Clear Fracture - Belvedere Translation Module

After subscribing to the Clear Fracture Translator AMI through AWS Marketplace, launch the AMI from the EC2 console in a supported AWS Region. The AMI is pre-configured with Docker, translation models, training data, and a systemd service that starts the translation service automatically on boot.

1. Launch the AMI Launch the AMI using the EC2 Launch Instance Wizard. Recommended minimum instance type: t3.large or larger. The translation models are large and inference can be CPU-intensive. Smaller instance types such as t2.micro or t3.micro are not recommended for production translation workloads. For heavier concurrent use, consider t3.xlarge or a larger instance type.

When configuring the instance, use a security group that allows inbound TCP traffic on port 5001 from trusted client IP ranges or from an approved load balancer security group.

If SSH access is required, allow inbound TCP traffic on port 22 only from trusted administrator IP ranges.

The AMI includes two EBS volumes: Root volume: operating system, Docker image, and application files. Models volume: pre-populated translation models and training data mounted at /mnt/models.

Do not shrink the models volume below its default size. You may increase the volume size at launch if you plan to add custom models later.

2. Connect to the Instance Wait for the EC2 instance to reach the running state and pass status checks. This typically takes a few minutes.

To connect by SSH, use the default operating system user: ec2-user

Example SSH command: ssh -i /path/to/your-key.pem ec2-user@<instance-public-dns-or-ip>

3. Verify the Service The translation service runs as a systemd service named: ctranslate2.service Check the service status: sudo systemctl status ctranslate2.service

The service should show as active (running).

If needed, start the service manually: sudo systemctl start ctranslate2.service

To view service logs: sudo journalctl -u ctranslate2.service -f

You may also run the included diagnostic script: sudo diagnose_ctranslate2.sh

The diagnostic script reports container status, recent logs, service state, and network configuration.

4. Access the Web Interface The Translator web interface is available on port 5001.

Open the following URL in a browser: http://<instance-ip>:5001/

Additional interface and documentation endpoints are available at: http://<instance-ip>:5001/beta http://<instance-ip>:5001/help http://<instance-ip>:5001/api/docs

The /api/docs endpoint provides Swagger documentation for the REST API.

5. Use the REST API

The API base URL is: http://<instance-ip>:5001

List available target languages: curl http://<instance-ip>:5001/targetlangs

Translate text: curl -X POST http://<instance-ip>:5001/translate -H "Content-Type: application/json" -d '{"text": "Bonjour tout le monde", "src_lang": "fr", "target_lang": "en"}'

Example response: { "data": { "bleu": "55.29", "lang": "French", "text": "Hello everyone" }, "status": "ok" }

For batch translation, use the bulk translation endpoint: POST /bulktrans

6. Common Service Commands Run these commands on the instance as needed: sudo systemctl status ctranslate2.service sudo systemctl start ctranslate2.service sudo systemctl stop ctranslate2.service sudo systemctl restart ctranslate2.service sudo journalctl -u ctranslate2.service -f sudo docker logs -f ctranslate2

7. Security Notes The service listens on HTTP port 5001. Restrict inbound access to trusted IP ranges, private networks, or an approved load balancer. Do not expose the service broadly to the public internet unless you have added appropriate network controls, authentication, monitoring, and operational protections. SSH access should be limited to trusted administrators. Clear Fracture recommends applying operating system updates periodically and monitoring EC2 and AWS Marketplace billing for unexpected usage.