[Optimized Sub-Agent Transition and Certificate Rotation, OpenClaw New Version Fixes Gemini Context Truncation Error] According to monitoring by Beating, the personal AI agent project OpenClaw has officially released version v2026.5.19. This incremental update focuses on multi-model fault tolerance, secure communication on mobile platforms, and lifecycle synchronization of sub-agents. In terms of model ecosystem adaptation, the system specifically addressed a critical interruption defect for Google Gemini. The new version proactively discards Gemini thought signatures truncated by the system's compression mechanism before replaying conversations, preventing malformed Base64 data from causing the assistant's next response to terminate abruptly. For fault tolerance with domestic large models, the system reclassified HTTP 429 payloads thrown when Moonshot (Kimi) balances are exhausted. These errors are no longer crudely treated as general rate limits but are now precisely identified as billing blocks, triggering proper billing guidance and fallback to alternative models. On the cross-device communication layer, the Android client introduced a smooth TLS certificate rotation mechanism. When the gateway's TLS fingerprint changes, the system will display both the old and new SHA-256 fingerprints for user verification and approval, completely replacing the previous hard crashes caused by public key pinning. The underlying sub-agent transition has also undergone rigorous race condition fixes. The new version mandates that handoff signals in the queue must fully integrate into the parent context before being marked as declared. This improvement, combined with the parent session's timeout-resistant keep-alive mechanism, fundamentally eliminates the risk of losing sub-agent results under extreme load conditions. Additionally, the browser extension now prioritizes Chrome CDP-managed diagnostics, eliminating false failure reports caused by cold-start HTTP probe competition. The WhatsApp channel has fully lifted restrictions on document-level delivery of original images and videos. From eliminating cross-generation concurrency losses to accurately identifying vendor error codes, OpenClaw is refining runtime boundaries to gradually solidify the scheduling stability of multi-level agent systems under highly complex networks and long sessions. [Original Link]