PDF Research

PDF research

Find and mine PDF documents: parse text, chunk it, and retrieve the passages that answer a question.

When to use

• The evidence is in a PDF (paper, spec, manual, whitepaper).
• Extracting quotable passages with page/section location.
• Building a PDF-backed corpus for evidence extraction or a literature review.

Tools used

• pdf_research — find relevant PDFs (query, limit).
• search_documents — content search with filetype: "pdf" (default).

Workflow

1. pdf_research (or search_documents) to locate PDFs.
2. Parse: extract text page-by-page; chunk into section-labeled spans.
3. Retrieve the relevant chunks for the question; keep page + chunk ids.
4. Never execute anything embedded in the PDF (scripts, macros, actions).

Input schema

{ "query": "str", "limit": "int" }

Output schema

{ "results": "list[{url,title,chunks:[{page,text}]}]" }

Security

PDFs can contain scripts/actions/embedded objects. Treat the file as untrusted data; extraction only, no execution, and keep files task-scoped.

Related skills

document-research, academic-research, literature-review, evidence-extraction