v0.64.0 v0.64.3
low risk
mlc-ai/web-llmSep 25, 2026 → Sep 25, 20260 days between releases
What's New
Shifts focus from intelligence upgrades to reliability and speed improvements
New Features
- Optional capture of full successful inference SSE responses for debugging
Improvements
- Agents now focus on the latest user request to avoid loops
- Chat recovery and streaming resilience are improved
- Warm-session completions finalize progress more reliably
- Completed model streams retry once if they return no answer
- Faster turn startup and clearer billing surfaces
Bug Fixes
- Agents stop unproductive tool loops sooner with smarter failure classification
- Chat transcript and app-state recovery is more robust after failed writes
- Completed model streams retry once if they return no answer
- Agents avoid getting pulled back into earlier greetings
Upgrade Notes
This patch release focuses on stability and recovery. No configuration changes are required, though the new SSE response capture feature is optional for debugging.
The Two Releases
v0.64.0
OpenHuman v0.64.0
Sep 25, 2026· 27 days ago
The provided release notes are for OpenHuman, not mlc-ai/web-llm.
Full release notesv0.64.3
OpenHuman v0.64.3
Sep 25, 2026· 26 days ago
The provided release notes are for OpenHuman, not mlc-ai/web-llm.
Full release notes