Credit Lifecycle
Sleepwalker uses credits for billable actions.
Lifecycle
Section titled “Lifecycle”| State | Meaning |
|---|---|
estimated | Sleepwalker calculated the expected credit cost. |
reserved | Credits were held before queued work started. |
settled | Credits were consumed for completed billable work. |
released | Reserved credits were returned because work did not run or did not complete billable work. |
failed | The billable action failed and no settled charge was recorded for that work. |
AI Visibility
Section titled “AI Visibility”AI Visibility is metered per probe (one prompt on one platform = one probe).
1 probe = that model's credits_per_probe (platform defaults range from 1 to 6 credits)A probe that selects a non-default model costs that model’s own per-probe price, which varies by model. See List visibility models for the exact credits_per_probe, and Create visibility run for how to select one. A run’s reserved credits are the sum of its probes’ individual prices, so a run pinned to a premium model reserves more than one credit per probe.
If a request contains more probes than the account can fund, Sleepwalker queues probes in order until the next probe no longer fits the balance and marks the rest as skipped.
The completed probes remain available. Skipped probes are reported as not run because of insufficient credits.
Probe credits are settled when Sleepwalker returns a usable response or citations. If a probe fails without a usable result, the reserved credit is released.
Cancelling a run (Cancel visibility run) releases the reserved credits of every probe that had not started. Probes already executing finish and settle normally; the remaining release happens when the last one completes.
Content Intelligence
Section titled “Content Intelligence”Content Intelligence runs are metered by analysis depth:
| Depth | Credits |
|---|---|
score | 3 |
full | 6 |
Single actions have their own fixed credit costs:
| Action | Credits |
|---|---|
| Page serialization | 1 |
| Prompt suggestion batch | 1 |
| Trend discovery | 1 |
Content score with trends | 2 |
Content score without trends | 3 |
Synchronous single actions reserve credits before work starts. If the action returns a billable result, the reservation is settled in the same request. If the page is blocked, content is empty, or the action cannot complete billable work, the reservation is released.
Scheduled saved tests use the same reserve, settle, and release lifecycle as manual saved runs. If an account has no usable credits, the scheduled run is skipped before provider or browser work starts.
Failures
Section titled “Failures”Invalid requests that are rejected before work begins do not consume credits.
Actions that cannot start do not consume credits.
When credits are reserved and work cannot be completed, unused credits are released rather than settled.