Annotation and Metadata

This is the data path behind annotated links. It connects the Build Pipeline, where metadata is read and generated, to the Popup System, where annotation fragments are shown to readers.

Reading Path

Step	Page	Why it matters
1	LinkMetadata.hs	Central metadata lookup, annotation creation, and fragment generation
2	Annotation.hs	Dispatcher that routes URLs to source-specific metadata extractors
3	GTX.hs	Parser and writer for the line-based metadata database format
4	LinkMetadataTypes.hs	Shared metadata types used across the backend
5	annotations.js	Browser-side annotation data loading and caching

Scrapers and Cleanup

Area	Start here
arXiv	annotation-arxiv.hs
BioRxiv and MedRxiv	annotation-biorxiv.hs
Local gwern.net pages	annotation-gwernnet.hs
OpenReview	annotation-openreview.hs
PDFs	annotation-pdf.hs
Author cleanup	Metadata.Author.hs
Title cleanup	Metadata.Title.hs
Abstract formatting	Metadata.Format.hs

Where It Surfaces

• Popup System shows how metadata appears in popups and popovers.
• annotation-blockquote templates define rendered annotation HTML.
• generateSimilar.hs reuses metadata for embedding-based recommendations.
• Functional Taxonomy lists the broader set of annotation and metadata files.