Agents

An Agent in Evalion represents the AI system being tested and evaluated. Agents serve as the central entity that responds to simulated user interactions during test runs, acting as both the subject under evaluation and the conversational partner for personas in your test scenarios.

Evalion supports both Voice AI Agents and Text AI Agents, allowing you to test conversational systems across different communication channels.

Agent Configuration

Each agent requires specific setup parameters to establish proper testing connections and behavior:

1. Simulation Type

Agents are configured with specific simulation modes that determine how tests are conducted:

Inbound Voice: Agent receives incoming calls from simulated user personas, mimicking scenarios where customers initiate contact (e.g., calling a support hotline or booking service) .
Outbound Voice: Agent initiates calls to simulated user personas, testing scenarios where your system reaches out proactively (e.g., appointment reminders, follow-up calls) .

2. Agent Name

The Agent Name is a crucial element in your testing environment, providing a clear and organized way to distinguish between multiple voice or text AI agents.

3. Connection Method

Agents are configured with specific connection methods that determine how your AI agent is accessed during testing:

Phone: Agent connection is established via your voice agent's contact number, enabling direct phone-based testing.
API: An agent connection is established via an accessible API endpoint, enabling seamless integration for text or voice-over-IP testing.

When using the API connection, configure the following parameters:

Field	Description
API Endpoint	The endpoint URL where the API request will be sent
Headers (JSON)	Headers to include in the API request (JSON format)
Payload (JSON)	Request payload to send to the API (JSON format)

4. Language

This ensures your agent's preferred language of communication, ensuring that test scenarios and personas interact in the appropriate language context for accurate evaluation.

5. Agent Prompt

A comprehensive description of your agent's purpose, behavior guidelines, and operational instructions that serves as the foundational reference for how the agent should interact with users and handle various scenarios.

The agent prompt plays a key role in generating more accurate test results, providing context for expected behavior, helping identify when your agent deviates from intended functionality, and ultimately boosting your confidence in the effectiveness of your testing process.

Agent Role in Testing

Agents form the foundation of all testing activities in Evalion, serving as the central component that interacts with personas within defined scenarios to generate measurable test results. During test execution, agents are evaluated against custom metrics and scenarios to assess their conversational performance, technical reliability, and user experience quality.