Readers

The catalog indexes recognized formats; opening depends on capability in src/readers/core.js in araru-web. The Server resolves ID to source and never exposes physical path.

Common flow:

1. Reading gets the book and format;
2. loads engine/parser on demand;
3. requests content, index, or page;
4. renders within the common shell;
5. limited prefetch prepares neighbors;
6. progress is persisted;
7. cleanup frees resources when switching/closing.

Endpoints are in API/content. Details by format in formats.