Skip to content

πŸ”¬ Model Explorer

The Model Explorer is an interactive testing workbench for deployed ML models. Test predictions, run parameter sweeps, compare models side-by-side, and inspect model architecture β€” all from the UI.

Quick Access

Deploy a model from the Assets page β†’ open Model Explorer β†’ the model is auto-selected.

Workflow

flowchart LR
    D[Deploy Model] --> S[Select in Explorer]
    S --> I[Configure Inputs]
    I --> P[Run Prediction]
    P --> R[View Results]
    R --> C[Compare Models]
    R --> W[Parameter Sweep]

Features

Single Prediction

  1. Select a running deployment from the sidebar
  2. Use the Form or JSON editor to set inputs
  3. Click Predict to get outputs, latency, and confidence
  4. Results are displayed as charts and tables

Model Comparison (Compare mode)

Select two deployed models and run the same inputs against both. A comparison table shows:

  • Output values per metric
  • Relative % difference (Ξ”)
  • Latency comparison

Parameter Sweep

Pick a numeric parameter and define a sweep range:

  • Min / Max β€” the value range
  • Steps β€” number of data points

Results are visualized as an animated bar chart showing how outputs vary across the range.

Model Introspection

The Model Info tab shows real metadata from the loaded model:

  • Framework and version
  • Input/output shapes
  • Layer count and total parameters
  • Feature names

External API Connection

Switch to External mode to test any REST API endpoint (e.g., a production model behind an API gateway):

  • Enter endpoint URL and optional API key
  • Run predictions against external services

Live Logs

View real-time logs from the model server, useful for debugging prediction errors.

Prediction History

All predictions are saved in session and accessible from the History tab. Click any entry to reload its inputs.

REST API

Endpoint Method Description
/api/explorer/schema/{model_id} GET Get model input/output schema
/api/explorer/predict POST Run a prediction
/api/explorer/sweep POST Run parameter sweep
/api/explorer/model-info/{deployment_id} GET Get model introspection data
/api/explorer/logs/{deployment_id} GET Get deployment logs
/api/explorer/sessions GET List exploration sessions
/api/explorer/sessions POST Create exploration session
/api/explorer/sessions/{id} GET / DELETE Get or delete session