feat: PDF text extraction, content size guardrails, and dynamic file offloading

[coctostan]

Summary

• PDF text extraction: fetch_content now detects application/pdf, extracts text via pdf-parse, with graceful handling of corrupt/empty/oversized PDFs
• get_search_content size guardrails: New maxChars parameter (default 30K, hard cap 100K) prevents oversized content from flooding LLM context
• Dynamic file offloading: tool_result interception writes large results (>30K) to secure temp files, replacing with preview + path so the model can bash grep instead of loading everything into context
• Documentation: README updated with PDF support, maxChars parameter, content size management section, and changelog

Files Changed

File	Change
package.json	pdf-parse dependency, version bump to 1.2.0
extract.ts	PDF detection and pdf-parse text extraction
extract.test.ts	3 PDF extraction tests
truncation.ts	New — content truncation with configurable limit
truncation.test.ts	7 truncation tests
offload.ts	New — secure temp file offloading for large content
offload.test.ts	8 offload tests
tool-params.ts	normalizeGetSearchContentInput with maxChars validation
index.ts	maxChars schema, truncation wiring, tool_result handler, cleanup
README.md	PDF support, maxChars docs, content size management, changelog

Test Plan

• 19 new tests added (3 PDF, 7 truncation, 8 offload, 1 constant)
• All 110 tests passing
• Manual: fetch a real PDF URL and confirm text extraction
• Manual: fetch large page and confirm file offload behavior

[socket-security]

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff	Package	Supply Chain Security	Vulnerability	Quality	Maintenance	License
	pdf-parse@​2.4.5

View full report