GeezerVison

Pull requests

• fix: Hindi and Bengali paragraphs misclassified as Title

  #4695·shlokbam opened Nov 11, 2026·

  ·

• feat: report page progress from partition_pdf and partition_image

  #4694·Rutvij97 opened Nov 10, 2026·4 tasks done·

  ·
  1

• fix(docx): stop long vertical merges from dropping table rows from text_as_html

  #4693·inchang-ing opened Nov 10, 2026·

  ·
  1

• fix(chunking): measure inter-chunk overlap in tokens under token chunking

  #4691·davrens101 opened Nov 10, 2026·

  ·

• fix: preserve spaces around inline formatting in table chunks

  #4690·peterolkhov opened Nov 9, 2026·1 of 2 tasks·

  ·

• fix(text): treat a lone carriage return as a line break

  #4689·GPT-AGI opened Nov 9, 2026·1 of 3 tasks·

  ·

• fix(pptx): order text inside a moved group by its position on the slide

  #4688·chrikrah opened Nov 9, 2026Contributor·1 of 3 tasks·

  ·

• perf(chunking): normalize each table cell once when splitting tables (about 1.7x faster)

  #4687·deadstrobe5 opened Nov 9, 2026Contributor·1 of 3 tasks·

  ·

• fix(cleaners): allow any character in remove_sentence_punctuation exclusions

  #4686·anubhavlal07 opened Nov 9, 2026·2 of 3 tasks·

  ·

• fix(metrics): honor agg_name in get_mean_grouping

  #4685·anubhavlal07 opened Nov 9, 2026·2 of 3 tasks·

  ·

• fix(pptx): preserve merged table cell spans

  #4684·GPT-AGI opened Nov 9, 2026·1 of 3 tasks·

  ·

• fix(csv): keep every field of CSV and TSV rows wider than the first row

  #4683·iammanoj807 opened Nov 9, 2026·2 of 3 tasks·

  ·
  1

• perf(pdf): vectorize aggregate_embedded_text_by_block's IoU and drop redundant slicing

  #4681·KRRT7 opened Nov 9, 2026Contributor·

  ·

• perf(pdf): skip the per-character word walk for text boxes with no overlapping link

  #4680·KRRT7 opened Nov 9, 2026Contributor·

  ·

• perf(text_type): run the non-alpha check before the spaCy-backed cap-ratio check

  #4679·KRRT7 opened Nov 9, 2026Contributor·

  ·

• fix: Accept non-seekable file-like objects in partition()

  #4678·shashank11yadav opened Nov 9, 2026·1 of 2 tasks·

  ·
  1

• fix(documents): give elements that concatenate to the same hash input distinct ids

  #4676·Eapp1e opened Nov 8, 2026Contributor·1 of 3 tasks·

  ·

• fix(staging): keep joined text within the model attention window

  #4674·henricop opened Nov 7, 2026·2 of 3 tasks·

  ·
  1

• fix(cleaners): honor paragraph_split in blank_line_grouper

  #4672·mmporong opened Nov 7, 2026·1 of 3 tasks·

  ·

• fix(email): keep the bytes of a binary file inside a multipart/* attachment

  #4671·shashank11yadav opened Nov 6, 2026·1 of 3 tasks·

  ·
  1

• chore: start 0.27.38-dev after 0.27.37 release

  #4669·ryannikolaidis opened Nov 6, 2026Contributor·

  ·

• fix(embed): default VertexAI embedding model to gemini-embedding-001

  #4667·pcbeingused333 opened Nov 6, 2026·1 of 3 tasks·

  ·

• fix(html): preserve multi-block list items and their child elements

  #4666·ssingh231219 opened Nov 5, 2026·

  ·
  1

• fix(staging): render ListItem elements as Markdown list items

  #4665·kennywu411-dev opened Nov 5, 2026·

  ·

• fix: complete rel values on blank target links

  #4664·Vietpr opened Nov 4, 2026·1 of 2 tasks·

  ·
  1