This project provides an engineering framework to mitigate Large Language Model (LLM) hallucinations when processing traditional Chinese divination arts like Qimen Dunjia and Ziwei Doushu. It targets developers and researchers working with LLMs in specialized, knowledge-intensive domains, offering a method to ensure accuracy and control over complex, rule-based calculations.

How It Works

The framework employs a hybrid approach, combining strict prompt engineering with external, structured knowledge bases and executable logic scripts. This architecture aims to replace LLMs' inherent "metaphysical hallucinations" with transparent, controllable, and verifiable calculation processes. By decoupling divination logic into distinct modules and references, it ensures precise understanding and application of traditional rules.

Quick Start & Requirements

No installation or specific requirements are detailed in the provided README snippet.

Highlighted Details

• Includes dedicated skills for qimen-dunjia (奇门遁甲), featuring charting logic, constraint analysis, and relationship judgments.
• Offers ziwei-doushu (紫微斗数) skills covering star-charting logic, core calculation principles, the Four Transformations system, and natal chart pattern analysis.
• The core design emphasizes transparency and control over LLM outputs in complex, rule-based domains.

Maintenance & Community

No information on maintenance, community channels, or contributors is available in the provided text.

Licensing & Compatibility

No licensing information is provided in the README snippet.

Limitations & Caveats

The project's primary focus is on addressing LLM hallucinations within specific traditional divination domains; its applicability to other complex, rule-based systems would require further investigation. The current scope is limited to the two mentioned divination arts.