EvalCase
A single test case inside an EvalSuite.
Shape: object
| Property | Required | Type | Description | Rules |
|---|---|---|---|---|
id |
no | string (uuid) |
The case identifier. | — |
suiteId |
yes | string (uuid) |
The identifier of the suite this case belongs to. | — |
seq |
yes | integer (int32) |
The sequence number within the suite (starts at 0). | pattern `^-?(?:0\ |
query |
yes | string |
The query text sent to the agent. | — |
expectedOutput |
no | string |
The expected output. Referenced by the containsExpected check. |
— |
expectedTools |
no | array of string |
The tool names looked up by the toolCalled check. An empty list does not mean the check accepts any tool call — the check is still defined separately in the suite’s checks field. |
— |
context |
no | string |
Text given to the model as extra context. | — |
sourceRunId |
no | string (uuid) |
The run the case was generated from. null when hand-written. |
— |
sourceKind |
no | null oneOf EvalCaseSource |
— | — |
promotedAt |
no | string (date-time) |
The promotion time. null when hand-written. |
— |