If Document AI returns an empty schema for your processor version, check which version you queried. Pretrained versions (e.g. Google Stable) expose only a stub schema with no entity types - the populated schema with all your custom properties exists only on processor versions you trained yourself. Point your schema read at your own trained version and the entity types show up as documented.

Context: Stack Overflow 79196878 (accepted answer, 2 votes): with a Custom Extractor, the Python API's view of the processor schema showed no entity types, contradicting the docs. The accepted answer tested it: the code is fine, but pretrained processor versions like Google Stable ship only a stub schema.