ai4finance-foundation/finrl

FinRL is an open-source Python framework for financial reinforcement learning. Reinforcement learning is a machine learning approach where an agent learns by trial and error: it takes actions in some environment, sees the results, and adjusts its strategy over time to earn more reward. FinRL applies this idea to financial markets, where the agent learns trading decisions by interacting with simulated market data. The project is positioned as the original, classic version of the framework. The maintainers now point new users toward a separate project called FinRL-X, also known as FinRL-Trading, described as the next-generation production-oriented stack. The repository at hand keeps the original train-test-trade pipeline alive for education, benchmarking, and research prototyping. A comparison table in the README lists differences between the two: the classic FinRL uses a three-layer architecture with five named DRL algorithms (A2C, DDPG, PPO, SAC, TD3), manually wired data processors, and basic broker support, while FinRL-X adds modular layers, machine-learning components, and richer risk controls. The codebase is organized around three core layers: market environments, DRL agents, and financial applications. The README walks users through cloning the repo, creating a Python virtual environment, installing with pip, then running example scripts that download DOW 30 stock data from Yahoo Finance and run a stock trading workflow. Example scripts are referenced for data download, training, and evaluation. FinRL is part of a broader ecosystem from the AI4Finance Foundation, alongside FinRL-Meta (gym-style market environments), ElegantRL (DRL algorithm library), and FinGPT (financial large language models). The README points to a Discord community, ReadTheDocs documentation, tutorials on YouTube, and academic papers on arXiv.