API reference
Savings
Dashboard
BetaWe just launched Apps: hosted models, actions and pages, on the same bill as your inference.Read the Apps docs

Savings

GEThttps://api.inferencebureau.com/v1/savings

What mixed-model sessions saved your team over a period: what opted-in turns cost, against what they would have cost on the model each asked for, net of the classifier.

Query

  • from, toYYYY-MM-DDoptional

    UTC days, both included, at most 366 apart. Defaults to the first of this month through today.

  • projectstringoptional

    One project, by id. Defaults to the whole team.

Response

  • totalobjectrequired

    baseline_usd is the requested models priced at the tokens actually used; served_usd and classifier_usd are settled charges; net_saved_usd is the baseline minus both, and can be negative.

  • declined_estimate_usdstringrequired

    Estimated savings of suggestions your app declined.

  • by_task, by_projectarrayrequired

    The same figures per task class and per project, largest saving first.

  • estimatebooleanrequired

    Always true: the baseline is an estimate. Charges are exact.

  • truncatedbooleanrequired

    True when a day had more classified turns than the report reads at once, so totals are a lower bound.

Notes

  • Freshnessoptional

    Days before today are stored each night and corrected for three nights as late charges settle. Today is computed when you ask.