try-monty

A small example project demonstrating a self-improving agent that uses a local Ollama LLM to generate and refine Monty-compatible Python snippets and execute them via pydantic-monty.

Quick start

• Prerequisites: Python 3.12+, Ollama running locally and accessible to the ollama Python client.
• Install dependencies (editable install recommended for development):

python -m pip install --upgrade pip
python -m pip install -e .[dev]

• Create a sample config (creates config/agent_config.toml):

python examples/create_sample_config.py

Edit config/agent_config.toml to set model, max_attempts, and logging settings.

Usage

Run the example agent from the project root (after creating a config):

python -m src.monty_ollama_agent

Or import and instantiate the agent in your own script:

from src.monty_ollama_agent import SelfImprovingAgent

agent = SelfImprovingAgent(config_path="config/agent_config.toml")
result = agent.run("your task description", {"a": 1, "b": 2})      # This is where the problem to be solve and inputs defined 

Tests

Run the test suite with pytest. If you installed the dev extras above you can run directly; otherwise set PYTHONPATH so tests can import src:

# If pytest is installed in the venv
pytest -q

# Or, with PYTHONPATH set (works if pytest is in the venv but package not installed)
PYTHONPATH=$(pwd) .venv/bin/pytest -q

Files of interest

• src/monty_ollama_agent.py — main agent implementation
• examples/create_sample_config.py — utility to create a starter config/agent_config.toml
• config/agent_config.toml — configuration for model, prompts, and logging
• tests/ — pytest unit tests (run with pytest)

Notes & recommendations

• The project requires Python >=3.12 (see pyproject.toml).
• The module no longer auto-creates a config on import — use examples/create_sample_config.py to create a starter file.
• The agent executes generated code via pydantic-monty; treat generated code as untrusted and run it in a controlled environment.

License

See the repository LICENSE file.