QuantVibe

•
3
•

Integrates Qlib's quantitative machine learning for stock scoring with Vibe-Trading's LLM agent for executing trading plans via a secure, read-only MCP server.

QuantVibe acts as a bridge between Qlib, Microsoft's comprehensive quantitative finance platform, and Vibe-Trading, an LLM-powered trading agent. It enables Qlib to train machine learning models that generate stock scores, which are then securely delivered to Vibe-Trading via signed signal files and a read-only MCP server. This innovative integration allows for ML-driven trading signals to inform an AI agent's decision-making, offering robust features like signed signal contracts, evaluation gates for model performance, and a track record for monitoring actual signal outcomes, all while ensuring isolated environments for each component.

주요 기능

010 GitHub stars
02Comprehensive track record in SQLite for historical signal logging and settlement
03Read-only FastMCP server for real-time signal access by LLM agents
04Evaluation gate for model performance (IC/ICIR/hit-rate) before signal publication
05Docker support for easy deployment and dependency management across Python versions
06Secure signal contract with schema validation and SHA-256 checksums

사용 사례

01Evaluating the real-world performance of stock scoring models through a persistent track record
02Connecting quantitative models to LLM-based trading agents for automated paper trading
03Developing and backtesting ML-driven trading strategies with Qlib