v0.2.16

L-Chris released this 05 Aug 09:05

· 3 commits to main since this release

a4033fc

Feature

Added provider-based PDF document OCR with PaddleOCR and MinerU, cached results, and switchable original/reflow views.
Added native reflow rendering for OCR images, formulas, and structured tables, including merged cells and pagination.

Improvement

Added direct multimodal scanned-PDF summaries that send page images to vision models in a single request.
Made PDF view switching reuse the active reader and persist mode separately, avoiding full-book reopen and OCR cache rewrites.
Simplified OCR setup with built-in official service endpoints and secure token storage.

Fix

Fixed OCR mode changes that required extra interaction before content rerendered and corrected hidden-title alignment.
Fixed missing OCR images, partial formula parse failures, and lost centered caption/formula alignment.

Assets 3