Voice Agent Test Script

A tool for automated testing of voice agents using predefined test cases executed through phone calls.

Overview

This project provides a script for running automated tests against voice agents. It allows you to define test cases in a CSV file, make real phone calls to test voice agents, evaluate the interactions, and output the results to a CSV file.

Features

• Run multiple test scenarios defined in CSV files
• Execute concurrent calls to test voice agents
• Use OpenAI's GPT models to simulate caller behavior
• Transcribe and analyze call recordings
• Generate detailed test reports as CSV

Requirements

• Node.js
• TypeScript
• Vapi.ai account and API key
• OpenAI API key
• Deepgram API key

Setup

1. Clone this repository
2. Install dependencies:

   npm install
   

3. Set up environment variables in a .env file:

   VAPI_API_KEY=your_vapi_api_key
   VAPI_NUMBER_ID=your_vapi_number_id
   OPENAI_API_KEY=your_openai_api_key
   DEEPGRAM_API_KEY=your_deepgram_api_key
   

Usage

1. Create a CSV file with test cases in the following format (an example my_test.csv is provided in the repo):

   ID,Number,Title,Instruction,NumTest,Test1,Test2,...
   1,+11234567890,Test Title,Test instruction for the voice agent,1,Expected behavior 1,...
   

2. Run the script:

   npx ts-node test-voice-agent.ts my_test.csv
   

3. Review the generated my_test_out.csv file for test results.

CSV Columns

• ID: Unique identifier for the test
• Number: Phone number to call
• Title: Title of the test
• Instruction: Instructions for the assistant making the call
• NumTest: Number of test iterations to run
• Test1, Test2, etc.: Test criteria to evaluate

License

[Your license information here]