Sessions

VectorScope sessions let you save and restore your entire workspace.

What Gets Saved

A session includes:

  • All layers - Source and derived

  • All points - Vectors, labels, metadata

  • All transformations - Type, parameters

  • All projections - Type, parameters, pre-computed coordinates

  • Column configuration - For tabular data sources

  • Random seeds - For reproducible t-SNE

What Doesn’t Get Saved

  • Selection state

  • Viewport layout

  • View sets

  • UI settings

Saving a Session

  1. Click Save in the toolbar

  2. Enter a session name

  3. Click Save

Files are saved to the scenarios/ directory:

  • {name}_config.json - Structure and metadata

  • {name}_data.npz - Vectors and coordinates

Quick Save

After saving once:

  • Click Save to overwrite the current session

  • Click Save As to save with a new name

Opening a Session

Method 1: From Browser

  1. Click Open in the toolbar

  2. Select a session from the “Saved Sessions” list

  3. The session loads, replacing current data

Method 2: From Files

  1. Click Open in the toolbar

  2. Click the file picker

  3. Select the *_config.json file

  4. Optionally select the *_data.npz file

New Session

  1. Click New in the toolbar

  2. Confirm if you have unsaved changes

  3. All data is cleared

Session Files

Config File (JSON)

Contains the structure:

{
  "name": "My Session",
  "description": "Analysis of embeddings",
  "layers": [
    {
      "id": "uuid-here",
      "name": "Source Data",
      "dimensionality": 30,
      "is_derived": false,
      "column_names": ["dim_0", "dim_1", ...],
      "feature_columns": ["dim_0", "dim_1", ...],
      "label_column": null
    }
  ],
  "transformations": [...],
  "projections": [...],
  "point_metadata": {...}
}

Data File (NPZ)

Contains NumPy arrays:

  • layer_{id}_vectors - Point vectors for each layer

  • layer_{id}_ids - Point IDs for each layer

  • projection_{id}_coords - Pre-computed projection coordinates

Importing External Data

You can create session files externally:

  1. Create a *_config.json with the structure above

  2. Create a *_data.npz with the arrays

  3. Use Open → file picker to load them

This is useful for:

  • Loading embeddings from neural networks

  • Batch processing

  • Integration with other tools

Scenarios (Built-in)

The scenarios/ directory contains pre-built sessions:

  • Iris - Classic iris flower dataset

  • Wine - Wine cultivar classification

  • Breast Cancer - Cancer diagnosis features

  • Digits - Handwritten digit pixels

  • Diabetes - Disease progression features

These are available in the “Test Scenarios” section of the Open dialog.

Best Practices

  1. Save often - VectorScope data is in-memory only

  2. Name descriptively - Include data source and date

  3. Save before experiments - Easy to roll back

  4. Backup scenario files - They’re just JSON + NPZ

  5. Version control - Session files work well with git