deepagents>=0.6
langchain-openai>=1.0.0
python-dotenv
tavily-python  # baseline agent in the benchmark (benchmark.py)
