Skip to content

Inspector View

Clicking an experiment card opens the inspector. The layout has four main areas.


The central brain viewer displays the evolved connectome topology:

  • Sensor nodes (purple, left column) - brain inputs from the environment
  • Hidden nodes (yellow, center) - evolved internal processing neurons
  • Actuator nodes (green, right column) - brain outputs that control the agent
  • Connections - green lines for excitatory (positive weight), red for inhibitory (negative weight)
  • Signal particles - animated dots flowing along connections during replay playback, showing live signal propagation

The generation watermark (large faint number, top-left) shows the current generation. The fitness HUD (top-right) shows the best fitness at the current generation.

Click any node to see its ID, type, activation function, and connections in a tooltip.

The brain panel has six view modes, selectable from the tab bar at the top:

ModeDescription
2DForce-directed graph with signal particles (default)
3DInteractive Three.js rotatable brain with depth separation
FlowSankey-style diagram showing pathway strength as flow width
RegionsSVG wiring diagram with nodes grouped into region boxes
ActivityTemporal heatmap of all node activations over time
FindingsAutomated analysis report with topology, pathways, correlations

Click any node (in 2D or 3D view) to open the Node Inspector in the side panel:

  • Node type, ID, activation function, bias, and region
  • Live activation value
  • All incoming connections with source label, weight, and source activation
  • All outgoing connections with target label, weight, and computed signal
  • Hover a connection in the inspector to highlight it on the canvas

When connected to a local or sandbox engine, the node inspector shows ON/OFF toggle buttons on each connection. This enables ablation - disabling connections to test their functional necessity:

  1. Click a node to open the inspector
  2. Click ON next to a connection to disable it (ablate)
  3. The viewer sends the modified brain to the engine for re-simulation
  4. The replay updates showing behavior without that connection
  5. Click OFF to re-enable, or Reset to restore the original brain

A “Modified” badge shows when ablations are active, with a summary of changes.


The timeline at the bottom shows the evolution history:

  • Fitness mode - best fitness (bright line) and average fitness (dim line) across all generations, with a filled area showing the fitness range
  • Species mode - species count over generations (toggle with the Fitness/Species buttons)
  • Checkpoint markers - small triangles at the bottom edge indicate generations where a checkpoint was saved
  • Generation scrubber - drag the slider to move through generations; the brain viewer updates to show the topology at each checkpoint
  • Replay progress - a thin line along the bottom shows how far through the current replay tick sequence you are

Hover over the timeline to see exact values. Click to jump to a generation.


The right side panel shows:

  • Node Inspector - appears when a node is clicked (details, connections, ablation toggles)
  • Agent State - all agent state values during replay (bars for normalized values, numbers for counters)
  • Sensors (collapsible) - every sensor input value with purple bars
  • Outputs (collapsible) - every actuator output value with blue bars
  • World - the simulation world view (see below)
  • Stats - node count, connection count, species count, fitness, and domain-specific metrics
  • Events (collapsible) - scrolling log of simulation events, color-coded by type

The viewer renders two world types, auto-detected from the replay data:

A square canvas in the side panel showing a 2D grid with food (green), water (blue), dangerous items (red), and the agent (teal dot with glow).

A full-width strip between the brain panel and timeline showing the train route:

  • Scrolling track that follows the train position
  • Signal heads colored by aspect (green, yellow, red)
  • Station platforms with abbreviated names
  • Level crossing markers (X symbols)
  • Kilometer markers along the track
  • Speed readout (top-right) and position (top-left)

The viewport auto-scrolls to keep the train centered. The world type is determined automatically from the replay data dimensions.


ControlAction
Play/Pause button (or spacebar)Start/stop replay playback
Generation scrubberDrag to move through evolution generations
Timeline clickJump to a specific generation
SpeedPlayback runs at approximately 60 ticks per second

When you scrub to a different generation, the viewer loads that checkpoint’s brain topology and starts a new replay simulation for it.