You choose the route and may change it.
TEMRAL records the selected provider and model inside each governed run.
How TEMRAL works
Point the same SDK at the TEMRAL base URL. Your provider choice, models, prompts, streaming behavior, and rollback path stay under your control. Each admitted result returns with measured savings and a cryptographic receipt.
POST /v1/chat/completionsKeep the same SDK, payload shape, models, provider keys, and streaming contract.
Your application sends the request through the TEMRAL endpoint without a workflow rewrite.
Policy and measured workload fit determine whether the request enters the agreed reduction envelope.
The selected provider and model remain attributable while TEMRAL preserves the application contract.
The outcome returns with measured cost, scope, integrity, and recovery custody in one record.
What gets measured
A percentage never upgrades the proof standard. It changes only the admitted reduction target for a workload that fits.
The agreed comparison basis is bound before a savings claim exists.
The selected tier applies only where the workload fits its admitted envelope.
The returned result must clear the workload acceptance contract.
Measured savings, scope, provider route, and integrity close together.
Evaluation binds the baseline, acceptance contract, tier, and scope before anything can become a savings claim.
Bring one workload