lyra81604/zhengxi-views

What it solves

This project addresses the issue of AI models hallucinating or fabricating the views of fund managers. It provides a verifiable, evidence-based knowledge base for the investment views and methodology of Zheng Xi (Deputy General Manager of E Fund Equity Investment Management and Fund Manager), ensuring that every conclusion can be traced back to specific public disclosures, reports, or interviews.

How it works

The system functions as an "Agent Skill" that combines a curated knowledge base with executable scripts:

• Verifiable Corpus: A collection of original texts from 2012–2026, including quarterly/annual reports, manager notes, and media interviews.
• Distilled Methodology: A framework derived from the corpus where every rule is backed by original quotes, allowing the AI to deduce views on new topics using Zheng Xi's logic rather than guessing.
• Real-time Data Integration: Python scripts that can fetch data for any of the 27,000+ funds in the market to compare them against Zheng Xi's managed funds or score them using his specific investment criteria.
• Cross-Platform Compatibility: Designed to be integrated into AI agents like Claude Code, Tencent WorkBuddy, Cursor, and ChatGPT via custom instructions and knowledge files.

Who it’s for

• Investors and researchers who want to study Zheng Xi's investment philosophy.
• Analysts looking to verify if a fund manager's actual holdings align with their stated views (say-do consistency).
• Users of AI agents who want a structured, evidence-based way to analyze funds and market trends using a specific professional framework.

Highlights

• Strict Traceability: Every answer is required to provide original citations; the AI is explicitly forbidden from fabricating quotes.
• Methodological Deduction: When a topic isn't in the corpus, the AI uses a distilled framework to simulate how the manager would think, clearly labeling these as deductions.
• ** uma-market Comparison**: Ability to fetch real-time data for any fund in the market to perform side-by-side comparisons.
• Automated Research Pipeline: Can be connected to MCP (Model Context Protocol) and automation tools to generate daily structured analysis reports based on market data and the manager's framework.