openclaw - 💡(How to fix) Fix [Bug]: Gateway timeouts on all OpenAI calls after auto-update to 2026.5.2 [3 comments, 4 participants]

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

After automatic update from 2026.4.29 to 2026.5.2, agent never replies. models.list takes 18-40s, event_loop_delay up to 17000ms, profile sha256:a0ebe3f799a6 times out on every run. API key works via direct curl. Setting api=openai-completions did not fix it.

Steps to reproduce

1. Install OpenClaw 2026.5.2 on Ubuntu 24.04 VPS
2. Configure openai/gpt-4.1-mini as the agent model
3. Open webchat and send any message to the agent
4. Observe agent shows "thinking" indefinitely and never replies
5. Check logs: models.list takes 18-40s, event_loop_delay up to 17000ms, profile sha256:a0ebe3f799a6 times out on every embedded run

Expected behavior

Agent replies normally as it did on version 2026.4.29

Actual behavior

Agent shows "thinking" indefinitely and never replies. Logs show:

• models.list takes 18-40 seconds on every connection
• event_loop_delay up to 17000ms
• llm-idle-timeout: openai/gpt-4.1-mini produced no reply before the idle watchdog
• Profile sha256:a0ebe3f799a6 times out on every embedded run
• fetch timeout after 60000ms on api.openai.com/v1/responses
• API key works perfectly via direct curl to /v1/chat/completions

OpenClaw version

2026.5.2 (8b2a6e5)

Operating system

Ubuntu 24.04

Install method

npm global

Model

openai/gpt-4.1-mini

Provider / routing chain

openclaw -> api.openai.com/v1

Additional provider/model setup details

api set to openai-completions. Direct curl to api.openai.com/v1/chat/completions works with same key. Timeout occurs on /v1/responses endpoint which OpenClaw uses internally.

Logs, screenshots, and evidence

[agent/embedded] llm-idle-timeout: openai/gpt-4.1-mini produced no reply before the idle watchdog
[fetch-timeout] fetch timeout after 60000ms url=https://api.openai.com/v1/responses
[diagnostic] liveness warning: eventLoopDelayMaxMs=17498 eventLoopUtilization=0.999
[agent/embedded] Profile openai:default timed out. profileId=sha256:a0ebe3f799a6
models.list took 40396ms on every connection

Impact and severity

Affected: Single user, all webchat sessions Severity: Critical (blocks all agent functionality completely) Frequency: 100% of attempts since auto-update to 2026.5.2 Consequence: Agent is completely unusable, all messages timeout

Additional information

Last known good version: 2026.4.29 First known bad version: 2026.5.2 Broke after automatic update while machine was idle. Tried setting models.providers.openai.api=openai-completions — did not fix. Direct curl to /v1/chat/completions works perfectly with same API key.