How it works

For contributors and readers evaluating the implementation. To use the extension in your library, start with Publish your first API pages.

Sphinx runs a small JDK doclet over your sources before reading documents. The doclet uses the compiler’s view of declarations, overloads and Javadoc links and writes a versioned XML model. The extension turns that model into native Sphinx objects, cross-references and inventory entries. Source comments are never interpreted as reStructuredText, and no raw HTML is emitted.

The extractor never loads application classes, so static initializers and annotation processors do not run. Results are cached by the content of sources, dependencies, options and the JDK, so any change re-extracts the whole project.

Design records

These live in the repository:

• Architecture decision and alternatives considered

• XML model schema

• Roadmap

Tested environments and known limits are on Compatibility and limits.

Contribute

The contributor guide covers local setup, checks, browser validation and releases.