I'm building an agent from scratch as a learning exercise.

Project structure

• harness.py: agent control loop that orchestrates model calls, tool calls, context, and more
• llm.py: handle model API calls
• tools.py: actions that the agent can perform

Getting started

python main.py