Full-Text Ingestion

Input Priority

ingest_paper_fulltext accepts any of:

• doi
• paper_url
• pdf_url
• local_pdf_path

You must pass at least one.

Parse Modes

• auto (default): tries parser fallback chain
• grobid: force GROBID path
• simple: force basic parsing fallback

In auto mode, ScholarMCP now attempts grobid -> simple.

Common Failure Cases

• DOI page has no downloadable PDF URL:
  • retry with pdf_url or local_pdf_path
• Remote downloads disabled:
  • set RESEARCH_ALLOW_REMOTE_PDFS=true
• Local ingestion disabled:
  • set RESEARCH_ALLOW_LOCAL_PDFS=true
• Throttling or timeout pressure:
  • increase SCHOLAR_REQUEST_DELAY_MS and/or RESEARCH_TIMEOUT_MS