About this project

Feynman is an AI-powered research agent that streamlines scientific inquiry through a combination of CLI tools and a local workbench. It is built on the Pi agent runtime and integrates with AlphaXiv for paper search and analysis. The tool provides a suite of specialized agents—Researcher, Reviewer, Writer, and Verifier—to handle tasks ranging from literature reviews and citation graph analysis to auditing claims against codebases and planning replication experiments. Key capabilities include: - Research Workflows: Supports deep research, literature reviews, paper ranking based on citation prestige and reproducibility, and automated ML training recipe discovery. - Scientific Tooling: Includes extensive connectors for biological and chemical data, such as OpenAlex, PubMed, ChEMBL, PubChem, and various genomic/proteomic databases. - Local Workbench: A standalone interface for managing research projects, sessions, and artifacts, featuring previews for PDFs, Jupyter notebooks, LaTeX, and chemical structures (KET/RXN/CDXML). - Extensibility: Users can install specific research skills into existing repositories or Codex environments. It supports local model providers (Ollama, vLLM, LM Studio) and hosted APIs. - Observability: Provides logs, distributed traces, and artifact lineage to ensure research transparency and provenance. Feynman is designed for researchers who require source-grounded evidence, allowing users to verify claims via direct links to papers, documentation, and repositories.