finrl

This listing points to the FinRL Trading repository maintained by the AI4Finance Foundation, which now hosts the successor to the original FinRL deep reinforcement learning library. The current framework is an open source, modular infrastructure for quantitative trading that unifies data processing, strategy composition, backtesting and broker execution behind a single interface: the target portfolio weight vector. Because every strategy outputs weights, the same code runs unchanged in backtests, paper trading and live execution. It is built for quant researchers, students and practitioners comfortable in Python 3.11 or later. The data layer pulls from Yahoo Finance, Financial Modeling Prep and WRDS with SQLite caching. The strategy layer composes stock selection, portfolio allocation, timing adjustment and risk overlays, with reference strategies covering allocation paradigms, rolling stock selection with deep reinforcement learning and adaptive multi asset rotation. Backtesting runs on the bt engine with multi benchmark comparison, and execution goes through Alpaca with multi account support and pre trade risk checks. Configuration uses Pydantic settings, and the design is described in an accompanying arXiv paper. The software is free under the Apache 2.0 license and installs from PyPI with pip. There is no commercial tier; users bring their own data subscriptions and brokerage accounts. The repository is active with recent commits, open issues and pull requests, and a Discord community for support.