Turn per-user preference personalization on or off for this workspace.

Owner or admin only: it changes what every member's recalls do. The value is pinned on the workspace's Tenant CR and applied by the tenant operator as a config change. When the requested value equals the fleet value (which the control plane reads off the operator's forwarded config, so it is the real one), the pin is cleared instead and the workspace follows the fleet again — a fleet-wide switch keeps reaching it.

The workspace pod is killed and recreated. The tenant chart deploys with strategy: Recreate and no termination grace period, so the pod is gone within 30 s of the call: in-flight requests fail, and a running cognify is interrupted and rolled back on the next boot (recover_stale_pipeline_runs), so its documents have to be processed again. Warn the user before flipping this while an ingest is running. The new value is live after that restart, not on return, and a second flip is refused (409) until that restart has settled. Turning it off stops reading and writing preferences; preference data already stored is kept.

Request Body

{"enabled": bool}

Response

{"tenant_id": str, "enabled": bool, "overridden": bool, "changed": bool} — enabled is the effective value now, overridden is false when the workspace follows the fleet value, changed is false when nothing had to be written.

Error Codes

• 403 Forbidden: caller is neither the owner nor an admin of this tenant
• 404 Not Found: unknown tenant, or no Tenant CR exists for it
• 409 Conflict: the workspace is scheduled for deletion, or suspended on its Tenant CR (the operator does not reconcile either, so the restart would not happen); or its previous change is still rolling out (one restart at a time — retry once it is back up); or another change to the workspace raced this one — retry
• 502 Bad Gateway: the control plane rejected or could not serve the request
• 503 Service Unavailable: caller is not the owner, and the workspace could not be reached to confirm they are an admin