Choose models for Subagents

A subagent is a helper agent DataMind Oda starts while it works on a task. By default, a subagent runs on the model of the chat that started it; model selection lets Oda choose a different model and reasoning effort for a subagent, from a list of routes you allow. See Subagents for what subagents are and how Oda starts them.

Turn model selection on

1. Open the Subagent settings page (under Plugins in Settings).
2. Turn model selection on and allow the models a subagent may use. An enabled setting must name at least one model.

Model selection is off by default, and the list of allowed models starts empty.

Which chats the choice applies to

• A chat decides when it starts. A chat reads the setting as it starts; when the setting is on, the allowed models are recorded for the chat before its first request, and the chat works from that record from then on.
• Later edits do not change chats that already exist. Changing the setting affects only chats that start afterwards.
• A chat restored from earlier keeps its past decision. A chat restored without a recorded decision stays without model selection, even when the setting is on now.
• A subagent takes the decision of the chat that started it.

How the model and effort are chosen

With model selection on, Oda may choose a model and reasoning effort for a subagent from the allowed list, instead of taking the chat's model:

• Every choice must resolve to an allowed route. The allowed list names routes, each a provider and a model. A provider outside the list is refused, and it never appears among the routes Oda can look up; a route outside the list is refused with child LLM route "<provider>/<model>" is not allowed for this Session before any subagent starts.
• Oda can look up the routes it may use. Its model-listing tool lists the providers, their models and the reasoning efforts a model offers, always trimmed to what the chat allows; a lookup changes nothing about the chat. A model named without its provider is refused (`model` requires `provider`), and a provider named as a blank string is refused (`provider` must be non-empty).
• A subagent starts on the chat's model. When Oda chooses a different model for it, the inherited reasoning effort is cleared unless an effort is chosen together with the model, so the new model resolves its own default.
• Some choices are refused outright. A choice that names a provider without a model, or a model without a provider, is refused with child LLM `provider` and `model` must be supplied together; where model selection is off, a named choice is refused with child model selection is disabled for this tool instance.

Choosing the model for the chat itself is covered in Chat with Oda.

Questions, answered

Where do I turn model selection on?

In the Subagent settings page: turn model selection on and allow the models a subagent may use. It is off by default, and an enabled setting must name at least one model. Subagent model selection covers the controls.

Which models may a subagent use?

The models on the allowed list recorded for its chat. Oda can look up the routes it may use, and the lookup is trimmed to what the chat allows; Oda chooses only from those routes.

Does turning it on change my existing chats?

No. Each chat keeps the decision it started with: the setting is read when a chat starts, and later edits affect only chats that start afterwards. A chat restored without a recorded decision stays without model selection.

Do subagents inherit the choice?

Yes. A subagent takes the decision recorded for the chat that started it.

Why was a model choice refused?

Model selection may be off (the choice is refused with child model selection is disabled for this tool instance), the route may be outside the chat's allowed list (child LLM route "<provider>/<model>" is not allowed for this Session), or the choice may name only half of the route (child LLM `provider` and `model` must be supplied together).

Why did the reasoning effort reset when the model changed?

When Oda chooses a different model for a subagent and no effort is chosen together with it, the inherited reasoning effort is cleared, so the new model resolves its own default; a specific effort is kept only when it is chosen together with the model.

What if the lookup finds nothing?

The lookup says which case it is:

• (no LLM providers): no provider is inside the chat's allowed list.
• (no advertised models for <provider>): that provider advertises no models.
• (no advertised reasoning efforts): that model offers no reasoning efforts.

Does the lookup change anything about the chat?

No. The lookup only lists the routes a subagent may use; it changes nothing about the chat.