v0.3.0

0.3.0 (2026-04-25)

Features

• add debugging script for extracting tables from page 18 of the financial law PDF (3583dda)
• add script to extract tables from page 58 of the financial law PDF (58bd66c)
• added debug for small number (7c111bc)
• enhance table extraction and OCR region handling for improved accuracy (1e58eb3)
• enhance table extraction fallback logic and add debugging scripts for strategy comparison (cbd1793)
• fix table selection (bee6ea2)
• fixed table to be more accurate (042cc9c)
• implement container discard security layer for table extraction (79d8b9c)
• implement topmost linked reference strategy for robust footnote detection (65eb6c6)
• simplify RAG table formatting to plain CSV style without tags for better LLM context (becfba3)
• sync between debug and extract (2173b52)
• update table extraction format and add debugging script for specific pages (db1ad12)
• use full-page OCR if any unreadable image regions are detected (c9bbfc9)

Bug Fixes

• bug related to footer (d97ccde)
• ensure topless and bottomless tables are detected and extracted (c01b168)
• ignore tiny symbols and footnote markers in image OCR extraction (fb856b1)
• test new approach for footers (ffb0379)

Documentation

• add image assets for README (5df661f)
• add visual examples of edge-case table extraction to README (c354c88)
• update README with new CSV table format and advanced edge-case handling (0253257)