Skip to content

v1.2.0

Choose a tag to compare

@PVLPM PVLPM released this 31 Oct 18:31
· 33 commits to main since this release

Highlights

  • Live peak and off-peak status. On the Quota page, providers whose published pricing has time-based rules now show a compact Peak, Off-peak, or Mixed status with the time of the next change. Open a status to see that provider's models grouped by their exact pricing policy, the upcoming windows or recurring weekly schedule in your local time zone, the current or next public-holiday exception, and whether the period raises, lowers, or leaves the API price unchanged. Z.ai plan schedules are described as quota points, never as a dollar price.
  • Historical peak and off-peak breakdown. A model's detail page now splits each provider route's recorded tokens and API-equivalent cost into Peak and Off-peak rows for the selected range.
  • Temporal pricing controls. Settings has a new Temporal pricing section that lists every provider with time-based pricing. A provider with a discovered account starts on; each switch keeps an explicit on or off choice.

Improvements

  • Estimates follow the price in effect at completion. Recorded steps are priced with the peak or off-peak rate that applied when each step completed, instead of one current rate. Cached estimates are invalidated when the bundled pricing snapshot changes.
  • Pricing data ships with the package. Time-based rules come from a pinned, validated models.dev snapshot inside the package, so status and historical estimates need no runtime network request.
  • Honest pricing labels. A window that changes different token types by different multipliers is described as varying by token type instead of being shown as one multiplier.
  • Dollar values use $. Monetary rows on quota cards render $50.27 instead of 50.27 USD. Other currencies keep their ISO code.

Fixes

  • Settings rows stay aligned. Labels, controls, and eye buttons keep their columns at narrow window widths instead of drifting out of alignment.
  • Hidden cards look hidden. Cards turned off in Settings → Visibility are visibly subdued while editing.

About peak and off-peak status

Peak and off-peak status appears only for providers whose published pricing defines time-based rules. A provider is Mixed when its models are in different states at the same moment, and Unknown when the applicable rule cannot be resolved; the dashboard does not guess a phase. Z.ai plan schedules that adjust quota points are kept separate from API price changes and are never presented as a dollar amount.

Historical peak and off-peak usage uses the completion time recorded for each step and the rule covering that instant. Steps that fall outside a known schedule, or on a route without a supported rule, stay out of the split instead of being assigned to one phase.