#Doesn´t give any respont
1 messages · Page 1 of 1 (latest)
First: what does openclaw --version show?
For a Web UI run stuck for 12h where /stop and the button do nothing, I’d restart the Gateway process. That clears in-memory running tool/process state and usually unsticks the UI:
openclaw gateway restart
openclaw gateway status
Then open the dashboard again and start a fresh chat/session with /reset or /new if the old thread still looks wedged.
If it comes back stuck after restart, please paste the safe diagnostics output:
openclaw status --all
openclaw logs --follow
Docs for the debug flow: https://docs.openclaw.ai/help/faq#first-60-seconds-if-something-is-broken and gateway troubleshooting: https://docs.openclaw.ai/gateway/troubleshooting
I did the restart and it says everthing is fine. I´m on the Version 2026.5.22 and the Web-UI stays in process
If the Gateway restart reports healthy but that same Web UI chat still says “in process”, it may just be the browser/client state for that tab being stale after the gateway lost the old run.
Try this in order:
openclaw dashboard
Open the fresh URL it prints, then hard-refresh the page (Ctrl+Shift+R / Cmd+Shift+R). In that chat, send:
/new
If the input is still locked and you can’t send /new, clear the OpenClaw dashboard site data for that browser tab, reopen openclaw dashboard, and start a new session.
If it still shows stuck after that, please paste these outputs so we can see whether the server still thinks there’s an active run or whether it’s only the UI:
openclaw sessions --all-agents --active 1440 --json
openclaw status --all
Don’t paste secrets/tokens if anything looks sensitive. Docs: https://docs.openclaw.ai/cli/sessions and https://docs.openclaw.ai/help/faq#first-60-seconds-if-something-is-broken
I was make a new session but it doesnt fix the problem. I have do an update, restart the gateway, disabeled the API Key from OpenRouter, The doctor command. Nothing was doing anything
At this point I’d stop changing settings and capture the failing run. Disabling the OpenRouter key only helps if you also switched to another working provider/model; otherwise the new session can still start and then fail/hang on model auth.
Please run these and paste the relevant output:
openclaw --version
openclaw models status
openclaw status --all
Then open logs in one terminal:
openclaw logs --follow
While that is running, start a brand-new Web UI session and send a tiny prompt like hi. Copy the log lines that appear right after that. That should show whether this is model auth/provider, a stuck tool/process, or a UI abort bug.
Also: if the dashboard says “in process” but logs show an error, paste the error text. If this is a repeatable 2026.5.22 bug after a clean run, it should go to GitHub with the sanitized logs: https://github.com/openclaw/openclaw/issues
It doesnt show any error
No error in logs is useful: it usually means the prompt is not actually reaching the gateway, or the dashboard WebSocket is connected to the wrong/stale gateway/auth state.
Please do this clean test:
openclaw gateway status
openclaw dashboard
Open exactly the URL printed by openclaw dashboard in a private/incognito window, then send hi in a new session.
If logs still show nothing when you press send, check the browser dev console/network tab for failed WebSocket/auth errors. The dashboard is an admin WebSocket surface, so stale token/sessionStorage can make the UI look connected-ish while actions don’t land.
If logs do show the request but no reply appears, paste:
openclaw models status
openclaw infer model run --prompt "hi" --json
That separates “Web UI can’t submit” from “model/provider can’t answer”. Dashboard/auth docs: https://docs.openclaw.ai/web/dashboard
The Web-UI shows compacting over and over again
That changes the diagnosis: a fresh session repeatedly compacting usually means the initial context is too large for the selected model, or the compaction model/provider is failing and retrying. Since you disabled OpenRouter, check whether compaction or the default model still points at OpenRouter.
Please run:
openclaw models status
openclaw config get agents.defaults.model
openclaw config get agents.defaults.compaction
Also check if any startup context file got huge:
wc -c ~/.openclaw/workspace/AGENTS.md ~/.openclaw/workspace/MEMORY.md ~/.openclaw/workspace/BOOTSTRAP.md 2>/dev/null
If BOOTSTRAP.md exists and is large/old, that is suspicious. It should only be first-run scaffolding; move it aside, restart, and try a new session:
mv ~/.openclaw/workspace/BOOTSTRAP.md ~/.openclaw/workspace/BOOTSTRAP.md.bak
openclaw gateway restart
If your active model is a small/local model, switch to a larger/cloud model for the main agent or set a dedicated compaction model in agents.defaults.compaction.model. Docs: https://docs.openclaw.ai/concepts/compaction and https://docs.openclaw.ai/gateway/configuration-reference#agentsdefaultscompaction
Ok, /reset hat mir da geholfen. Ich habe die Models nochmal entfernt und neu hinzugefügt, danach /reset und jetzt geht es wieder. Danke
Super, freut mich. Dann war es sehr wahrscheinlich eine kaputte/alte Model-Konfiguration plus Session-State, und /reset hat den hängenden Kontext sauber neu gestartet.
Wenn das wieder auftaucht: erst openclaw models status, dann /reset, bevor du zu viel an der Gateway-Konfiguration änderst.
Wenn das damit wirklich gelöst ist, bitte die Lösung markieren: Rechtsklick auf die hilfreiche Nachricht → Apps → ✅ Mark Solution. Das hilft Answer Overflow, die Lösung für andere zu indexieren.