Summary

"Verifying" model stage fails if server doesn't issue response in 10 seconds. This is too short for let's say, Qwen3 14B running on a laptop with a small GPU, llama.cpp based server. Please make it 2 minutes. That won't hurt anyone.

Steps to reproduce

Use a local model that takes longer that 10 seconds to finish its verification response.

Expected behavior

Verification step should be more patient. Wait 2 minutes.

Actual behavior

Verification fails. Have to manually configure the local model -OR- temporarily use a fast small mode on LLM server.

OpenClaw version

2026.2.26

Operating system

Windows 11, WSL Ubuntu

Install method

install script

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

https://github.com/openclaw/openclaw/blob/main/src/commands/onboard-custom.ts

Current value:

const VERIFY_TIMEOUT_MS = 10000;

Desired value:

const VERIFY_TIMEOUT_MS = 120000;