router (Router and Model Console)

Shared front door: load ../olares-shared/SKILL.md for suite routing, active-profile selection, platform entry points, and the auth proceed/stop gate. Load its auth reference only when login, profile switching, token storage, or auth recovery is actually needed.

Use olares-cli router <verb> --help for authoritative syntax.

When to use

• Give Olares access to a cloud vendor's models, or inspect what it already has.
• Install, configure, or diagnose a model application running on this machine.
• Decide what a caller may put in the model field: an alias, a group of models, or a default category.
• Issue keys and quotas for software that calls models, and read what it spent.
• Send work to a model from the command line: chat, embeddings, reranking, web search and scraping, translation, images, video, speech in either direction, and OCR.

Mental model: every call goes through Router, one AI gateway per Olares. Router holds providers, models, keys, quotas and the usage record; it runs no model itself. A local model runs inside a model application, whose own Model Console downloads the weights, launches the engine and serves the OpenAI-compatible endpoint Router forwards to. So a local model has two states, not one: the platform says whether the container is up, the application says whether the model can answer, and a container reports running minutes before the weights have loaded. Never read one as evidence for the other. Architecture and identity is the long form, and is worth reading before the first write.

All verbs require Olares 1.12.7+ because Router ships as the router Market listing, which asks for that line. Router is an admin-only application: a non-admin profile cannot see its entrance, so every verb here reports it is not installed. Check olares-cli profile whoami for the role and olares-cli market list --mine for the application before concluding anything is missing.

Fast paths