Add offset/continuation support for search_pdf when results are too large

[vlad-ds]

Problem

The search_pdf tool can return responses that are too large for the MCP context when searching yields many matches across a large PDF. This causes failures when the response exceeds context limits, preventing users from seeing all search results.

Proposed Solution

Add offset/continuation parameters to search_pdf to enable paginated search results:

• offset: Starting index for search results (default: 0)
• limit: Maximum number of matches to return (default: current behavior)
• Include metadata about total matches found and whether more are available
• Maintain consistent search state for pagination

Example Usage

// Get first 20 search results
search_pdf({ path: "document.pdf", search_pattern: "budget", offset: 0, limit: 20 })

// Get next 20 search results
search_pdf({ path: "document.pdf", search_pattern: "budget", offset: 20, limit: 20 })

Implementation Notes

• Should work with existing max_results parameter (offset within max_results)
• Include response metadata:
  • total_matches_found: Total number of matches discovered
  • has_more_results: Boolean indicating if more results are available
  • next_offset: Suggested offset for next request
• Consider caching search results temporarily to avoid re-scanning entire PDF
• Ensure consistent ordering of results across paginated requests

Related

This complements the existing max_results and max_pages_scanned parameters but focuses on chunking the response rather than limiting the search scope.

Priority

Medium - affects usability when searching large PDFs with many matches