RL

Launch the reviewed managed RL recipe

Changed on

Uses ordinary workload admission, billing and tenant ownership. 202 confirms acceptance. Reuse the exact body and Idempotency-Key after an uncertain response. A changed configuration or runtime returns 409 review_changed. After an earlier uncertain submission, keep the original key and check existing runs before starting another attempt. 503 rl_unavailable means new managed launches are paused or the runtime is unqualified. A completed run does not by itself establish improved model quality.

post/v1/rl-runs

Request

  • Base URL
    / (relative, and the document was published on a file host, not at the API’s origin)
  • Auth
    HTTP bearer

Headers

Idempotency-Keystring required

Required stable key per logical submission. Replay the same payload with the same key. Different payload returns 409.

Request body

recipe_id'reasoning-gym-leg-counting'
recipe_version'1.0.0'
mode'evaluate' | 'train'
evaluation_tasksinteger
training_stepsinteger
max_cost_usdnumber

Deprecated compatibility field. Customer spending limits are not enforced.

seedinteger
include_tracesboolean

Opt in to prompt, model response and verifier text in live task events.

review_tokenstring required

Response

Success

idstring required
workload_idstring required
statusstring required
revisioninteger required

Changes

    • info

      the request property default value 5 was removed

    • info

      request property deprecated

    • warning

      the request property's max was set to 100.00

    • warning

      the request property's min was set to 1.00

    • info

      the request property default value 5 was added

    • info

      request property reactivated