v2.2.0 v0.28.0
ruvnet/RuViewAug 28, 2026 → Aug 31, 20263 days between releases
What's New
v0.28.0 adds serving enhancements and speech stacks; v2.2.0 introduced a guided installer and Nasiko CLI bridge.
New Features
- Production-level serving enhancements for Minimax-H3 on GPU and NPU
- Scheduler-managed paged KV cache for diffusion models
- Expanded full-duplex and real-time speech stack
- New models including PersonaPlex, Nemotron VoiceChat, and IndexTTS 2.5
- New TTS integrations and improvements including TTS timestamps
Improvements
- Advanced MiniMax H3 deployment stack with modular pipelines and continuous batching
- Scheduler-managed paged KV cache for diffusion models
- Expanded full-duplex speech serving capabilities
- Quantization support for Qwen2.5-Omni NVFP4 W4A4 and SVDQuant W4A4
- Batched Chat Completions and speech token-usage headers
Upgrade Notes
These releases belong to different repositories (ECC and vLLM-Omni). For vLLM-Omni v0.28.0, note the rebase onto vLLM 0.28.0 and the introduction of the scheduler-managed paged KV cache.
The Two Releases
ECC 2.2.0: Guided Setup, Antigravity 2.0, and the Nasiko CLI Bridge
Aug 28, 2026· 2 months ago
v2.2.0 adds a guided installer for multiple AI harnesses, native Antigravity 2.0 support, and an experimental Nasiko CLI bridge, significantly expanding the distribution catalog.
Full release notesv0.28.0
Aug 31, 2026· 2 months ago
A major release featuring production-level serving enhancements for Minimax-H3, the introduction of a scheduler-managed paged KV cache for diffusion models, and a comprehensive full-duplex and real-time speech stack.
Full release notes