Annotate Stage

Purpose

The annotate stage classifies extracted sections and optionally produces more complete annotation outputs.

Main responsibilities

• load section parquet outputs
• apply the packaged section-classifier model
• optionally perform fuller annotation behavior
• preserve document-type context when available

Main inputs

• section parquet outputs
• classifier model
• optional document-type metadata

Main outputs

• classified section parquet
• fully annotated section parquet when enabled

Dependency note

This stage depends on model availability. If the model is missing, the stage should fail clearly or skip clearly rather than silently pretending annotation succeeded.

Metadata role

This stage can enrich downstream outputs with document-type and processing-stage information.

Contributor note

Changes here should preserve compatibility with the section parquet contract and should document any new output schema or annotation semantics.