pdf-reading

name	pdf-reading
description	PDF parsing, page chunking, and evidence note preparation for Idea2Repo. Use when extracting page text, stable chunk ids, paper notes, or PDF reader prompt inputs.

PDF Reading

Parse downloaded PDFs into stable chunks before any analysis.

Workflow

1. Read docs/reference/pdf_manifest.json and process only downloaded records.
2. Validate and parse each PDF with the repository parser.
3. Chunk by page or section with stable chunk_id values.
4. Write docs/reference/pdf_chunks.json.
5. Prepare paper notes in docs/reference/paper_notes/<paper_id>.md with page, quote, and chunk id for every important claim.

Guardrails

• Mark corrupted, missing, or low-text PDFs as incomplete instead of inferring content.
• Extract limitations and negative results, not only strengths.
• Do not let downstream novelty or score use claims without page, quote, and chunk id.