weak shore
Problem Description
I’m getting millisecond-level response speed when using the built-in chat box inside oMLX.But when using OpenClaw to send the same Problem Description
I get millisecond responses in oMLX’s built-in chat box, but the same request takes 80–90 seconds via OpenClaw. Is this normal?
Environment
Device: Mac mini M4 24GBModel: qwen3.5-9b-mxl-4bit (local oMLX deployment)
Issue Details
oMLX native chat replies instantly. In OpenClaw, even a simple "Hello" needs 80–90s; complex prompts often timeout or stop mid-stream.
Troubleshooting confirms the oMLX backend and streaming work fine. When OpenClaw calls oMLX, Node.js CPU hits 99%–108%.
All OpenClaw Skills are disabled; only a browser is running for log checking.
Any help diagnosing this is greatly appreciated. I can provide additional logs and details anytime. Thanks!
22:58:18
warn
diagnostic
liveness warning: reasons=event_loop_delay interval=36s eventLoopDelayP99Ms=21.8 eventLoopDelayMaxMs=12415.1 eventLoopUtilization=0.509 cpuCoreRatio=0.531 active=1 waiting=0 queued=1
22:58:25
warn
agent/embedded
[trace:embedded-run] startup stages: runId=0b9d9e06-701e-48d0-808f-db8de7b7e776 sessionId=5afb2219-404f-45a2-9e5d-0281a8049404 phase=attempt-dispatch totalMs=13846 stages=workspace:0ms@0ms,runtime-plugins:1ms@1ms,hooks:0ms@1ms,model-resolution:1512ms@1513ms,auth:5850ms@7363ms,context-engine:0ms@7363ms,attempt-dispatch:6483ms@13846ms
22:59:00
warn
agent/embedded
[trace:embedded-run] prep stages: runId=0b9d9e06-701e-48d0-808f-db8de7b7e776 sessionId=5afb2219-404f-45a2-9e5d-0281a8049404 phase=stream-ready totalMs=35159 stages=workspace-sandbox:1ms@1ms,skills:0ms@1ms,core-plugin-tools:4279ms@4280ms,bootstrap-context:4ms@4284ms,bundle-tools:5079ms@9363ms,system-prompt:12368ms@21731ms,session-resource-loader:1019ms@22750ms,agent-session:2ms@22752ms,stream-setup:12407ms@35159ms
22:59:10
warn
fetch-timeout
fetch timeout reached; aborting operation
fickle agate