This repo contains 3 basic demos that were discussed during the macroexpand-2 meetup on 08/09/2025.
It shows how to use ObneyAI's Grain library to build agents of varying complexity.
To actually run the examples, you'll need a python virtual environment in the root of the project directory running at least Python 3.12. You will need to install the requirements in the requirements.txt file as well. You'll also need an open router API key or familiarity with DSPY to adjust the config in extras/dspy.clj.
This shows a naive chatbot with no conversational memory.
This shows a chatbot with conversational memory.
This shows a naive researcher coding agent that can accept simple requests and use a read only set of tools to investigate the current project. It's primary purpose is to demonstrate an agent of reasonable complexity and how Behavior Trees keep that complexity under control. It also demonstrates a re-entrant general purpose Tool Use subtree in order to display the optimally modular property of behavior trees.