Glossary - Tuner Documentation

The terms used across Tuner, in alphabetical order. Each links to the page that covers it.

Term	Definition
Agent	The voice AI application you monitor with Tuner, such as a customer support bot, appointment scheduler, or sales assistant. Each agent has its own evals and settings.
Agent Remote ID	The identifier that links an agent in Tuner to the calls you send for it, sent as `agent_remote_identifier`. For Retell, Vapi and ElevenLabs it’s the agent’s ID on that platform; for SDK and API agents Tuner generates it. Shown on Agent Settings > Agent Setup. See API keys and IDs.
Alert	A notification sent when calls matching the conditions you set reach a threshold in a time window, for example more than 5 red-flagged calls in an hour. An alert can combine several conditions. Delivered by email or webhook. See Alerts.
API key	A workspace key starting with `tr_api_` that SDKs, the public API and MCP clients send as `Authorization: Bearer <key>`. Created on API Keys and shown once. See API keys and IDs.
Call Logs	The page that lists every call for the selected agent, with filters and search. Open a call to see its transcript, evals, voice metrics, extracted data and traces. See Call Logs.
Call Outcome	How a call ended, defined by you as a success or a failure. Add outcomes from Tuner’s outcome library or create your own. Call outcomes drive your success rate. See Call outcomes and intents.
Call Simulation	Real SIP test calls between your agent and a Simulated Caller: Tuner dials in for inbound agents, or your agent dials out for outbound agents. Each simulated call is scored against the same evals, intents and call outcomes configured for production. See Call simulation.
Caller Turn	One stretch of the caller speaking in a Dataset call, with its start and end time in the original recording. A replay plays a call’s caller turns in order, waiting for your agent to answer after each one.
Correlation ID	The `X-Correlation-Id` header Tuner adds to each inbound simulated call. Your integration sends it back with the call so Tuner can link the call to its run. See How simulated calls are linked.
Dataset	Your agent’s collection of Dataset calls, built on the Dataset tab under Simulations. Replay it to reproduce a failure, prove a fix, or compare setups on your own calls. See Datasets.
Dataset Call	The caller’s side of one real call, cut into caller turns, ready to replay against your agent. You add one from a production call or an uploaded stereo recording. See Add Dataset calls.
Eval	A check of whether your agent did something specific or followed a guideline during a call, for example confirming appointment details before ending the call. Use one from the predefined eval library or write your own. Evals run automatically on every call. See Evals and guardrails.
Guardrail	A rule that defines what “bad” looks like for your agent, such as a red-flag rule or an alert. See Evals and guardrails.
Hallucination	When the agent gives incorrect, made-up, or misleading information, such as a wrong price or a booking that didn’t happen. The predefined Hallucination eval checks for it.
Intent	What the caller wanted from the call. For an appointment scheduling agent, intents might include “Book Appointment,” “Reschedule,” “Cancel,” or “Check Availability.” Set up on Agent Settings > Analysis Setup. See Call outcomes and intents.
Label	A tag applied to a call when a label rule matches, used to categorize calls without marking them as a problem. See Red flags and labels.
Overview	The dashboard for the selected agent: Total Calls, Success Rate, Failed Calls, Calls with Red Flags and Red Flag Rate, the Call Intent Breakdown and Call Outcome Breakdown charts, and recent red flags. See Overview dashboard.
Pressure Test	A Simulated Caller that stress-tests your agent with difficult situations, such as unhappy customers, scope-boundary requests, or attempts to bypass your process. Pressure Tests are generated from your configured evals. See Call simulation.
Red Flag	A tag applied to a call when a red-flag rule matches, for example when an eval fails. Red flags appear on the Overview dashboard and in Call Logs, so you can see which calls need attention first. See Red flags and labels.
Reference Transcript	What the caller said in each caller turn of a Dataset call, taken from the source call’s transcript or transcribed by Tuner. You can edit it and mark it Verified. A replay always plays the audio, and uses the transcript to score transcription accuracy.
Replay	A simulated call that plays the caller turns of a Dataset call into a live call with your agent, so the caller is identical on every run. Every replay is graded against your agent’s Pass/Fail evals. See Replays.
Replay Run	A group of replays started together from New Replay: up to 10 Dataset calls, each replayed 1 to 20 times, with at most 20 calls in total.
Routine Caller	A Simulated Caller that follows your agent’s normal, cooperative workflows. Routine Callers are generated from your agent’s configured intents to test the happy path. See Call simulation.
Rule	The conditions that apply a label or red flag to a call, set on Agent Settings > Evaluation Rules. A rule has one or more conditions on evals, voice metrics, data fields or call metadata, and matches when all or any of them match. See Red flags and labels.
Runs	The tab under Simulations that lists your simulation runs as cards. Open a run to see its calls with #, Type, Evals, Outcome, Intent and Dur. columns, and click a call for its full details. See Run statuses.
Scenario	One situation for a Simulated Caller to play: the caller’s persona, intent and what to test. Tuner writes new Scenarios for each simulation run. Scenarios in the Scenario Library show their origin: Manual for ones you wrote, Saved from run for ones kept from a run. See Scenario Library.
Sentiment	The emotional tone of the caller in the conversation, shown on each call’s Overview tab.
Simulated Caller	The AI caller that talks to your agent on a simulated call. Each one is a Routine Caller or a Pressure Test. See Call simulation.
Simulation Mix	The ratio of Routine Callers to Pressure Tests in a simulation run. You set it with a slider in the Run Simulation dialog.
Simulation Run	A group of simulated calls started together from the Run Simulation dialog: 5 to 20 calls, 10 by default. Runs appear on the Runs tab.
Simulation Setup	The Agent Settings > Simulation Setup tab, where you set how Tuner reaches your agent for simulation. Inbound agents use the SIP Settings card for the SIP URI and credentials Tuner dials. Outbound agents use the Outbound Trigger Endpoint card. See Inbound simulation and Outbound simulation.
SIP	Session Initiation Protocol, the signaling standard voice platforms use to start, manage, and end calls. Tuner uses SIP so simulation calls travel the same path as a real phone call. See Inbound simulation.
SIP Call-ID	A unique identifier carried with each SIP call. Tuner can use it to link a simulated call to its run. See How simulated calls are linked.
SIP URI	A SIP address, the internet equivalent of a phone number, for example `sip:agent@sip.vapi.ai`. In inbound simulation, Tuner dials your agent’s SIP URI; in outbound simulation, Tuner gives your system a SIP URI to dial. See Inbound simulation and Outbound simulation.
Transcript	The written record of the conversation between the caller and the agent, turn by turn, with speaker labels and timing.
Transcription accuracy	On a replay call, the share of the caller’s words your agent’s speech-to-text heard correctly, with a Good, Fair or Poor rating. See Transcription accuracy.
Turn-taking Settings	The Advanced turn-taking values a replay uses to decide when your agent has finished speaking and when to play the next caller turn: initial agent wait, silence threshold, minimum speech, inter-turn delay, and max wait. See Replays.
Voice metrics	Measurements Tuner computes for every call, such as per-turn latency for each pipeline stage and the longest agent monologue. See Voice metrics.
Workspace ID	The identifier of your workspace, sent as `workspace_id` in API calls. Shown on Workspace Settings > General Settings. See API keys and IDs.

Next steps

Evals and guardrails

How evals and guardrails judge your agent’s calls.

Call simulation

Run simulations to test your agent.