v0.1.2 - Precision OCR for URLs & Token Reconstruction

Screen-Pin 0.1.2

This release focuses on high-precision OCR extraction for web URLs, alphanumeric hashes, code tokens, and small screen text, eliminating character fusion and spurious spaces.

Highlights & Key Improvements

• High-Fidelity Link & Alphanumeric OCR: Replaced cubic spline filtering with Lanczos-3 interpolation in the image scaling pipeline. This preserves single-pixel stroke contrast in tiny screen fonts (~8px text height) and prevents character fusion (such as t5 muting into IS or 6m into EmM).
• Intelligent URL Reconstruction: Added automatic URL path reassembly (repair_line_urls) to eliminate spurious spaces accidentally inserted by word-segmenting engines within web addresses (e.g., PH 13iLq $\to$ PH13iLq).
• Google Docs/Sheets/Drive ID Disambiguation: Added prefix disambiguation for Google Drive file IDs (/d/l $\to$ /d/1), ensuring spreadsheet and document links remain 100% clickable and valid upon copying.
• Protected Web Link Routing in OCR Pipeline: Refined is_technical heuristics to prevent Tesseract language model priors from overwriting superior native Windows OCR recognition on web URLs and domains.
• Clippy & Linter Hardening: Refactored URL matching and character checks to meet strict clippy standards without warnings.

Downloads

• Windows x64:
  • NSIS installer
  • Portable ZIP
• macOS:
  • Apple Silicon (arm64)
  • Intel (x64)