openclaw - 💡(How to fix) Fix [Bug]: OpenCalw hang after a couple of dialog with skill call using local model qwen3.5 27b [1 participants]

05:17:44+00:00 warn agent/embedded {"subsystem":"agent/embedded"} [timeout-compaction] LLM timed out with high prompt token usage (95%); attempting compaction before retry (attempt 1/2) diagId=ovf-mnh0wwt9-pLPYYQ 05:17:44+00:00 warn agent/embedded {"subsystem":"agent/embedded"} [compaction-diag] end runId=aaef06dd-cb89-45a6-8d54-95ddfb4cf834 sessionKey=agent:main:main diagId=ovf-mnh0wwt9-pLPYYQ trigger=timeout_recovery provider=ollama/qwen3.5:27b attempt=1 maxAttempts=2 outcome=failed reason=already_compacted_recently durationMs=276 05:17:44+00:00 warn agent/embedded {"subsystem":"agent/embedded"} [timeout-compaction] compaction did not reduce context for ollama/qwen3.5:27b; falling through to normal handling 05:17:44+00:00 warn agent/embedded {"subsystem":"agent/embedded"} Profile ollama:default timed out. Trying next account... 05:17:44+00:00 warn agent/embedded {"subsystem":"agent/embedded"} {"event":"embedded_run_failover_decision","tags":["error_handling","failover","assistant","surface_error"],"runId":"aaef06dd-cb89-45a6-8d54-95ddfb4cf834","stage":"assistant","decision":"surface_error","failoverReason":"timeout","profileFailureReason":"timeout","provider":"ollama","model":"qwen3.5:27b","profileId":"sha256:9c018ec112cf","fallbackConfigured":false,"timedOut":true,"aborted":true} embedded run failover decision 05:21:25+00:00 warn agent/embedded {"subsystem":"agent/embedded"} Profile ollama:default timed out. Trying next account... 05:21:25+00:00 warn agent/embedded {"subsystem":"agent/embedded"} {"event":"embedded_run_failover_decision","tags":["error_handling","failover","assistant","surface_error"],"runId":"f46b19d7-bb5f-4da0-9803-18d87d1e5b9b","stage":"assistant","decision":"surface_error","failoverReason":"timeout","profileFailureReason":"timeout","provider":"ollama","model":"qwen3.5:27b","profileId":"sha256:9c018ec112cf","fallbackConfigured":false,"timedOut":true,"aborted":true} embedded run failover decision

TL;DR

The issue can be mitigated by adjusting the timeout settings or optimizing the model usage to reduce prompt token usage.

Guidance

• Review the logs to identify patterns in timeout occurrences and high prompt token usage, which may indicate a need for model optimization or increased timeout thresholds.
• Consider adjusting the timeout settings for the qwen3.5:27b model to allow for longer processing times, potentially reducing the frequency of timeouts.
• Investigate the compaction-safeguard and timeout-compaction mechanisms to understand their roles in handling high prompt token usage and timeouts, and assess whether adjustments to these mechanisms can help mitigate the issue.
• Evaluate the feasibility of implementing a fallback configuration for the ollama provider to handle timeouts more gracefully.

Example

No specific code snippet can be provided without more context on the implementation details of OpenClaw and its interaction with the qwen3.5:27b model. However, the logs suggest that the issue is related to timeout handling and model performance.

Notes

The provided information lacks specific details on the OpenClaw configuration, model optimization, and the exact nature of the "hang" behavior. Further investigation into these areas is necessary to provide a more targeted solution.

Recommendation

Apply workaround: Adjust timeout settings or optimize model usage to reduce prompt token usage, as this approach directly addresses the identified issues of timeouts and high prompt token usage, potentially mitigating the hanging behavior without requiring a full understanding of the underlying system's intricacies.