The BESSER Agentic Framework (BAF) is part of the BESSER (Building Better Smart Software Faster) project. It aims to make the design and implementation of agents, bots and chatbots easier and accessible for everyone.

Check out the official documentation.

Quick start

Requirements

• Python 3.11
• Recommended: Create a virtual environment (e.g. venv, conda)
• Install the package:

pip install besser-agentic-framework

This command will install the base package with the core dependencies, but will omit some optional dependencies.

You can add the following tags to the installation:

• extras: It will install the necessary dependencies for some additional agent functionalities (e.g., RAG, Speech-to-Text, plotly, opencv).
• llms: Necessary dependencies to run LLMs (openai, replicate, transformers)
• tensorflow: Necessary for the SimpleIntentClassifier. Since tensorflow is a very heavy package, this allows to install it only if necessary
• torch: To install PyTorch (necessary for some HuggingFace models)
• docs: Dependencies to compile the project documentation (the one you are reading now)
• all: It installs all the dependencies at once

This is how you would install the package with additional dependencies:

  pip install besser-agentic-framework[extras,llms,tensorflow]

If you cloned this repository, you can install the dependencies in 2 ways:

pip install -e .[extras]

or by referencing to the requirements files:

pip install -r requirements/requirements-extras.txt

Note that if you want to set your agent's language to Luxembourgish, you will need to manually install the spellux library.

Example agents

• greetings_agent: Very simple agent for the first contact with the framework
• weather_agent: Introducing entities
• llm_agent: Introducing Large Language Models (LLMs)
• rag_agent: Introducing Retrieval Augmented Generation (RAG)
• telegram_agent: Introducing the TelegramPlatform

For more example agents, check out the BAF-agent-examples repository!