Skip to main content
GET
Get test case result

Authorizations

X-API-Key
string
header
required

API key for authentication

Path Parameters

task_id
string
required

Test run or benchmark the case was run in

Example:

"a3b2c1d0-e5f4-3210-abcd-ef1234567890"

test_uuid
string
required

The test whose result to read, as test_uuid on the case

Example:

"b1c2d3e4-f5a6-7890-bcde-f12345678901"

Query Parameters

model
string | null

Which model's answer to read. Required for a benchmark, which runs every test once per model

Example:

"openai/gpt-4.1"

Response

Successful Response

test_case_id
string | null

ID of the test case within the run

name
string | null

Name of the test

test_uuid
string | null

ID of the test this case ran, which is what you pass to read the case on its own

Example:

"b1c2d3e4-f5a6-7890-bcde-f12345678901"

test_type
enum<string> | null

What the test asks of the agent, which decides how a reader draws the case

Available options:
response,
tool_call,
conversation,
general
passed
boolean | null

Whether the case passed

reasoning
string | null

The judge's reasoning, or the tool-call diff for a tool-call test

output
TestOutput · object | null

The agent's output for this case

test_case
object | null

The test case definition that was run

inputs
object | null

Extra request fields sent to the agent for this case, the agent's default_inputs with this case's overrides applied

judge_results
JudgeResult · object[] | null

One verdict for each evaluator

latency_ms
number | null

How long the agent took to respond, in milliseconds

cost
number | null

Cost of this case (USD)

unanswered
boolean
default:false

Whether this case produced no answer because the agent or the judge could not be reached, in which case reasoning carries the error and passed is not a verdict on the agent

not_run
boolean
default:false

Whether this case never started, because a user stopped the run first. It is counted neither as passed nor as failed