#Model reply slowness

1 messages · Page 1 of 1 (latest)

low beacon
#

Anyone feels the openclaw with codex oauth is taking more time for response. It was like 2 sec for a response to hi, now it takes like 20 sec. Anyone else felt that? Or I need check some settings?

opal deltaBOT
prime boughBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

opal deltaBOT
lament vesselBOT
#

I’m checking the docs and source for anything specific to Codex OAuth latency, so I can give Macx a real answer instead of guessing.

#

Could be real upstream latency, but there are a couple OpenClaw-specific things worth checking.

For Codex OAuth, OpenClaw currently uses auto transport by default, which is WebSocket first with SSE fallback, and /fast maps to priority processing for openai-codex/*. So if replies suddenly got much slower, I’d check these first:

  1. openclaw --version
  2. openclaw status
  3. openclaw models status
  4. openclaw doctor

Also worth trying:

  • /fast status
  • /fast on

One more sneaky one: if you have old custom settings under models.providers.openai-codex, they can shadow the built-in Codex OAuth path and hurt routing/fallback behavior. openclaw doctor should warn about that.

Docs: