retriva-pdf-injector-patterns

name	retriva-pdf-injector-patterns
description	Guidance for building a local PDF injector for Retriva.

Retriva PDF Injector Patterns

Core input shape

The input is a recursive directory containing one or more .pdf files.

Preferred processing flow

1. recursively scan the root directory
2. detect candidate PDF files
3. extract embedded text page by page
4. preserve page numbers and document-level metadata
5. emit document/chunk objects compatible with the existing indexing pipeline
6. keep enough metadata for citations and debugging

Guardrails

• Do not require OCR for PDFs with extractable text
• Skip encrypted or unreadable PDFs gracefully with diagnostics
• Keep chunk metadata page-aware for citations