PDF Redaction MCP Server

# Changelog All notable changes to this project will be documented in this file. ## [Unreleased] ### Added - **Batch text redaction**: `redact_text` tool now accepts a list of texts (`texts_to_redact`) instead of a single text, enabling efficient batch processing - **Redaction tracking**: Server now maintains a record of all texts marked for redaction for each PDF - **Automatic duplicate prevention**: The `redact_text` tool automatically skips texts that have already been redacted - **New tool: `list_applied_redactions`**: Lists all texts that have been marked for redaction for a specific PDF or all loaded PDFs - Accepts optional `pdf_path` parameter to filter by specific PDF - Returns detailed list of redacted texts per PDF - Useful for progress tracking and audit trails ### Changed - **Breaking Change**: `redact_text` parameter changed from `text_to_redact` (string) to `texts_to_redact` (list of strings) - `redact_text` now returns detailed summary including: - How many texts were newly redacted - Which texts were skipped (already redacted) - Instance counts per text - Page distribution - `close_pdf` now also clears redaction tracking data for the closed PDF - `load_pdf` initializes redaction tracking for newly loaded PDFs ### Performance Improvements - Batch redaction significantly reduces overhead compared to multiple individual calls - Single PDF scan for multiple texts instead of multiple scans - O(1) duplicate detection using hash-based tracking ### Documentation - Updated README.md with batch redaction examples and performance tips - Updated QUICKSTART.md with batch workflow examples - Updated PROJECT_SUMMARY.md with architectural details of new tracking system - Added performance comparison section showing benefits of batch operations - All examples now demonstrate batch redaction as the recommended approach ### Migration Guide **Old API (v0.1.0):** ```python await redact_text(pdf_path, text_to_redact="John Doe") await redact_text(pdf_path, text_to_redact="123-45-6789") ``` **New API (v0.2.0):** ```python # Single text (must be in a list) await redact_text(pdf_path, texts_to_redact=["John Doe"]) # Multiple texts (recommended) await redact_text(pdf_path, texts_to_redact=["John Doe", "123-45-6789"]) ``` **For MCP Clients:** The change is transparent - just pass multiple texts in the instruction: ``` # Old way (still works but less efficient) Redact "John Doe" in document.pdf Redact "123-45-6789" in document.pdf # New way (recommended) Redact ["John Doe", "123-45-6789"] in document.pdf ``` ## [0.1.0] - 2025-10-10 ### Initial Release - Basic PDF loading and text extraction - Single-text redaction - Area-based redaction - PDF saving with redactions applied - PDF resource management - MCP server implementation with FastMCP 2