v0.30.0 v1.10.0
vectorize-io/hindsightSep 25, 2026 → Sep 26, 20260 days between releases
What's New
Releases belong to different repositories (vLLM-Omni vs ChatbotXIO).
Breaking Changes
- diffusion_batch_size has been removed; use max_num_seqs instead.
New Features
- Realtime event queue implementation
- Progressive dashboard charts and media loading
- Idempotency keys and retry headers added to API
- Bot messages can now be marked as read
- Realtime sender cost tracking
Improvements
- Unified full-duplex serving framework
- Native cross-stage KV and multimodal payload transfer
- Improved batching and memory efficiency
- Rebased onto vLLM 0.30.0
Bug Fixes
- Forward commentAnchor to getUserData prompts
- Close refactor audit gaps
- Refine state seeding and coverage
- Preserve cache consistency
- Fix bug upload pdf error
Upgrade Notes
These releases originate from different repositories (vLLM-Omni vs ChatbotXIO). v0.30.0 features a full-duplex serving framework, while v1.10.0 introduces an API with idempotency keys. Direct upgrade is not possible; a migration to the new repository is required.
The Two Releases
v0.30.0
Sep 25, 2026· 26 days ago
A major release for vLLM-Omni featuring a unified full-duplex serving framework, new model support (MAGI-2, MiniMax H3), and expanded hardware support.
Full release notesv1.10.0
Sep 26, 2026· 26 days ago
This release introduces a realtime event queue, progressive dashboard charts, and adds idempotency keys and retry headers to the API.
Full release notes