Skip to main content

​
What is a Phone Connection?

A Phone Connection tells Galtea how to reach your AI agent over the phone. Instead of calling an HTTP API, Galtea dials your agent’s phone number and runs the conversation as a real voice call — this is Direct Inference over the phone. You configure the phone number once, and Galtea uses it to drive voice evaluations against your telephony agent automatically. The simulated caller’s language is taken from each test case (see below), so one connection serves test cases in different languages.

​
Use Cases

  • Evaluate voice agents — Galtea calls your agent’s phone number directly for each test case, exercising the full telephony stack
  • Multi-turn conversation evaluation — For Behavior test cases, Galtea drives a full conversation using the test case’s persona, goal, and scenario, with the conversation simulator generating each user turn.
  • Single-turn evaluation — For Accuracy and Security test cases, Galtea speaks the test case’s input line into the call and evaluates the agent’s reply — the voice equivalent of a single-turn HTTP run.
  • Standardize call configuration — Define a reusable phone number and call protocol for repeatable voice evaluations
  • Manage multiple agents — Create separate phone connections for different agents or environments

​
Creating a Phone Connection

To create a Phone Connection:
  1. Navigate to your product in the Galtea Dashboard
  2. Go to the “Phone Connections” section
  3. Click “New Phone Connection”
  4. Configure the connection properties as described below

​
Phone Connection Properties

​
Text
required
A unique name for the phone connection within the product. Example: “Production Voice Agent” or “Staging IVR”.
​
Text
required
The phone number Galtea dials to reach your AI agent, in E.164 format (a leading + followed by the country code and number). Example: +14155552671.
​
Checkbox
Controls whether the voice agent opens the conversation once the call connects. Leave it unchecked (the default) when the simulated caller should speak first, which suits agents that wait for a greeting from the user. Check it for interviewer-style agents that greet the caller first.

​
How the caller’s language is chosen

The simulated caller’s spoken language is not set on the phone connection — it comes from each test case’s language. Because a phone connection is attached to a Version and a version runs many test cases, this lets one connection cover test cases in different languages. With a multilingual text-to-speech provider, the test case’s language drives the synthesis language and the voice is selected automatically (e.g. a Spanish test case → the caller speaks Spanish with a natively-accented voice). When a test case has no language set, the provider falls back to its configured default voice.

​
Adding background noise to a call

Like language, background noise is not set on the phone connection — it comes from each test or test case. Two optional fields control it:
  • Background Noise Profile — the kind of noise: clip-backed (office, street, car) or synthetic (white, pink). Leave unset for a clean recording.
  • Background Noise Level — how loud the noise is relative to the caller’s speech: light, medium, or heavy.
Both must be set together; setting only one is rejected. Galtea mixes the noise into the simulated caller’s audio during the call, so it tests how your agent behaves under noisy conditions (e.g. a caller phoning from a busy street). The saved call recording contains the noise, so what you hear on playback in the dashboard is what your agent heard. A test carries a default background noise setting, and each test case can override it. A test case created without its own setting inherits the test’s setting at creation time; setting a test case’s noise explicitly (including clearing it) overrides that default for that case alone.
Background noise only applies to voice tests run over a phone connection. It has no effect on tests run over an Endpoint Connection.

​
Relation to Versions

A Phone Connection becomes the conversation target of a Version: when a version references a phone connection, Galtea dials that number to run each evaluation. A version’s conversation target is mutually exclusive — a version connects to your AI product either through an Endpoint Connection or through a Phone Connection, never both. Choose an endpoint connection when Galtea should call your HTTP API, or a phone connection when Galtea should place a voice call.

​
Relation to Voice Evaluation

Phone Connections power voice evaluation: because Galtea drives a real phone call, the conversation, transcription, and resulting inferences reflect how your agent behaves over the telephony channel. This makes phone connections the entry point for evaluating voice agents end-to-end. Phone connections run any test type — no additional configuration is required beyond attaching the phone connection to the version:
  • Behavior test cases run multi-turn: the conversation simulator drives each user turn from the test case’s persona, goal, and scenario.
  • Accuracy and Security test cases run single-turn: Galtea speaks the test case’s input line into the call and evaluates the agent’s reply.
For single-turn (Accuracy/Security) tests scored by deterministic metrics, leave Agent Speaks First unchecked (the default). When the agent speaks first, the call has two turns (the agent’s greeting plus the scripted line), which deterministic metrics — designed for a single turn — cannot score correctly. Multi-turn (Behavior) tests are unaffected.
The API automatically transcribes the agent’s audio responses using speech-to-text, so no manual transcript preparation is required. Each recorded agent turn is transcribed before evaluation runs, and the written transcript is what metrics score — your agent’s voice output is evaluated as accurately as any text-based agent. Both sides of every turn are recorded and stored on the InferenceResult for that turn:
  • Agent audio (what your agent said): stored as a content-parts envelope on actualOutput; access the envelope via the SDK’s .actual_output_data field and the scored transcript via .actual_output.
  • Simulated user audio (what the synthetic caller said): stored as a content-parts envelope on input; access the envelope via the SDK’s .input_data field and the spoken transcript via .input.
This means the full phone conversation — both the synthetic user side and the agent side — can be played back turn by turn in the dashboard’s audio player and read programmatically via the SDK. See InferenceResult for the envelope shapes.

​
Best Practices

Choose names that clearly identify the agent’s purpose and environment, such as “Production Voice Agent” or “Staging IVR”.
Provide the phone number with a leading + and country code (e.g., +14155552671). Numbers in other formats are rejected.
Set each test case’s language to the language your agent operates in. With a multilingual text-to-speech provider the simulated caller then speaks that language with a natively-accented voice (e.g. Spanish), with no per-connection configuration needed.
Set the background noise profile and level on the test so every generated or CSV-uploaded test case inherits realistic ambient noise automatically. Override it on individual test cases only when you need a specific case to be quieter, louder, or silent.
If your agent greets the caller (interviewer-style), check Agent Speaks First. Leave it unchecked when your agent waits for the user to initiate the conversation.

Concepts overview

How Galtea’s concepts connect — diagram + per-entity quick reference.

Endpoint Connection

Connect to your AI product over an HTTP API instead of the phone.

Phone Connection Service SDK

Manage phone connections programmatically using the Python SDK.

Version

Attach a phone connection to a version for voice evaluation.