Create Replay Run
Start a deterministic replay run against one or more voice clip sets.
A replay plays each selected clip set’s recorded caller turns at the agent, turn by
turn, so the caller side is byte-identical between runs — a metric delta is
attributable to a pipeline change rather than conversational variance. It places one
call per (clip set, repetition), for len(voice_clip_set_ids) * repeat_count calls
total, and is inbound-only.
selected_eval_ids of null or [] means no evals — a latency-only run. This
deliberately differs from Create Run, where null means “all evals”: a benchmark run’s
headline metric is latency, and silently scoring every eval would misrepresent what was
measured. Pass explicit eval IDs to grade the replayed calls.
Unlike Create Run, scenarios are written synchronously (no LLM involved — each
scenario is a snapshot of one clip set’s clips) and test calls are queued immediately,
so generated_count reflects the true total on the 202 response, not zero. Poll List
Runs for progress.
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 launch a deterministic replay run against one or more voice clip sets.
1 - 10 elements1 <= x <= 201 <= x <= 12User-supplied overrides for a replay's timing policy.
Every field is optional; an unset field falls back to the corresponding default on
app.services.simulation_clip_replay.ReplaySettings -- the single source of truth
for these defaults, deliberately not duplicated here.
Response
Successful Response
One selectable background environment.