MakingChatbots / ivr-tester

An automated testing framework for Interactive Voice Response (IVR) call flows

IVR Tester

Automate the testing of your IVR call flows!

Follow me on LinkedIn to learn about updates to this project

IVR Tester automates the testing of IVR flows by calling them, interpreting prompts and replying with DTMF tones based on fluent test definitions.

Features:

• Fully automates testing call flows
• Test multiple scenarios in parallel
• Expressive test definitions help document call flow
• Record audio of tests
• Record transcriptions of tests
• Supports Google Speech-to-Text and AWS Transcript for transcribing calls
• Open-source

const config = { transcriber: googleSpeechToText({ languageCode: "en-GB" }) };

new IvrTester(config).run(
  { from: "0123 456 789", to: "0123 123 123" },
  {
    name: "Customer is provided a menu after their account number confirmed",
    steps: [
      {
        whenPrompt: similarTo("Please enter your account number"),
        then: press("184748"),
        silenceAfterPrompt: 3000,
        timeout: 6000,
      },
      {
        whenPrompt: similarTo(
          "press 1 for booking a repair or 2 for changing your address"
        ),
        then: hangUp(),
        silenceAfterPrompt: 3000,
        timeout: 6000,
        },
     ],
  }
);

Quick Start

1. Create a Twilio account (referral link for $10 free if you upgrade), load it with money and rent a phone number
   1. Store an authentication token in environment variables:

   export TWILIO_ACCOUNT_SID=ACXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
   export TWILIO_AUTH_TOKEN=your_auth_token

2. Configure your environment for either Google or Amazon's transcription service
3. Install and start ngrok

   npm install ngrok -g
   ngrok http 8080

4. Run the tests

   # Local port that IVR Tester will listen on
   export LOCAL_SERVER_PORT=8080
   # URL that ngrok exposes to the outside world
   export PUBLIC_SERVER_URL=$(curl -s localhost:4040/api/tunnels | jq -r .tunnels[0].public_url)
   
   node test.js

How it works

Under the hood this orchestrates:

1. Establishing a bi-directional audio stream of the call to the IVR flow - using Twilio
2. Transcribing the voice responses from the flow - using Google Speech-to-Text
3. Using the test to conditionally respond with DTMF tones to transcripts

Writing tests

When	Overview
contains	Prompt contains a piece of text
matches	Prompt matches regular expression
similarTo	Prompt is similar to a piece of text
isAnything	Prompt can be anything

Then	Overview
press	Produces DTMF tones
hangUp	Terminates the call
doNothing	Doesn't do anything

Development

Documentation

Where possible the documentation is generated from the code using the following script in the root directory or individual packages:

yarn docs

The documentation is automatically generated and committed as part of the CI pipeline when merged to the main branch.

The official website can be previewed locally by running:

docsify serve docs

License