Skip to content

v1.12.0

Latest

Choose a tag to compare

@jakmro jakmro released this 08 Apr 13:48
· 3 commits to main since this release
Immutable release. Only release title and notes can be modified.
d5c922b

Changes:

  • Prompt prefill for KV cache warming and prefill benchmarking
  • Thinking/reasoning output support via enableThinking option
  • Speaker diarization
  • Speaker embedding extraction
  • Voice activity detection moved to new CactusAudio class (renamed
    from CactusVAD)
  • Language option for streaming transcription
  • Updated cactus engine to v1.12

Breaking

  • CactusVAD / useCactusVAD renamed to CactusAudio / useCactusAudio

Full Changelog: v1.10.4...v1.12.0