In this post, you’ll create a voice transcription pipeline. The objective is to use Amazon Transcribe to process an entire conversation into channels and then insert the results into an RDS MySQL database instance. To accomplish this will take two AWS Lambda functions: an HTTP application to retrieve an MP3 file and submit to Amazon Transcribe, and a callback function upon completion of the transcription to store the results into a MySQL database.
- PHP 7.4 (update
serverless.ymlfor other versions)
- Composer installed globally
- Node.js and npm
- Serverless Framework
- AWS account
Vonage API Account
To complete this tutorial, you will need a Vonage API account. If you don’t have one already, you can sign up today and start building with free credit. Once you have an account, you can find your API Key and API Secret at the top of the Vonage API Dashboard.
Clone the nexmo-community/voice-channels-aws-transcribe-php repo from GitHub, and navigate into the newly created directory to proceed.
Use Composer to Install Dependencies
This example requires the use of Composer to install dependencies and set up the autoloader.
Assuming you have Composer installed globally, run:
You will need to create AWS credentials as indicated by
Also, create a new AWS S3 bucket and make note of the URL for later use.
Link the App to Vonage
Create a Vonage Application Using the Command Line Interface
Install the CLI by following these instructions. You’ll use this to create a new Vonage Voice application that also sets up an
event_url for the app running in AWS Lambda:
nexmo app:create aws-transcribe https://<your_hostname>/webhooks/answer https://<your_hostname>/webhooks/event
NOTE: You’ll be using
<your_hostname> as a placeholder in this command. Later, after you know the URLs provided by deploying to AWS Lambda, you’ll need to update these pieces of the URLs via the Vonage API Dashboard settings for your application.
IMPORTANT: This will return an application ID and a private key. The application ID will be needed for the
nexmo link:app command as well as the
.env file later. A file named
private.key will be created in the same location/level as
server.js, by default.
Obtain a New Virtual Number
If you don’t have a number already in place, obtain one from Vonage. This can also be achieved using the CLI by running this command:
Link the Virtual Number to the Application
Finally, link the new number to the created application by running:
nexmo link:app YOUR_NUMBER YOUR_APPLICATION_ID
Rename the provided
.env.dist file to
.env and update the values as needed:
NOTE: All placeholders noted by
<> need to be updated.
Install the serverless-dotenv-plugin with the following command:
npm i -D serverless-dotenv-plugin
Deploy to Lambda
With all the above updated successfully, you can now use
Serverless to deploy the app to AWS Lambda.
Note: Make sure to visit the Vonage API Dashboard and update the
event URLs for your application with what is provided by the deployment.
Migrate Transcription to a Database
If you only require the transcription, all is done. However, to automate migrating the transcription results to the database will require another function to be deployed. Clone this nexmo-community/aws-voice-transcription-rds-callback-php repo to another location and follow the instructions in the README to get it up and running. The instructions are identical to what was done above for the first function.
Create The Trigger
Set the Rule as follows:
- Event Pattern
- Build event pattern to match events by service
- Service Name = Transcribe
- Event Type = Transcribe Job State Change
- Specific status(es) = COMPLETED
- As the Target select the Lambda function #2 created above
- Scroll down and click Configure Details.
- Give the rule a meaningful name and description, and enable it.
- Click Create rule to complete it.
Now you’re ready to test.
With the deployment completed, you should be able to place a call to your virtual number from any phone. You will hear a message about being connected, and then the recipient number will be called.
After you hang up, the MP3 file will be retrieved from Vonage and uploaded to AWS S3. Following that, a transcription job will be started. The job can be monitored in the AWS Console website after login.
Upon completion of the transcription, CloudWatch will trigger the Lambda function to parse the transcription and insert to the database.