Model service configuration¶
Use provider-backend to manage model-supply Backends, Endpoints, and Routers. These commands configure model services; they do not invoke Genesis models.
moi-cli -o provider-backend list --kind llm
moi-cli provider-backend --help
Configure model services and routing¶
Task |
Commands |
Notes |
|---|---|---|
Inspect or manage a Backend |
|
Confirm the service address, model configuration, and impact before a write. |
Manage Endpoints |
|
Use the Backend identifier for later operations. |
Inspect or update a Router |
|
A Router change can affect later model-service selection. |
Probe Backend models |
|
Record non-sensitive results only. |
Choose llm, embedding, or parser with --kind. Before creating, updating, or deleting a Backend or changing a Router, confirm the service address, model configuration, secret source, and impact. Do not expose secrets in command lines or logs; record only non-sensitive results when probing Backend models.
Discover models and set up an LLM Backend¶
The model command group lists models available to the selected workspace and can set up an LLM Backend in one operation. It complements the fine-grained Backend, Endpoint, and Router management in provider-backend.
moi-cli -o model list
moi-cli -o model list-embedding-models
Setting up an LLM Backend requires a Backend name, model-service address, and comma-separated model list:
moi-cli model setup-llm-backend \
-name "<backend-name>" \
-address "<model-service-address>" \
-models "<model-id-1>,<model-id-2>"
Use probe-llm-models or probe-embedding-models to discover candidate models. A probe result does not mean that a Backend has been created or enabled; confirm the secret source and permission scope before writing service addresses or model configuration.