unitagain/WenShape

What it solves

WenShape is designed for long-form novel writing, addressing the common issues of consistency, traceability, and maintainability in AI-assisted storytelling. Instead of treating the LLM as a black box that generates a whole book at once, it breaks the writing process into manageable, structured modules to prevent hallucinations and maintain a coherent narrative across many chapters.

How it works

The system uses an orchestrator that coordinates specialized agents (Writer, Editor, Archivist) and a dedicated context engine. It organizes data into a hierarchy of volumes, chapters, and drafts, storing them in human-readable formats like YAML, Markdown, and JSONL.

To maintain consistency, it employs a "fact summary system" that uses distance-based indexing. This ensures that facts closest to the current chapter have higher weight, while still retaining important world-building details. For fan-fiction, it includes a crawler service that can extract data from wikis (like Wikipedia or Fandom) to automatically propose character and world-building cards.

Who it’s for

Authors and writers who want to use AI to write long-form fiction while maintaining strict control over plot, character arcs, and world-building settings.

Highlights

• Structured Writing Workflow: Explicitly manages the hierarchy of volumes, chapters, and summaries to keep the narrative ordered.
• Card-Based Setting Management: Uses dedicated cards for characters, world-building, and style to keep assets reusable and searchable.
• Distance-Aware Context Injection: Reduces hallucinations by using a logarithmic chapter distance decay to prioritize relevant facts.
• Fan-Fiction Integration: Built-in tools to search and crawl web pages to quickly build a source material's world-view.
• Multi-Provider LLM Support: A unified gateway supporting OpenAI, Anthropic, DeepSeek, Gemini, Qwen, and other OpenAI-compatible APIs.