Skip to main content
POST
Check judge generation settings against the model
You are reading the docs of Galtea 5.12. Read this page in the latest version.

Authorizations

Authorization
string
header
required

API key authorization. Pass your API key in the Authorization header as a Bearer token. Both new (gsk_*) and legacy (gsk-) API keys are accepted, e.g. Authorization: Bearer gsk_... or Authorization: Bearer gsk-....

Body

application/json
evaluatorModelName
string
required

The judge model to send the probe to. Must be a model a metric may select.

Minimum string length: 1
Example:

"gpt-5-nano"

judgeGenerationSettings
object | null

The settings to try. Omit a setting to leave the provider default in place.

Response

The probe ran. outcome says whether the provider accepted the settings, refused one, or never answered.

outcome
enum<string>
required

What the call established. accepted means the provider took the values; a setting a model does not support is discarded silently and still counts as accepted, so this never reports that a value changed the judge behaviour. refused means the provider was reached and rejected a value. unreachable means the provider never answered, so nothing was learned about the values and the call is worth repeating.

Available options:
accepted,
refused,
unreachable
Example:

"accepted"

error
string | null
required

The provider's own refusal message, naming the parameter and its limit. Set only on refused: an unreachable outcome carries no message, because the failure is ours and says nothing about the values that were sent.

Example:

"temperature: range: 0..1"