Deployment & Serving

Catvish isn't just for training; it's a complete platform for serving your models. You can test models in the Playground or spin up a production-ready API server.

Interactive Playground

The Playground is a sandbox environment for testing your models immediately after training.

• Load any trained `best.pt` or exported `best.onnx`.
• Drag and drop test images from your file explorer.
• Adjust "Confidence Threshold" and "IoU Threshold" sliders in real-time to see how they affect detection.

Local API Server

Catvish includes a built-in FastAPI server. This allows other applications (like a web app or mobile app) to consume your model via HTTP requests.

Starting the Server

1. Navigate to Tools > Deploy.
2. Select the Model you want to serve.
3. Choose the Port (default: 8000).
4. Click Start Server.

API Documentation

Once running, you can access the model via:

curl -X POST "http://localhost:8000/predict" \ -H "accept: application/json" \ -H "Content-Type: multipart/form-data" \ -F "file=@image.jpg"

{
  "inference_time": 0.015,
  "predictions": [
    {
      "class": "person",
      "confidence": 0.98,
      "bbox": [100, 50, 200, 300] // x, y, w, h
    }
  ]
}