Extract Page Structure

• PageStructure
• article metadata
• ordered block inventory
• embed and asset references
• blockCompleteness
• headerFidelityRequirements
• noteTextRequirements

Procedure

1. Parse the document from static source HTML only.
2. Extract article metadata such as headline, deck, byline, date, breadcrumbs, and story links.
3. Walk the article body in source order and classify content-bearing blocks.
4. Record embeds and special containers such as Birdkit, Datawrapper, ai2html, and quote sections.
5. Produce a normalized PageStructure payload with stable ordering and block indexes.
6. Emit header fidelity evidence such as deck presence versus visible rendering, byline/date arrangement, and explanatory note ordering.
7. Compute blockCompleteness from matched versus expected content-bearing blocks.

Validation

1. Preserve source order for every classified block.
2. Return empty arrays rather than fabricated blocks when a component type is absent.
3. Flag likely extraction gaps when blockCompleteness indicates missing content.

Stop And Escalate If

1. Source HTML is incomplete enough that ordering cannot be trusted.
2. More than a small minority of content-bearing blocks remain unclassified without explanation.

Completion Contract

Return:

1. page_structure
2. metadata_summary
3. embed_summary
4. block_completeness
5. warnings