1.11.1 v0.64.10
medium risk
Textualize/textualSep 30, 2026 → Sep 30, 20260 days between releases
What's New
Shift from CV correctness fixes to Chat UI performance and security hardening.
Breaking Changes
- Explicit memory engine selection requires migration configuration
- Tool-output compaction enabled by default alters tool output formats
New Features
- Faster agent loop with warm prompt caching
- Chat UI improvements including collapsible source lists
- Explicit memory engine selection with migration support
- Enhanced MCP & Composio integrations
- Tool-output compaction enabled by default with REPL-style tools
Improvements
- Faster agent loop with warm prompt caching
- Chat UI improvements like collapsible source lists
- Sharper security protections for prompt-injection
- Browser IP blocking using full non-global range definitions
- Tool-output compaction for cleaner tool outputs
Bug Fixes
- Silence watchdog now warns instead of killing running turns
- Answered approvals no longer ghost into background deck
- Refusals now carry a clear deny reason
Upgrade Notes
Migrate memory engine configuration if you rely on the previous default. Tool-output compaction is now enabled by default, which may alter tool output formats.
The Two Releases
1.11.1
v1.11.1: Correctness Fixes for LoRA, TensorRT & Crowd mAP
Sep 30, 2026· 22 days ago
A correctness patch that fixes silent errors in LoRA checkpoints, TensorRT inference, and crowd mAP metrics, while renaming a gradient clipping method.
Full release notesv0.64.10
OpenHuman v0.64.10
Sep 30, 2026· 21 days ago
Focuses on performance, responsiveness, chat UI polish, memory engine selection, and security hardening.
Full release notes