llm-config
Tier models
When a config isn’t pinned, the smart router picks a tier per request. These flags choose the model for each tier. They work oncreate and update.
Embedding flags
These flags set where a config’s embedding calls go. They work oncreate and update.
On
update, two more flags apply:
Changing or clearing the embedding model moves the config to a different vector space, so embeddings made before the change aren’t comparable with new ones. When that happens,
update shows the old and new embedding model and asks you to confirm. Without a terminal, it refuses unless you pass --yes.