Projects, books, chapters and scenes

Your manuscript has four levels:

Project (one work)
└── Book (a volume)
    └── Chapter
        └── Scene  ← prose lives here

While there is only one book, that level is hidden in the UI — the top bar shows chapter and word counts instead, and the plan view doesn't draw a book row. Add a second book and the level reappears. So "new project = new book" holds in practice, without giving up multi-volume support.

Why prose is split into scenes

A scene is the smallest unit for prose, search, export and AI context:

• autosave, word count and mention scanning all operate per scene;
• full-text search returns scenes, and clicking one jumps you back to it;
• when generating, the summaries of the previous and next scenes are fed to the model.

The writing view is a continuous manuscript by default: every scene in a chapter is shown as one flowing document, so it reads and edits like a whole chapter while staying separate underneath.

Scene summaries earn their keep

Each scene takes a one-to-three sentence summary. It isn't just a table of contents for you — earlier scenes' summaries become the "story so far" in the AI's context. Accurate summaries are the cheapest way to stop the AI contradicting itself.

The plan view has "生成摘要" next to each scene if you'd rather have the AI summarise existing prose.