-
-
Notifications
You must be signed in to change notification settings - Fork 80.7k
Model Setup 'Test & use' fails on configless gateways: prepared model runtime owner was not committed after replacement #111520
Copy link
Copy link
Closed
Labels
P0Emergency: data loss, security bypass, crash loop, or unusable core runtime.Emergency: data loss, security bypass, crash loop, or unusable core runtime.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.ClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.ClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.ClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:auth-providerAuth, provider routing, model choice, or SecretRef resolution may break.Auth, provider routing, model choice, or SecretRef resolution may break.impact:ux-release-blockerA non-technical user is blocked without terminal, logs, config, or support.A non-technical user is blocked without terminal, logs, config, or support.issue-rating: 🦪 silver shellfishThin issue quality; more reproduction proof or environment detail is needed.Thin issue quality; more reproduction proof or environment detail is needed.maintainerMaintainer-authored PRMaintainer-authored PR
Description
Metadata
Metadata
Assignees
Labels
P0Emergency: data loss, security bypass, crash loop, or unusable core runtime.Emergency: data loss, security bypass, crash loop, or unusable core runtime.clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.ClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.ClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.ClawSweeper does not recommend queueing a new automated fix PR for this issue.impact:auth-providerAuth, provider routing, model choice, or SecretRef resolution may break.Auth, provider routing, model choice, or SecretRef resolution may break.impact:ux-release-blockerA non-technical user is blocked without terminal, logs, config, or support.A non-technical user is blocked without terminal, logs, config, or support.issue-rating: 🦪 silver shellfishThin issue quality; more reproduction proof or environment detail is needed.Thin issue quality; more reproduction proof or environment detail is needed.maintainerMaintainer-authored PRMaintainer-authored PR
Type
Fields
Priority
None yet
Summary
On a gateway with no explicit model configured (model resolved from a provider env key, e.g.
OPENAI_API_KEY→openai/gpt-5.6-sol), the Control UI first-run Model Setup → "Test & use" flow fails deterministically with:Because activation never commits,
openclaw.setup.detectkeeps returningsetupComplete: false, so every operator-admin device is permanently bounced to/settings/model-setup?firstRun=1from the default chat landing. On a multi-user trusted-proxy gateway this locks all admins out of the chat UI.Where it throws
rebindInputToCommittedConfiguredOwner(src/agents/prepared-model-runtime.owner.ts:101) requires exactly one committed owner withprovenance === "configured":candidates.length === 0and the rebind throws every time — this is not a transient race. Reproduced repeatedly on an idle gateway.Repro
agents.defaults.modelunset andmodelsunset in openclaw.json; provider key present via env (OPENAI_API_KEY).firstRun=1).Workaround (verified)
openclaw models set openai/gpt-5.6-solon the gateway host (writes the explicit default into openclaw.json) + gateway restart.setup.detectthen reports complete and the takeover disappears for all devices.Environment
mainagent at/home/openclaw/.openclaw/agents/main/agent.openclaw.setup.activaterequest completing but the prepared-runtime rebind rejecting with the error above.Side observation
While debugging, the gateway logs show a steady 5s-interval spam of
question.listresponses failing withFORBIDDEN missing scope: operator.questionsfrom a connected Control UI client — likely a UI poller not gating on granted scopes. Filing here as a footnote since it surfaced in the same logs; can split into its own issue if preferred.