OpenClaw 2026.6.8-beta.1: GLM-5.2 and Haiku 4.5 catalog, web-search auto-fallback removed
Beta prerelease adding GLM-5.2 and Claude Haiku 4.5 to the provider catalog, richer Telegram/WhatsApp delivery, and a native usage-footer renderer. Key-free web search providers no longer get selected automatically when no API-backed provider is configured — anyone relying on that implicit fallback must now opt in explicitly.
PUBLISHED2026-06-14
OBSERVED2026-08-11
AGE1mo
SOURCES1
- Providers: GLM-5.2 support and Claude Haiku 4.5 catalog entries; provider-qualified model IDs normalized across OpenRouter and Google Vertex paths
- Web search: key-free providers (Parallel Free, DuckDuckGo, Ollama, Codex Hosted Search) are explicit opt-ins instead of automatic fallbacks
- Telegram: structured text with tables, lists, expandable blockquotes, preserved intentional line breaks, CLI-backed rich replies; WhatsApp honors configured ACP bindings
- Usage hooks: built-in full footer renderer, default template, per-turn usage state, credential-aware limits, fixed-decimal formatting, warnings on broken templates rather than silent bad output
- Agent commands:
/btwworks in CLI-backed sessions; CLI usage-error exits are classified as failures, not successful runs - Gemini CLI runs in an isolated runtime home so ambient Google machine credentials cannot override the selected auth profile
- HTTP session/model override surfaces now require admin privileges
- Memory: oversized OpenAI embedding batches split before hitting 431s; SQLite skips WAL on NFS volumes; QMD search stays available in transient mode
- Discord auto-thread title generation capped at 60s and a 4,096-token reasoning budget, clamped to the model output cap
COMMUNITY
No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.