Atomic Agents Introduction
Atomic Agents is a framework designed to be modular, extensible, and easy to use. It is built on the concept of Atomic Design, aiming to create components that are small and single-purpose. This framework provides a set of tools and agents that can be combined to develop powerful applications. It utilizes Instructor for its foundation and leverages Pydantic for data validation and serialization.
Atomic Agents Features
Modular and Extensible
Atomic Agents is designed with modularity in mind. Each component in the framework is as small and single-purpose as possible, allowing for easy extension and customization.
Built on Top of Instructor
The framework is built on top of Instructor, which allows it to leverage the power of various AI APIs such as OpenAI, Cohere, Anthropic, Gemini, and more.
Pydantic for Data Validation and Serialization
Atomic Agents uses Pydantic for data validation and serialization, ensuring that data handling is efficient and error-free.
Quick and Easy Installation
To install Atomic Agents, you can simply use pip:
bash pip install atomic-agents
Alternatively, you can install the necessary dependencies from the repository using the following commands:
bash
python -m venv venv
source venv/bin/activate # On Windows, use venv\Scripts\activate.bat
pip install -r requirements.txt
pip install -e .
Atomic Agents Usage Examples
Quickstart Guide
A quickstart guide is available in the quickstart notebook, providing a step-by-step introduction to using Atomic Agents.
Thorough Documentation
The framework comes with comprehensive documentation in the docs directory, including API documentation and general guides on how to create new tools.
Examples Directory
All examples are located in the examples directory, with each example thoroughly documented for clarity and ease of understanding.
Atomic Agents Instructor & Model Compatibility
Atomic Agents is compatible with various AI models and APIs. It depends on the Instructor package, which supports multiple AI APIs like OpenAI, Cohere, Anthropic, Gemini, and more. Additionally, Atomic Agents can work with Ollama or LMStudio, with the option to set the mode to JSON if your local server does not support tool-calling.
Atomic Agents Contributing Guidelines
Contributions to Atomic Agents are welcome. If you wish to contribute, follow these steps:
- Fork the repository.
- Create a new branch.
- Make your changes.
- Commit and push your changes.
- Open a pull request.
Atomic Agents License
Atomic Agents is licensed under the MIT License. For more details, refer to the LICENSE file in the repository.
Atomic Agents Philosophy
The Atomic Agents framework is inspired by Atomic Design, aiming to create small, reusable components that can be combined to build complex applications. The framework's philosophy is to provide a set of tools and agents that are easy to use, allowing developers to focus on innovation and creativity.
For a more detailed deep-dive into Atomic Agents, you can find an article on Medium.
Atomic Agents FAQs
Q: What is the philosophy behind Atomic Agents?
A: The philosophy behind Atomic Agents is to create a modular, extensible, and easy-to-use framework inspired by Atomic Design. It aims to provide small, reusable components that can be combined to build powerful applications.
Q: How can I contribute to Atomic Agents?
A: To contribute to Atomic Agents, fork the repository, create a new branch, make your changes, commit and push them, and finally, open a pull request.
Q: Which AI models and APIs are compatible with Atomic Agents?
A: Atomic Agents is compatible with various AI models and APIs supported by the Instructor package, including OpenAI, Cohere, Anthropic, Gemini, and more.
Q: How can I install Atomic Agents?
A: You can install Atomic Agents using pip:
bash pip install atomic-agents
Alternatively, you can install the necessary dependencies from the repository by following the instructions in the Quickstart section.
Q: Where can I find the documentation for Atomic Agents?
A: The documentation for Atomic Agents can be found in the docs directory of the repository. It includes API documentation and general guides on how to use and extend the framework.
