Run AI-Generated Scenarios
Start an AI-generated simulation run.
Tuner generates new test scenarios for your agent, a mix of functional and adversarial,
places a call for each, and grades them with the agent’s PASS_FAIL evals. Scenarios are
generated asynchronously, so poll List Runs for progress. To re-test scenarios you’ve
already saved, use Run Saved Scenarios.
Authorizations
Use Tuner API key (tr_api_...) or user session token. Find your API key in Workspace Settings > API Keys.
Path Parameters
Tuner's internal numeric ID for the agent.
Workspace ID. Find this in Workspace > General Settings.
Body
Request to start a generated simulation run.
Tuner generates functional_count + adversarial_count test scenarios (5 to 20 in
total), places one simulated call per scenario against the agent, and grades the calls
with the agent's PASS_FAIL evals.
ID of the caller persona (language and accent) the simulated caller uses. Get valid IDs from List Caller Personas.
1
Number of functional (routine) scenarios to generate.
x >= 04
Number of adversarial (pressure) scenarios to generate. functional_count + adversarial_count must be between 5 and 20.
x >= 02
Maximum length of each simulated call, in minutes (1-12). Defaults to 3.
1 <= x <= 123
ID of a simulation profile belonging to this agent. Its fields are given to the simulated caller as facts about themselves. Optional.
x > 012
ID of a background-noise environment to mix into the call. Get valid IDs from List Noise Environments. Must be sent together with noise_intensity.
x > 02
Gain applied to the background noise: 0.5, 1.0 or 1.5. Must be sent together with noise_environment_id.
1
IDs of PASS_FAIL evals to grade the calls with. Omit or send null to use all of the agent's evals. If provided, it must contain at least one ID.
IDs of call intents to detect on the calls. Omit or send null to use all of the agent's intents. If provided, it must contain at least one ID.
Response
Successful Response
The simulation run that was just started.
ID of the new simulation run.
Number of test scenarios created so far. 0 for Run AI-Generated Scenarios, because generation is asynchronous; the full count for replay runs and runs from saved scenarios.
Whether the run was accepted.
ID of the caller persona used. Null for replay runs.
The caller persona used. Null for replay runs.
ID of the simulation profile used, if one was given.
Name of the simulation profile used, if one was given.
Copy of the profile's fields as they were when the run started.
ID of the background-noise environment used, if one was given.
Gain applied to the background noise, if noise was used.
The background-noise environment used, if one was given.