Software Sushi AI Interface Starter with Gradio

1. Launch and connect

Choose the recommended x86_64 instance type in the offer, with at least 8 GiB RAM and 32 GiB encrypted gp3 root storage. Require IMDSv2 and attach your own instance profile with Session Manager permissions. Allow the outbound connectivity required by Session Manager and any selected provider or external database. No inbound application rule is needed. Wait for cloud-init and automatic first boot to finish, then open a Session Manager shell and switch to ubuntu.

sudo -iu ubuntu

2. Open a local browser tunnel

The application listens only on instance loopback port 7860. On your workstation, use AWS CLI with the Session Manager plugin and your authorized AWS profile. Supply your actual EC2 instance ID when prompted. Keep the port-forwarding session running and open http://127.0.0.1:7860 in your own browser. Keep both ends bound to localhost; do not expose the application port to the internet.

read -r -p "EC2 instance ID: " INSTANCE_ID
aws ssm start-session --target "$INSTANCE_ID" \
  --document-name AWS-StartPortForwardingSession \
  --parameters '{"portNumber":["7860"],"localPortNumber":["7860"]}'

3. Sign in after automatic startup

The instance creates its account at /home/ubuntu/.config/sushi-gradio/access.json with mode 0600. Read it only in your own unrecorded shell and enter the username and password in the Gradio login page. Keep credentials out of command arguments, logs, screenshots, recordings and support requests.

sudo -iu ubuntu cat /home/ubuntu/.config/sushi-gradio/access.json

4. Predict and independently verify

Enter helpful excellent service, select Submit and inspect the saved ID, class and probabilities. Try awful poor service for a negative example. The terminal client calls the same authenticated queue API; use a new filename for every result.

sushi-gradio predict --text 'helpful excellent service' --output ~/gradio-results/review.json
sushi-gradio read ~/gradio-results/review.json

5. Restart and back up

The service is sushi-gradio.service. Input accepts at most 512 characters and 64 English words, including a word from the original examples. The teaching model is not intended for consequential decisions. Back up /var/lib/gradio, ~/.config/sushi-gradio/access.json and ~/gradio-results together while writes are stopped or consistently snapshotted. Restarts and reboots retain the first identity and saved predictions.

sudo systemctl status sushi-gradio.service
sudo systemctl restart sushi-gradio.service

Software price and support

Software costs USD 0.03 per running instance-hour. EC2, EBS and data transfer are billed separately. The supplied local starter has no model-provider usage charge. Software Sushi provides AMI packaging and setup guidance. Contact info@softwaresushi.com for installation questions. Support coverage requires a separate agreement; this offer does not include upstream project support.

Documentation and terms

This is a repackaged open source software product wherein additional charges apply for independent AMI packaging, maintenance and setup documentation. Software Sushi is an independent packager and claims no upstream or AWS endorsement.

Original software and dependencies retain their own licenses. The offer uses the AWS Marketplace Standard EULA; see the offer for its legal and refund terms.

Official upstream documentation · Upstream source