#problem with model timeout

1 messages · Page 1 of 1 (latest)

undone ether
#

@clever glacier the problem is i prompt it to do something(like create a website) and sometimes it starts to reply and start doing stuff, sometimes it doenst say anything, but the end is always the same, after 30 seconds it goes into some kind of timeout, thats why im trying to see if i can make openclaw wait longer
here is a snippet of the tui demonstrating the problem(i could not upload a pic so i copypaste it)
can you continue making the site

This response is taking longer than expected. Send another message to continue.
local ready | idle
agent main | session main | ollama/glm-4.7-flash:latest | think medium | tokens 12k/203k (6%)

clever glacierBOT
#

Yep — that specific line is the TUI’s 30s streaming watchdog, not necessarily the whole agent runtime timeout. It fires when the TUI hasn’t seen new streamed output for ~30s, then tells you to send another message to continue.

For slow Ollama/local models, the knob that usually matters is the provider request timeout, e.g. in ~/.openclaw/openclaw.json:

{
  models: {
    providers: {
      ollama: {
        timeoutSeconds: 300
      }
    }
  }
}

Then restart the gateway/TUI. You can also launch TUI with a longer agent wait timeout:

# 
openclaw tui --timeout-ms 300000

But: with ollama/glm-4.7-flash, if it’s silent for long stretches or weak at tool/file work, extending timeouts may only hide the real issue. Local models can stall or fail tool-calling; if this is for building a site, try a more tool-reliable/cloud model as a sanity check.

Can you paste:

openclaw --version
openclaw status --all
openclaw models status

Docs: https://docs.openclaw.ai/providers/ollama and https://docs.openclaw.ai/web/tui