Skip to main content
POST
Python SDK

Authorizations

Authorization
string
header
required

Enter your API key (e.g. dai_sk_test_key_001)

Headers

Authorization
string

Path Parameters

task_id
string
required

Cookies

decimal_session
string | null

Body

application/json

Typed body for POST /replay/tasks/{id}/submit.

Every field is validated here rather than at the database, so a bad value comes back as a 422 naming the field instead of a 500 from the column type.

The bounds are not cosmetic. eval_score must be 0.0–1.0 because the preference-pair export downstream splits pairs on a score threshold: a score outside that range would put every pair on the same side of the split and silently skew the exported training data. eval_verdict is pass/fail/review — the same vocabulary score_replay_result emits and the batch passed/failed counters count — and status is the replay-task lifecycle set. extra="forbid" matches the decision payload: a mistyped field name should 422 rather than quietly mean "no value supplied".

replayed_trace_id
string | null

ID of the trace produced by re-running the prompt.

eval_score
number | null

Quality score for the replayed trace (0.0–1.0).

Required range: 0 <= x <= 1
eval_verdict
enum<string> | null

Verdict for the replayed trace.

Available options:
pass,
fail,
review
status
enum<string>
default:completed

Lifecycle state to record on the task.

Available options:
pending,
running,
completed,
failed,
skipped

Response

Successful Response