Different AMIs are provided for each AWS region. Follow these steps to find your assigned AMIs.

1. In the AWS EC2 management console for the region, select AMIs.

2. Select Private images from the search drop-down.

3. Search for the ID of the Voci AMIs shared with your account. These IDs were provided directly in an email from support@vocitec.com.

Visit the Amazon AWS Help for additional information about locating and using shared AMIs.

Used to install and launch Voci V‑Blaze instances in Amazon Web Services (AWS) environments

After locating the Voci AMIs, launch them in a manner appropriate for your organization and environment. Consult with your administrator for specific questions about VPCs, availability zones, security groups, or other EC2 details.

The minimum connectivity required is a security group that allows you to access the instances via SSH (port 22). When prompted by the launch dialogs, configure the instance to use a valid SSH keyset for the default system user (centos for CentOS AMIs, and ec2-user for Oracle Linux AMIs).

For Voci V‑Blaze instances, launch on a g4dn.4xlarge or g3.4xlarge instance. In most cases, 20GB standard gp2 EBS is sufficient. No performance gains are expected with other storage types or larger volumes.

For Voci V‑Spark instances, launch on an m4.xlarge instance or larger. In most cases, 100GB standard gp2 EBS is sufficient to start. Most AWS customers eventually use S3 and/or EFS storage to augment the single instance. For planning, you can use 15M per saved audio hour as a rough guideline for storage sizing.

Connect via SSH to perform some basic administrative and configuration tasks before using your instance. This can be done via any SSH client using the key you specified when launching the instance. Connect as user centos . For example:

ssh -i MyAmazonEC2Voci.pem centos@123.123.123.123

For additional detail on connecting to an instance, refer to:

• Instance licensing and updates

• Voci service testing and verification

• Additional customizations

The minimum connectivity required is port 22 (SSH) for remote administration and testing. If remote access to Voci services is required, set up your security groups to allow inbound port 17171/443 for V‑Blaze or inbound port 3000/443 for V‑Spark. The instance will need to have outbound access to license.vocitec.com:443 to retrieve licensing. For Voci-specific updates, allow outbound access to vocitec.com:443, but general Internet access may also be helpful for retrieving OS updates as necessary.

Configure licensing with your authorization token to initialize and operate Voci services. Edit the appropriate configuration files for your product to ensure the auth_token= line is set to your authorization token.

V‑Blaze licensing: /opt/voci/state/license/worker.cfg

V‑Spark licensing: /opt/voci/state/license/vspark.cfg

Voci support has also provided a personal URL that can be used to configure your instance for retrieving updates and upgrades via the standard yum utility. Simply copy and paste the URL provided via email into the following command, replacing REPO_URL with your own:

sudo yum install REPO_URL

The AMI can be updated at any time by running sudo yum update .

The additional customizations described on this page are optional.

1. Download additional models and set up /opt/voci/models.list if required. In a cluster or autoscale environment that starts up directly from AMIs, implement a shared EFS mount point for /opt/voci/models.list in order to minimize ASR service startup time.

   For examples of language model configuration and more information about models.list, see Language model configuration.

2. Customize OS, logins, security configuration, etc. to conform to your company policies and standards.

Restart services and check logs for service startup to validate that the service is online, configured correctly, and starting up properly.

Voci ASR:

sudo service vociserver restart
{ tail -n +1 -f /var/log/vociserver/err.log & } | sed -n '/serving on port 17170/q'

V‑Spark:

sudo service vspark restart
tail -f /var/log/vspark/{server,backend,license}.log

Use these commands to confirm availability and operation of the V‑Blaze REST API.

The curl commands shown below will hang for up to 5-10 minutes until the vociserver ASR service is available to process them. Startup activity can be monitored with top.

The first startup may seem unusually delayed since large amounts of language model data are often still being mirrored from S3 storage in the background. Allow some time for the first launch of an instance directly from AMI.

Transcribe a file:

#This is a multiple line command
$ curl -s -X POST -Foutput=text \-
Ffile=@/opt/voci/server/examples/sample1.wav \
http://localhost:17171/transcribe ; echo

Response:

And that it was resolved in a very professional manner. Your employees are very good.

Display the active models:

$ curl -s http://localhost:17171/models | python -m json.tool

Response:

{"models": ["eng-us:callcenter"]}

Display the server status:

$ curl -s localhost:17171/status | python -m json.tool

Response:

{"status": {"#active": 0,
"#conns": 1, "#done": 0, "#queued": 0,
"#streams": 0, "idlefor": 87, "lastactive": "2019-04-11 22:56:41.522462",
"maxconns": 32, "maxstreams": 16, "started": "2019-04-11 22:54:42.464289",
"totalerrors": 0, "totalidle": 112, "totalstreams": 224}}