Mochocyang/QMAI

What it solves

QMAI is a desktop writing system designed for long-form novels (2-3 million words). It addresses the common failures of standard AI writing tools, such as the AI forgetting previous plot points, inconsistent character personalities, chaotic timelines, and lost foreshadowing as the story progresses.

How it works

The system uses a structured memory engine that automatically processes each saved chapter to extract key entities, events, and relationship changes. This data is stored locally as Markdown and JSON files.

When generating new text, a Context Engine assembles a prioritized "context package" (including outlines, recent chapters, and canon rules) and uses a hybrid retrieval strategy—combining keyword, vector, and graph-based searches—to ensure the AI has the necessary background information within its token limits.

Who it’s for

• Web novel authors who need to maintain quality and consistency across long serializations.
• Story planners managing complex world-building, factions, and multi-threaded plots.
• AI-assisted writers seeking a tool that remains usable over millions of words.
• Students of writing who want to analyze and replicate the styles of successful works via the "Book Analysis" library.

Highlights

• Memory System: Automatically extracts chapter summaries, character states, and foreshadowing to prevent plot holes.

• Character Soul System: Defines narrative tone and specific character "DNA" (speech habits, decision patterns) to ensure consistent voice.

• Book Analysis Library: Deconstructs existing novels to extract 9-dimensional style metrics and character skills for reuse in new projects.

• Multi-dimensional Review: An AI-driven audit system that checks for "satisfaction density," setting consistency, and character alignment.

• Knowledge Graph: Visualizes characters, locations, and plot threads using Sigma.js to track complex relationships.

• Causality Tracking: A "Foreshadowing Debt" system that tracks unrecovered plot points and alerts the author.