Generation defaults
Applies on sign-in, Reset and first use of an input tab. Existing drafts and manually selected skills are preserved.
Automatic retry
Applies to new failures across all accounts. Waits 30 seconds, preserves the Image ID and counts each accepted retry against the original account's allowance. The selected model reloads automatically with safety checks; retries never switch models.
The limit excludes the initial request and is not reset by manual Retry. Closed/full queues, quota or storage restrictions delay admission without consuming attempts. Disabling cancels pending automatic retries, not accepted work. Existing failed History is not retried retroactively.
Create user
Accounts
RD accounts remain managed by RD. Accounts created here belong to GPU LAB and survive RD synchronization. Passwords cannot be retrieved after this dialog closes.
Rolling-hour generation allowance
Counts accepted requests across all models, including administrators and later failures. Deleting History does not refund usage. Rejected requests do not count. 0 pauses new submissions; accepted jobs keep running. Maximum 1,000 per hour.
Per-account overrides
Leave an override blank to inherit the system default. Usage renews progressively as requests become one hour old. Refresh usage before inspecting recent activity.
Plans & shared credit pools
Every account has one rolling seven-day pool shared by the website, all API keys and all models. Each accepted attempt, including Retry and failures, costs one credit. Rejections cost nothing; deleting History or changing Plans does not reset usage. Hourly limits also apply. Starter is the default Plan.
Tab structure & style
Reorder the fixed sections, choose the opening tab and adjust this settings page. All sections stay available; custom HTML/CSS is not accepted. Saved preferences apply to all administrators.
Configuration change log
Who changed what, when, with before/after values. Retained for 90 days; this is an operational log, not immutable compliance storage.