HKUDS/Vibe-Trading

What it solves

Vibe-Trading provides a comprehensive framework for building and deploying personal trading agents. It bridges the gap between high-level AI agent reasoning and the technical execution of financial trades, offering a unified system for market data ingestion, strategy backtesting, and live broker integration.

How it works

The project operates as an agentic system that empowers LLMs with a suite of trading capabilities. It includes a backend powered by FastAPI and a frontend built with React. The system integrates with multiple brokers and market data sources, allowing an agent to perform tasks such as fetching fundamentals, analyzing fund flows, and executing trades. It also features a "Shadow Account" for safe testing and a "Quant Library" for advanced financial calculations (e.g., Heston model, Risk Parity, and microstructure analysis).

Who it’s for

It is designed for traders and developers who want to automate their trading strategies using AI agents, as well as quantitative researchers who need a robust environment for backtesting and managing multi-broker portfolios.

Highlights

• Multi-Broker Support: Integration with a wide array of brokers, including Alpaca, Zerodha Kite Connect, and others.
• Extensive Quant Library: Built-in support for complex financial models like copulas, barrier options, and event-study tests.
• Agentic Tooling: Provides a set of tools that allow AI agents to perform market research, backtest strategies, and manage positions.
• Comprehensive Data Handling: Supports diverse asset classes including equities (US, HK, China, UK), futures, and crypto, with rigorous handling of data gaps and adjustments.