Deploying AI Models

Learn how to deploy AI models to your edge devices using Cliff.

Supported Model Formats

Cliff supports various model formats:

• ONNX
• TensorFlow SavedModel
• PyTorch (via ONNX conversion)
• TensorFlow Lite
• More formats coming soon

Uploading Models

From the Dashboard

• Navigate to the Models page
• Click Upload Model
• Select your model file
• Provide metadata (name, description, framework)
• Wait for upload and validation to complete

Via API

You can also upload models programmatically:

curl -X POST https://api.trycliff.com/models \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -F "file=@model.onnx" \
  -F "name=My Model" \
  -F "framework=onnx"

Deployment Process

Single Device Deployment

• Select your model from the Models page
• Click Deploy
• Choose your target device
• Configure deployment settings:
• Resource limits
• Environment variables
• Model parameters
• Click Deploy and monitor progress

Batch Deployment

To deploy to multiple devices:

• Select your model
• Click Deploy to Multiple
• Select devices from the list
• Configure settings (applied to all devices)
• Deploy

Model Configuration

Configure your model deployment:

• Resource Limits: CPU, memory, and GPU allocation
• Environment Variables: Custom configuration
• Model Parameters: Inference settings
• Auto-restart: Enable automatic restart on failure

Monitoring Deployments

Track your deployments:

• Real-time status updates
• Performance metrics
• Log streaming
• Error notifications

Updating Models

To update a deployed model:

• Upload the new model version
• Navigate to the deployment
• Click Update Model
• Select the new version
• Deploy (zero-downtime updates supported)

Best Practices

• Test models locally before deployment
• Use model quantization for smaller file sizes
• Monitor resource usage
• Set up alerts for failures
• Keep model versions organized