ArchitectureA supervised JVM-class runtime — OLTP on seven engines, OLAP on three. AI-native, MCP-native, observable as plain SQL.Read the architecture
Está viendo la edición Perú. Está viendo la edición Colombia. You're viewing the Pakistan edition. Cambiar a la edición global →Cambiar a la edición global →Switch to the global edition →

One Javadoc scan writes the scripting API catalogue every document is built from

A single scan of the Java sources produces the catalogue of the scripting library, one document per type with samples and deprecations. TypeScript declarations, the Docs app and the AI reference derive from it.

The scripting library was documented four times: a doclet for TypeScript declarations, a folder for the Docs app, an extractor for the AI reference, and the Javadoc itself. They disagreed. This release makes one scan the source and derives the rest.

The catalogue

  • One document per type, with an index, written as a classpath resource at build time. Samples are kept verbatim, an undocumented override inherits its description, deprecated members are kept and flagged, and abstract members are marked.
  • Complete 4,096 documents, every core, AI, scripting and document-engine type that exports members, with the JDK and third-party classes the library exposes. Of 565 linked names, one has no document, on purpose.

Derived from it

  • TypeScript declarations generated at build time and packed in the jar, byte-identical to the old doclet's output, no longer committed. The two doclets were removed.
  • The Docs app and MCP resources read the catalogue from the jar, so user documentation is as current as the code in the same build, with a deprecated chip where it applies.
  • The AI reference keeps code samples and never cuts prose: 1,118 classes and 6,798 documented members, up from 760 and 4,970, and 1,593 samples where before none reached the model and 103 members ended on a dangling "example usage".

Four documentations that could disagree became one that cannot.