Awesome AI Agents › Financial & Trading Systems

Lumiwealth/lumibot

⭐ 2060 Python repository created 2020-09-10

Lumibot is a Python library for building, backtesting and running algorithmic trading strategies, including strategies driven by LLM agents. A strategy is written as a Python class that subclasses Strategy and implements callbacks such as on_trading_iteration, where it requests market data, creates orders and submits them. The same strategy class runs in three modes without rewriting logic: backtesting against historical data with simulated fills, paper trading against a broker sandbox, and live trading against a real broker account, so switching mode means swapping the runner. Backtesting data sources such as YahooDataBacktesting are passed as classes to the backtest entry point, and runs produce artifacts that can be inspected afterwards. For paper and live execution the library ships broker adapters, with Alpaca used in the README quick start, wired into a Trader object that can run several strategies together. Beyond deterministic rule-based strategies, Lumibot includes an AI agent runtime in which one or more agents reason over evidence, call tools, write memory and optionally submit orders inside the same strategy loop used by backtests. Built-in agent tools cover market and account state, order inspection, DuckDB queries, documentation search, Alpaca news when credentials exist, technical indicators, SEC fundamentals and filings, FRED macroeconomic data, local memory and Telegram notifications. The documentation describes composing agents into teams, for example a research agent that assembles an evidence pack, bull and bear agents that argue opposite sides of a trade, and a trader or portfolio manager agent that checks cash, positions, open orders and risk limits before deciding. Hybrid designs are supported, where deterministic Python code enforces hard gates and agents handle open-ended reasoning. It targets quantitative developers and traders who want reproducible backtests and agent experiments in one codebase.

https://github.com/Lumiwealth/lumibot

tradingbacktestingfinancepythonmulti-agent