Installation
This guide covers installing the Proxy Base Agent (PBA) and its dependencies.
Prerequisites
- Python: Version 3.10 or higher.
- Operating System: Linux, macOS, or Windows.
- LLM Backend: You need a compatible local LLM setup. PBA currently supports:
- Hardware: Requirements depend heavily on the LLM you choose to run locally. Ensure your system meets the minimum requirements for your selected model.
Installation Methods
Choose the method that best suits your needs:
Method 1: Using pip
(Recommended)
This is the simplest way to install the latest stable release of PBA from PyPI.
# Install the core PBA package
pip install proxy-base-agent
# --- Framework Extras ---
# Install extras for the LLM backend you intend to use:
# For MLX (Apple Silicon):
pip install proxy-base-agent[mlx]
# For PyTorch:
pip install proxy-base-agent[torch]
# (Ensure you have a compatible PyTorch version installed separately if needed)
Method 2: Using uv
(Fast Alternative)
If you use uv
, Astral's fast package manager:
# Install the core PBA package
uv pip install proxy-base-agent
# --- Framework Extras ---
# For MLX (Apple Silicon):
uv pip install proxy-base-agent[mlx]
# For PyTorch:
uv pip install proxy-base-agent[torch]
Method 3: From Source (Development / Latest Features)
Install directly from the GitHub repository for development or to get the absolute latest code (potentially unstable).
-
Clone the Repository:
git clone https://github.com/TheProxyCompany/proxy-base-agent.git cd proxy-base-agent
-
Install in Editable Mode: We recommend using a virtual environment.
- Using
pip
:# Install core package and common dev dependencies pip install -e ".[dev]" # Install specific framework extras if needed pip install -e ".[mlx]" # or pip install -e ".[torch]"
- Using
uv
:The# Install core package and common dev dependencies uv pip install -e ".[dev]" # Install specific framework extras if needed uv pip install -e ".[mlx]" # or uv pip install -e ".[torch]"
-e
flag installs the package in "editable" mode, meaning changes you make to the source code will be reflected immediately when you run the agent.
- Using
Verifying Installation
After installation, you should be able to run the agent's setup wizard:
python -m agent
If the wizard starts without import errors, the installation was successful.
Next Steps
- Follow the Quickstart Guide to run the agent and interact with it.
- Explore the Core Concepts to understand how PBA works.
If you encounter issues, please open an issue on GitHub.