Optimized Sub-Agent Circulation and Certificate Rotation, OpenClaw New Version Fixes Gemini Context Truncation Bug

According to Perceive Beating monitoring, the personal AI assistant project OpenClaw has officially released version 2026.5.19. This incremental update focuses on multi-model fault tolerance, secure communication on mobile devices, and sub-agent lifecycle synchronization.

In terms of model ecosystem adaptation, the system has specifically addressed a critical disruption flaw related to Google Gemini. The new version will proactively discard Gemini thought signatures cut off by the system's compression mechanism before dialogue replay, preventing malformed Base64 data from causing the assistant's next response to terminate abruptly.

For domestic large-model fault tolerance, the system has reclassified the HTTP 429 payloads thrown when Moonshot (Kimi) balances are depleted. These errors are no longer rudely treated as general rate limits but are precisely identified as billing blocks, triggering the correct billing guidance and fallback to backup models.

On the cross-device communication layer, the Android client has introduced a seamless TLS certificate rotation mechanism. When the gateway's TLS fingerprint changes, the system will display both the new and old SHA-256 fingerprints for user verification and approval, completely replacing the previous hard crashes caused by fixed public keys.

The underlying sub-agent handover has also undergone meticulous race condition fixes. The new version mandates that handover signals in the queue must land completely within the parent context for the system to mark them as declared. This transformation, combined with the parent session's anti-timeout keep-alive mechanism, fundamentally eliminates the risk of losing sub-level execution results under extremely busy conditions.

Additionally, the browser plugin now prioritizes Chrome CDP-hosted diagnostics, eliminating false negative failures caused by cold-start HTTP probe races; the WhatsApp channel has completely opened up document-level delivery of high-resolution images and videos.

From eliminating cross-generation concurrent degradation to accurately identifying vendor error codes, OpenClaw is gradually solidifying the scheduling stability of multi-tiered intelligent agent systems in highly complex networks and long sessions by honing the fine boundaries of its runtime.