data efficiently.

Table of Contents

• OpenCustomsGPT
  • Table of Contents
  • Features
  • Examples
  • Prerequisites
  • Installation
    • Production
    • Development
  • Configuration
  • Usage
  • Project Structure
  • Roadmap
  • Contributing
  • License
  • Contact

Features

• Sydonia Integration: Direct connection to Sydonia database for real-time data access
• Data Management:
  • Support for key Sydonia tables (expandable to full schema)
• AI-Powered Analysis:
  • Automated data extraction and processing
  • Advanced analytics and reporting
  • Custom query generation through natural language
• Flexible Architecture:
  • Multi-LLM provider support through LiteLLM

Examples

example_1.mov

example_2.mov

Prerequisites

• Python 3.10
• Poetry for dependency management
• PostgreSQL database
• OpenAI API key or other supported LLM provider credentials

Installation

Production

For production deployment, please refer to our dedicated installation guide at OpenCustomsGPT-install.

Development

1. Clone the repository:

   git clone https://github.com/FrancoisChastel/OpenCustomsGPT.git
   cd OpenCustomsGPT

2. Install dependencies with Poetry:

   poetry install

Configuration

1. Set up environment variables:

   export DB_URI="postgresql+psycopg2://user:password@localhost:5432/dbname"
   export OPENAI_API_KEY="your-api-key"
   export BASE_URL_LITELLM="http://0.0.0.0:4000"

2. Configure LLM provider:

   • Edit configs/litellm/config.yaml to set your preferred LLM provider
   • Available configurations include different agent types in configs/ directory:
     • coder_agent.yaml: For code-related tasks
     • normal_agent.yaml: For general purpose tasks
     • small_agent.yaml: For lightweight operations
     • thinking_agent.yaml: For complex reasoning tasks
     • sql_aent.yaml: For sql-related tasks

Usage

1. Start the LiteLLM proxy:

   poetry run litellm --config configs/litellm/config.yaml

2. Launch the Chainlit interface:

   poetry run chainlit run src/app.py

3. Access the application at http://localhost:8501

Project Structure

OpenCustomsGPT/
├── LICENSE                 # Project license file
├── README.md               # Project documentation and overview
├── configs/                # Configuration directory
│   ├── coder_agent.yaml    # Configuration for coding-focused AI agent
│   ├── litellm/            # LiteLLM configuration directory
│   │   └── config.yaml     # LiteLLM settings (LLM provider configs)
│   ├── normal_agent.yaml   # Configuration for standard AI agent
│   ├── small_agent.yaml    # Configuration for lightweight AI agent
│   ├── sql_agent.yaml    # Configuration for sql AI agent
│   └── thinking_agent.yaml # Configuration for reasoning-focused AI agent
└── dockerfile              # Container configuration
└── poetry.lock             # Poetry dependency lock file
└── pyproject.toml          # Python project configuration and dependencies
└── src/                    # Source code directory
│   ├── agents/             # AI agents implementation
│   │   ├── cache.py        # Caching mechanism for agents
│   │   ├── config.py       # Agent configuration handling
│   │   ├── executor.py     # Agent execution logic
│   │   ├── prompts.py      # Agent prompt templates
│   │   ├── setup.py        # Agent initialization
│   │   ├── sql.py          # Database interactions
│   │   └── tools.py        # Agent utilities and tools
│   ├── data_schema/        # Data structure definitions
│   │   ├── main.py         # Main schema handling
│   │   └── tmp/            # Temporary schema storage
│   │       └── schema.txt
│   ├── main.py             # Application entry point
│   └── render/             # Output rendering
│       ├── agents.py       # Agent response rendering
│       └── messages.py     # Message formatting
└── work_dir/               # Runtime working directory
    └── helpers/            # Helper functions
        └── tools.py        # Additional utility tools

Roadmap

• Implement dashboarding solution
• Enhance SQL query generation
• Create plugin system for extensions
• Add verified tooling for LLM used
• Add targeted data-schema extraction
• Implement authentification and SSO
• Add nested agentic behavior
• Add multi-language support (french as a priority)
• Add Oracle support

Contributing

Contributions are welcome! Please check our Contributing Guidelines for details on our code of conduct and how to submit pull requests.

License

This project is licensed under the AGPL License. See the LICENSE file for more information.

Contact

For questions or feedback, please reach out at [email protected].