RD Agent
Basic Information
R&D-Agent (RD-Agent) is an open-source, multi-agent framework designed to automate data-driven research and development workflows using large language models and programmatic loops. The repository focuses on two complementary agent roles: 'R' for proposing new ideas and hypotheses and 'D' for implementing those ideas as runnable code, enabling iterative evolution of models and data. It targets a range of industrial scenarios including quantitative finance, medical prediction, Kaggle competitions and general model research by extracting formulas and features from papers and reports, implementing models, and running experiments. The project provides command-line scenarios, a web UI for monitoring, reproducible demo traces and documentation. It supports local deployment via Docker and Conda, installation from PyPI, and developer setup from source. Configuration supports multiple LLM backends and embedding providers for chat and retrieval workflows.
Links
Categorization
App Details
-
$$$$ -
$$$$