RedRavenApi reference

Get next attack prompt

Return the next user prompt and prior `messages` array, or `done` true when the run is finished. Returns 409 while an attack eval is still pending.

GET
/tests/{test_id}/attack-run/next

Return the next user prompt and prior messages array, or done true when the run is finished. Returns 409 while an attack eval is still pending.

Authorization

ApiKeyAuth OrganizationId
X-API-Key<token>

Your RedRaven API key.

In: header

X-Organization-Id<token>

RedRaven organization UUID that scopes the API key.

In: header

Path Parameters

test_id*string

RedRaven test UUID.

Formatuuid

Response Body

application/json

curl -X GET "https://example.com/tests/497f6eca-6276-4993-bfeb-53cbbbba6f08/attack-run/next"
{  "done": true,  "run_id": "string",  "status": "string",  "summary": {    "scenarios_total": 0,    "attempts_total": 0,    "attempts_failed": 0,    "asr": 0  },  "attempt_id": "string",  "conversation_id": "string",  "attempt_index": 0,  "turn_index": 0,  "phase": "warm_up",  "messages": [    {      "role": "string",      "content": "string"    }  ],  "prompt": "string",  "scenario_id": "string",  "technique_id": "string",  "attacker": {}}
Empty

Begin an attack run POST

Start an attack run from enabled scenarios and selected techniques. Credits are charged here: `sum(maximum_multi_turn) × techniques × attempts`

Submit an attack-run turn POST

Upload the target request, response, and full `messages[]` for the pending turn. Warm-up turns return `decision: continue`. Attack turns return `decision: pending`; poll eval-status before calling next.