LLM Agents Graph

This project implements a small agent system using LangGraph and HuggingFace transformers.
Two agents answer questions, and a critic evaluates which answer is better.

Features

• Runs two instruction-tuned models as agents
• Uses a third model as a critic to compare answers
• Simple graph-based pipeline with LangGraph

Getting Started

1. Clone the repository

git clone https://github.com/christalphilip/llm-agents.git
cd llm-agents

2. Install dependencies

pip install -r requirements.txt

Usage

python llm_agents.py

You will be prompted to enter a question. The agents and critic will generate responses and print the output. Outputs are also saved in agent_outputs.txt

Example Output

An example run is included in the example_outputs/ folder. Actual outputs from each run are saved to agent_outputs.txt (not tracked in version control).

Project Structure

llm_agents/
├── llm_agents.py
├── requirements.txt
├── README.md
├── LICENSE
├── .gitignore
└── example_outputs/ └── exercise.txt

License

This project is licensed under the MIT License.