In CloudRadial AutomationAI, the Settings > Model page sets one Default model for your whole workspace. When it is set, your runners use it instead of their own default model for every agent run and every AI Prompt node that doesn't name a model, so you can change the model your automations use without editing each runner's Key Vault. This article covers what the setting overrides, how to set and clear it, how to choose a name your runners accept, and what happens when the name is wrong. Admins can view the setting; only the workspace Owner can change it.
Runners older than version 2.2.0 have no fall-back: agent runs use the Default model as-is, so a name the runner's provider doesn't offer fails them, and AI Prompt nodes ignore the setting. A runner on version 2.0 or later updates itself automatically; for an older runner, see Upgrading a runner to a new version.
- What the Default Model Overrides
- Which Model a Job Uses
- Setting or Clearing the Default Model
- Choosing a Name Your Runner's Provider Accepts
- If the Name Is Wrong
- What the Default Model Does Not Cover
What the Default Model Overrides
Every runner has its own default model, configured in the runner's Key Vault along with its AI provider (see Choosing the runner AI provider). When you set a workspace Default model, AutomationAI sends it to the runner as the requested model for:
- Every agent run that names no model: the AI Playground, Agent nodes in workflows, Pulsars, and agent plays in playbooks
- Every AI Prompt node that names no model, on runner version 2.2.0 or later
It is one value for the whole workspace, used for every run whatever company the run is for. There is no per-company Default model.
Leave it blank and each runner uses its own default, as before.
Which Model a Job Uses
A job uses the first of these that is set:
- The model set on the saved agent, the Agent node, or the AI Prompt node
- The workspace Default model
- The runner's own default model
To use a different model for one job, set the model on that agent, Agent node, or AI Prompt node: that model overrides the workspace Default model for the job.
Note: The Model (optional) fields on agents and nodes still show the placeholder Runner default when a workspace Default model is set. A blank field means the job follows the order above.
Setting or Clearing the Default Model
You must be the workspace Owner to change the Default model.
- Go to Settings > Model.
- In the Default model field, enter the model name. It can be up to 64 characters, with no spaces.
- Select Save default model. The page confirms with "Default model saved."
To clear the setting, empty the Default model field and select Save default model. The page confirms with "Default model cleared." Clearing the field restores each runner's own default from the next run.
If the name is too long or contains a space, the save is refused with the message "The default model must be 64 characters or fewer and contain no spaces." The value you typed stays in the field so you can fix it, and the stored setting is unchanged.
A workspace Admin sees the page read-only, with the note "Read-only — only the owner can change this." A Member sees "You don’t have access to tenant settings. Ask an owner or admin." To learn more about roles, see Understanding workspaces, roles, and permissions.
Choosing a Name Your Runner's Provider Accepts
AutomationAI does not check the name against your runners. The name goes unchanged to the AI provider of whichever runner serves the job, so it must be valid there:
| Runner AI provider | Valid name |
|---|---|
foundry |
A deployment name, for example gpt-5.5 |
openai |
An OpenAI model ID, for example gpt-5.5 |
openrouter |
A namespaced ID, for example openai/gpt-5.5 |
anthropic |
An Anthropic model ID, for example claude-sonnet-5 |
foundry-claude |
A Claude deployment name in your own Foundry account |
If your runners use different providers, a single name may suit only some of them. On the others, set the model on the individual agents or nodes instead, or leave the Default model blank.
If the Name Is Wrong
- Runner 2.2.0 and later — a name the provider doesn't have, including one meant for a different provider, falls back once to that runner's own default. The run's detail shows a notice such as "Requested model → ran on default model (not available on this runner)." So on a runner whose provider doesn't match, the setting has no effect.
- Runners older than 2.2.0 — there is no fall-back. Agent runs use the name as-is, so a wrong name fails them, and AI Prompt nodes ignore the setting.
After you set or change the Default model, run a quick test in the AI Playground and check the run's detail to confirm which model answered.
What the Default Model Does Not Cover
- AI calls inside scripts that don't name a model (
Invoke-NodeAi,Node.InvokeAi,invoke_ai) keep using the runner's own default - The AI Playground's planning calls before a run, the Auto extension choice and the clarifying question, use AutomationAI's own model, not this setting
- The runner's AI provider is not affected; only the model name sent to it changes
If you are still having trouble, we're here to help! Submit a ticket here for assistance, and don't forget to check our status page to ensure there are no outages in your area.
Comments
0 comments
Please sign in to leave a comment.