Spaces:
Sleeping
Sleeping
Commit History
6 days ago
-
Support Chinese OCR
Verified signatureAdd Simplified/Traditional Chinese + English selection with pinned, verified recognition models. Preserve OCR limits and legacy defaults. Fix false deskew from filled charts and retain Chinese strokes. Validated with 36 real converter regression checks, including English scans, Chinese numeric tables, image controls, mixed pages and failure paths.aa4419a
7 days ago
-
Recognise clear English scan pages before Word and Excel extraction
Verified signatureAdd bounded Tesseract OCR in the existing service, with scan-page and time limits, numeric/column regression coverage, a confidence gate and mixed-document preservation. Require client opt-in so existing pages do not silently start OCR. Prefer detected table cells over alternate Markdown layouts.3ab1ac4 -
Improve PDF extraction accuracy and preserve document content
Verified signatureKeep one extraction per source table, numeric spreadsheet amounts and literal identifiers. Preserve Word page boundaries, source heading sizes, table widths and clear two-column reading order. Reject image-only pages requiring OCR rather than returning incomplete content.126bc03
-
Upgrade document parser dependencies
Verified signaturedef5409 -
Create PDF table extraction API
Verified signature92fea71 -
initial commit
Verified signature86d0fd9