Skip to content

quale inspect

Loads a checkpoint file and prints the best genome’s topology. Optionally performs pathway analysis showing which sensors influence each actuator.

quale inspect is how you look inside an evolved brain to understand what emerged. After a run completes, it tells you which neurons exist, how they are connected, and - with --pathways - which sensors actually influence each actuator. This is where you find out whether the agent evolved the behavior you intended or found some other strategy entirely.

quale inspect <checkpoint-file> [flags]
ArgumentDescription
<checkpoint-file>Path to a .quale-ckpt checkpoint file
FlagDefaultDescription
--pathwaysfalseShow sensor-to-actuator pathway analysis using reverse BFS

Without --pathways:

=== Checkpoint: Generation 500 ===
Population size: 200
=== Best Genome ===
ID: a1b2c3d4
Fitness: 42.50
Generation: 498
Nodes: 18 total (14 input, 2 hidden, 2 output)
Connections: 12 total (10 enabled, 2 disabled)
Sensors: [hunger thirst energy health ...]
Actuators: [move_n move_s move_e move_w eat drink]

With --pathways:

--- Hidden Nodes ---
Node 16: activation=Sigmoid bias=0.3421
Node 17: activation=ReLU bias=-0.1234
--- Enabled Connections ---
IN:hunger(0) -> H(16) weight=0.8234 inn=42
H(16) -> OUT:eat(14) weight=1.2345 inn=56
--- Output Node Biases ---
eat: bias=0.1234
drink: bias=-0.5678
--- Pathway Analysis ---
Sensors that influence each actuator (direct + indirect paths):
eat: hunger, item_nearest_color, item_nearest_smell
drink: thirst
move_n: food_nearby_n, hunger
Terminal window
# Basic topology inspection
quale inspect checkpoints/phase1/checkpoint_gen500.quale-ckpt
# Full pathway analysis
quale inspect checkpoints/phase1/checkpoint_gen500.quale-ckpt --pathways

Not yet implemented. Planned for future versions. Will replay a recorded simulation from a .quale-replay file.

quale replay

Currently takes no arguments, prints “replay is not yet implemented in v0.1” to stderr, and exits with code 1.