#03:32:56+00:00 warn agent/embedded {"subsystem":"agent/embedded"} [timeout-compaction] LLM timed out
1 messages · Page 1 of 1 (latest)
That warning by itself isn’t usually fatal — it means the model request timed out while the prompt was already huge (84% of context), so OpenClaw is trying an automatic compaction + retry.
Please share:
openclaw --version
openclaw status --all
And the follow-up log lines after that diagId, especially whether you see:
compaction succeeded ... retrying promptcompaction did not reduce context- another timeout/failover error
You can grab recent logs with: