What comes out
The output has three parts. nav is the corpus arranged as parts and the chapters beneath them, which is what a site's navigation is built from. pages is a page per chapter, each a list of blocks: a heading carries its level, its title and a slug to link to, and a prose block carries Markdown. loose holds chapters that belong to no part.
Because prose comes out as Markdown rather than as HTML, the site decides how to render it. Nothing about weaving assumes a browser.
Point it at a file, not a directory
Weave works out which corpus you mean from the path you hand it, and it takes the directory that path sits in. Give it a chapter, or a spine — a file whose only job is to hold the table of contents — and it reads that file's directory and everything beneath it:
Give it a directory instead and, unless that directory is a workspace in its own right, it reads the directory above the one you named. In a repository that means it can quietly weave everything you have, test fixtures included. The rule is short: pass a file.
More than one corpus in a project
A project can hold several corpora — a book of documentation beside the source it documents, for instance. They need no configuration of their own. Each is a directory of chapters, and each is woven by pointing at a file inside it, so the same workspace settings govern them all.