Routing

Control how requests are routed: strategy, policy, model aliases, and the decision behind every request.

📄️Speed options

Artemis Task's speed setting, as the active capability version measured it in-fold on its training prompts: per preset (`quality_first`, `balanced`, `cost_first`) and speed (`off`, `faster`, `fastest`), the share of the best single model's quality, dollars per 100k requests, and the answers' median and 90th-percentile latency. Not a guarantee on your traffic. `available` is false when the active version has no speed table.

📄️Get routing config

Your organization's routing mode, strategy, weights, speed and provider policy.

📄️Update routing config

Replaces the routing configuration: a field you leave out is reset to its default, so send back the whole config from `GET /v1/routing/config` with your changes. Two are kept when left out: `speed` (send `null` for off) and the model aliases (manage them with the alias endpoints).

📄️List routing decisions

List routing decisions

📄️Get a routing decision

Why a request went where it did: which router decided, every candidate considered, its score, and what was excluded.

📄️List model aliases

List model aliases

📄️Create or update an alias

Maps an alias (lowercase letters, digits, `-` and `_`) to a model ID, so requests can use `model: '<alias>'`.

📄️Preview a routing policy

Which of your providers would be eligible under the given policy fields. Anything omitted falls back to your saved routing config. Under Artemis Task the cost and latency caps exclude no provider here: Artemis Task checks them per request, on each model's expected cost and time to answer, and `perRequest` lists them. `projects` shows the same policy for each of your projects that narrows your regions or rules out a provider in its project policy (the reason names the project's rule).

📄️Delete a model alias

Delete a model alias