HKUDS/Vibe-Trading

"Vibe-Trading: Your Personal Trading Agent"

What it is

Vibe-Trading is an AI-powered personal trading agent. You chat with it in plain language (via Slack, Discord, Telegram, a web UI, email, or WeChat), and it can do real work: pull live market data, compute technical indicators, run backtests, and even place orders through broker connections. It's a full application — Python/FastAPI backend, React frontend, installable via pip install vibe-trading-ai — not a demo.

How it works

The core is an LLM given a toolbox: market-data loaders (Stooq, BaoStock, Eastmoney, etc.), a backtest engine, indicator calculators, and connectors to 18 brokers (Robinhood, Upbit, KIS, Scalable Capital, MT5, and more). You describe what you want ("backtest a momentum strategy on AAPL"), and the agent fetches data, runs the analysis, and reports back.

The standout engineering idea is a "grounding gate" that fights AI hallucination. When the model reports a number, it must declare the figure's role — observed (from data), derived (computed), cited, etc. — and the system checks that declaration against the actual tool outputs. A number that can't be verified gets cut from the answer. Backtest run cards even show hashed execution records and verified metric-to-CSV references, so you can audit where every reported metric came from.

Key features

  • Real trading + paper trading: "Shadow accounts" let the agent practice without risking money; live orders go through broker APIs.
  • Backtesting: Supports strategies with risk metrics (Sharpe, Sortino, Monte Carlo drawdown), walk-forward folds, and reproducible runs.
  • Multi-channel chat: Same agent accessible from Slack, Discord, Telegram, Feishu, WeChat, email, WebSocket, and a web UI — in 9 languages.
  • Scheduled reports: Automated PDF/HTML reports delivered by email, optionally AES-256 encrypted.
  • Long-running research: The agent keeps memory across conversation compactions, and context handling adapts to each model's actual window size.
  • Model-agnostic: Works with multiple LLM backends (OpenRouter is the shipped default; Codex and others are supported).

Who it's for

Individual traders and investors who want an AI assistant that does things rather than just talks — researching markets, validating strategies, and executing trades — with guardrails against the AI making up numbers. The project is actively maintained by HKUDS (Hong Kong University of Data Science) with a large contributor base.

Notable

The README carries a prominent security warning: scammers have impersonated the project with fake tokens and memecoins. The maintainers explicitly state they never launched any token — a sign the project is popular enough to attract fraud.

Related

  • Project
    EthanAlgoX/AIStock

    AI Stock is an agentic research workspace that helps investors analyze stocks, compare investment perspectives, and simulate trading strategies using LLM-powered tools.

  • Project
    simonlin1212/Vibe-Research

    Vibe Research is a local-first investment research workstation that integrates global stock market data with AI-assisted analysis to help users perform deep-dive company research, backtesting, and portfolio management.

  • Project
    x-pai/AlphaBot

    AlphaBot is an open‑source, Docker‑deployable web app that combines a LLM‑driven chat assistant with real‑time market data (global via Alpha Vantage, Chinese A‑shares via Tushare/AKShare). It offers AI‑generated technical and fundamental analysis, chart visualisation, risk assessment, and simple quant‑strategy suggestions for retail investors.

  • Project
    yenche123/liubai

    Liubai is an open‑source, AI‑enhanced personal note‑taking, calendar, and task‑management system. It stores information as modular “cards” that can include text, images, files, tags, and status. The platform works as a responsive web PWA, a WeChat mini‑program, and a VS Code extension, all syncing through a Laf‑based backend. Integrated LLMs (DeepSeek, Kimi, MiniMax, etc.) let you query or augment your notes directly from WeChat, while privacy controls let you keep cards local or block AI access. Offline‑first design, dark/light themes, and Chinese/English UI are built‑in. AGPL‑v3 license.

  • Project
    lc2panda/alphastream

    Alphastream is an AI‑driven financial analysis web app. It uses DeepSeek/any OpenAI‑compatible LLM with LangGraph to run 14 specialized agents (technical, fundamental, sentiment, risk, etc.) plus four investor‑persona agents. Real‑time market data comes from dual adapters (AKShare & BaoStock) covering Chinese, US, Hong‑Kong, crypto and macro sources. The Next.js + React front‑end shows a dark glass‑morphism UI with chat, dashboards and chart cards, while the Flask backend streams agent progress via SSE. All services can be launched with a single Docker‑Compose command.