metadata-extractor

/metadata-extractor — Metadata Extractor Generator

In-Process SDK only — metadata extractors deploy inside the ACS JVM as part of the Platform JAR. They map values found in a file's content into node properties at upload time — distinct from renditions/transforms (/transforms), which convert content into other content.

Generate a custom metadata extractor (and optional embedder) from requirements.

Input

Read REQUIREMENTS.md to identify metadata extraction requirements:

1. Resolve the Platform JAR project's Root path from Section 2 (Project Architecture).

   • If Section 2 contains no Platform JAR project, stop and explain that /metadata-extractor only applies to the in-process Platform JAR project.

2. Read the "Metadata extraction requirements" sub-section (Section 7 Behaviour Requirements or a dedicated section).

   • If none are present, stop and ask the user to run /requirements first (or provide a description as $ARGUMENTS).
   • Check coverage first: ACS routes extraction for common formats (Office, PDF, images, audio/video) through the Transform Service's Tika engine. Only scaffold a custom in-process extractor for a bespoke format or a custom source→property mapping that the standard extractors do not provide.

3. From Section 2, derive:

   • {platform-project-root} — . for Platform JAR only mode; {name}-platform/ for Mixed mode
   • {module-id} — the Platform JAR artifactId (bare artifact ID, e.g. my-extension). Read from <artifactId> in the platform pom.xml or derive as {platform-artifactId} from Section 2. Never use the full module.id property value as the directory name.
   • {java-package} — the Java package declared in Section 2
   • {package-path} — {java-package} with dots replaced by slashes (resource path)
   • {prefix} — the namespace prefix declared in Section 5 (Content Model Requirements)

4. Derive from extraction requirements:

   • {Name} — PascalCase extractor name (e.g. Contract, Invoice)
   • {sourceMimetype} — the MIME type(s) the extractor supports
   • The source→target mapping: each raw key in the file → a content-model property QName (e.g. vendor → {prefix}:vendor)