LocalAI: stop backends stuck busy for 10 minutes
Agent-pool work kept the Tesla T4 at 94 C after the agents were paused, and killing the backend only made it reload. Enable the busy watchdog so no single request can hold the GPU indefinitely. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
+2
-1
@@ -10,6 +10,7 @@
|
||||
localai_debug: "true"
|
||||
localai_watchdog_idle: "true"
|
||||
localai_watchdog_idle_timeout: "15m"
|
||||
localai_watchdog_busy: "false"
|
||||
localai_watchdog_busy: "true"
|
||||
localai_watchdog_busy_timeout: "10m"
|
||||
localai_p2p: "false"
|
||||
localai_federated: "false"
|
||||
|
||||
Submodule roles/ericomeehan.localai updated: 0559783297...dd5f03f074
Reference in New Issue
Block a user