Model Resolution

When you specify a model name (e.g., in prompt file), promptcmd tries to resolve it to a specific configuration defined in your config.toml.

Resolution Strategy

Resolution takes place in this order:

1. Parse as provider/model: If in provider/model-name format, extract the provider and model name, otherwise:
2. Check for variant: Is there a variant with this name? otherwise:
3. Check for group: Is there a load balancing group with this name?

If the resolution result is a group, each group member is resolved by name again, according to the above strategy (excluding step 3, since there cannot be nested groups).

INFO

During execution, if resolution results in a group configuration, the Load Balancer decides which group member to use.

Test resolution

Use the following command to see what a specified name resolves to:

promptctl resolve openai
promptctl resolve anthropic/claude-sonnet-4
promptctl resolve group_name
promptctl resolve variant_name