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
Click Save in the toolbar
Enter a session name
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
Click Open in the toolbar
Select a session from the “Saved Sessions” list
The session loads, replacing current data
Method 2: From Files
Click Open in the toolbar
Click the file picker
Select the
*_config.jsonfileOptionally select the
*_data.npzfile
New Session
Click New in the toolbar
Confirm if you have unsaved changes
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 layerlayer_{id}_ids- Point IDs for each layerprojection_{id}_coords- Pre-computed projection coordinates
Importing External Data
You can create session files externally:
Create a
*_config.jsonwith the structure aboveCreate a
*_data.npzwith the arraysUse 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
Save often - VectorScope data is in-memory only
Name descriptively - Include data source and date
Save before experiments - Easy to roll back
Backup scenario files - They’re just JSON + NPZ
Version control - Session files work well with git