← Changelog
NewDirectives

Aliases that pick their model: routing profiles

Define a routing profile in Directives, a pool of models with a baseline, and point an alias at it. Each request then runs on the cheapest pool model, or the best scoring one, that it and your Directives allow, above an optional quality floor. No classifier runs and nothing extra is billed.

A profile's objective is cost (estimated from list prices for the request) or quality (evaluated scores for its task class, the request's ib.task, or the mean across classes). Every pool model is checked as if the request named it: allowed and denied lists, zero data retention and no training, the tools, structured output and streaming it uses, its context window and the per-request charge ceiling.

The chosen model runs with provider failover, then the rest of the ranking and the baseline as fallback models. When no pool model qualifies the request runs on the baseline, and Savings counts profile requests against the baseline, in a row per profile.

Receipts and GET /v1/requests/:id carry routing_mode profile and a selection listing each pool model and why it was chosen or left out; the request's page in Logs shows it too. Edit profiles in the new Routing profiles group of the Directives table, or with profiles.<id> through the management API.

More in Routing profiles.