All pull requests

Pull requests

• fix(eval): estimate CIDEr corpus frequencies across all input rows

  #577·chivector opened Oct 25, 2026·

  ·

• fix(dedup): preserve field boundaries in exact multi-column deduplication

  #576·chivector opened Oct 25, 2026·

  ·

• fix(code): apply the Text file-name rule in CodeFileTypeContentFilter

  #575·raams95 opened Oct 22, 2026Contributor·

  ·
  1

• fix(mysql): identify reads by result set instead of the leading keyword

  #573·cyberspace-cs opened Oct 21, 2026Contributor·

  ·

• Add training data operators and opt-in resilient JSONL processing

  #572·yaoxiao67 opened Oct 21, 2026·

  ·
  1

• fix(sem-dedup): score similarity in blocks and keep the tail of a duplicate chain

  #571·cyberspace-cs opened Oct 21, 2026Contributor·

  ·

• fix(filters): hash the trailing characters in NgramHashDeduplicateFilter

  #570·cyberspace-cs opened Oct 20, 2026Contributor·

  ·

• fix(general_text): stop RemoveImageRefsRefiner from deleting long words

  #569·raams95 opened Oct 20, 2026Contributor·

  ·
  1

• fix(eval): return BLEU-n instead of BLEU-1 in BleuSampleEvaluator

  #567·raams95 opened Oct 20, 2026Contributor·

  ·
  1

• fix(dedup): normalize SimHash similarity by fingerprint size

  #565·BSAIPF opened Oct 20, 2026·

  ·

• fix(refine): load the stopword corpus once and tolerate missing cells

  #564·cyberspace-cs opened Oct 20, 2026Contributor·

  ·
  1

• fix(pipeline): harden compilation lifecycle and recovery

  #559·rikkurea-boop opened Oct 14, 2026Contributor·

  ·

• fix(streaming): prevent duplicate output when resuming after crashes

  #558·rikkurea-boop opened Oct 14, 2026Contributor·

  ·

• fix(storage): preserve zero-row batches and concurrent lazy flushes

  #550·rikkurea-boop opened Oct 13, 2026Contributor·

  ·

• fix(registry): harden enumeration, import recovery, and duplicate resolution

  #548·rikkurea-boop opened Oct 12, 2026Contributor·

  ·

• docs(litellm): clarify mutable instance configuration

  #546·rikkurea-boop opened Oct 12, 2026Contributor·

  ·

• fix(local-vlm): apply json schema guided decoding

  #545·rikkurea-boop opened Oct 12, 2026Contributor·

  ·

• fix(prompted-generator): propagate serving failures

  #544·rikkurea-boop opened Oct 12, 2026Contributor·

  ·

• fix(cache): respect selected cache across Text2Model and PDF2Model

  #540·rikkurea-boop opened Oct 12, 2026Contributor·

  ·