Extract PDF metadata (author, title, dates, etc.)

[monneyboi]

Summary

Extract all available metadata from PDF files during ingestion, including author, title, creation/modification dates, and other standard PDF Info dictionary fields.

Current State

• DocumentMetadata only stores: id, name, pdf_hash, text_hash, page_count, tags, created_at
• ExtractedDocument only returns: pdf_bytes, text, page_count

Available PDF Metadata

PDFs have an Info dictionary (accessed via doc.trailer.get(b"Info") in lopdf) containing:

Field	Type	Description
Title	String	Document title
Author	String	Person who created the content
Subject	String	Subject/topic
Keywords	String	Keywords (comma-separated)
Creator	String	Application that created the original (e.g., "Microsoft Word")
Producer	String	Application that converted to PDF
CreationDate	Date	When the document was created
ModDate	Date	When the document was last modified

Implementation

1. src/core/pdf/extractor.rs - Extract metadata from the Info dictionary using:

   if let Ok(info_ref) = doc.trailer.get(b"Info") {
       if let Ok((_, info)) = doc.dereference(info_ref) {
           // Extract Title, Author, Subject, Keywords, Creator, Producer
           // Use Object::as_datetime() for CreationDate/ModDate
       }
   }

2. ExtractedDocument - Add optional metadata fields

3. DocumentMetadata - Add new optional fields:

   • title: Option<String>
   • author: Option<String>
   • subject: Option<String>
   • keywords: Option<String>
   • creator: Option<String>
   • producer: Option<String>
   • pdf_created_at: Option<String> (from PDF CreationDate)
   • pdf_modified_at: Option<String> (from PDF ModDate)

4. Search index - Consider indexing title/author/keywords as searchable fields

Notes

• All new fields should be Option<T> since not all PDFs contain metadata
• lopdf has Object::as_datetime() for parsing PDF date format (D:YYYYMMDDHHmmSS+HH'mm')
• Metadata syncs automatically since it's part of the metadata blob in iroh-docs