All pull requests

Pull requests

• open_object: block_size=0 replaces the read_ahead flag

  #2126·shcheklein opened Oct 28, 2026Contributor·

  ·
  4

• fix(parquet): read only the picked columns with ParquetFileFormat

  #2123·shcheklein opened Oct 28, 2026Contributor·

  ·
  7

  

• fix(read_json): keep the chain's prefetch after nrows

  #2122·shcheklein opened Oct 28, 2026Contributor·

  ·
  7

  

• perf(parquet): prefetch small files in partial reads

  #2121·shcheklein opened Oct 28, 2026Contributor·

  ·
  13

  

• fix(deps): raise litellm cap, bump mkdocs-material for CVEs

  #2120·yathomasi opened Oct 28, 2026Contributor·

  ·
  7

  

• perf: read media metadata from headers, without a HEAD per file

  #2118·dmpetrov opened Oct 28, 2026Contributor·

  ·
  5

• Restore batch_map() and BatchMapper for batched inference

  #2117·dmpetrov opened Oct 28, 2026Contributor·

  ·
  3

• Checkpoints: keep finished rows when a UDF job is killed

  #2116·dmpetrov opened Oct 28, 2026Contributor·

  ·
  5

• Delta: drop results of overwritten and deleted files

  #2115·dmpetrov opened Oct 28, 2026Contributor·

  ·
  3

• README: data pipelines for robotics and physical AI, 2× faster with your coding agent

  #2114·dmpetrov opened Oct 28, 2026Contributor·22 tasks·

  ·
  4

• build(deps): update litellm requirement from <1.92,>=1.91.0 to >=1.91.0,<1.107

  maintenance
  #2110·dependabot[bot] opened Oct 26, 2026Bot·

  ·
  2

  

• Refresh listings when an object is deleted or changed in place

  #2108·dreadatour opened Oct 26, 2026Contributor·

  ·
  17

  

• Skip the etag check for pinned versions on version-aware clients

  #2106·dreadatour opened Oct 26, 2026Contributor·

  ·
  40

  

• fix: bound file slice reads by buffer byte capacity

  #2105·chivector opened Oct 26, 2026·

  ·

• fix(http): preserve the client host for dotted relative paths

  #2102·chivector opened Oct 25, 2026·

  ·
  1

• fix(cache): clean temporary files after interrupted downloads

  #2101·chivector opened Oct 25, 2026·

  ·
  1

• fix(http): normalize callbacks when opening uncached files

  #2100·chivector opened Oct 25, 2026·

  ·
  1

• Preserve empty path segments in rebasing, saves and listings

  #2088·dmpetrov opened Oct 21, 2026Contributor·

  ·
  8

• Azure: read public containers when no identity is configured

  #2083·dmpetrov opened Oct 20, 2026Contributor·

  ·
  3

• Fast UDF: stats response once in 3 sec

  #2080·dmpetrov opened Oct 19, 2026Contributor·

  ·
  7

  

• Skip cp/pull destinations that resolve outside the output directory

  #2062·syahra712 opened Oct 17, 2026·

  ·
  1

• Fast anonymous reads of public buckets; accept boto3 key names

  #2059·dmpetrov opened Oct 17, 2026Contributor·

  ·
  15

  

• Small buckets: one HeadBucket, import only the backend in use

  #2055·dmpetrov opened Oct 16, 2026Contributor·

  ·
  13

  

• Listing: support deleted markers

  #2052·dmpetrov opened Oct 15, 2026Contributor·

  ·
  3

• union and optional docs

  #2043·dmpetrov opened Oct 13, 2026Contributor·

  ·
  3