lex-semantic-priming

Spreading activation network for LegionIO cognitive agents. Nodes represent concepts; connections carry weights that strengthen each time they are traversed (Hebbian reinforcement).

What It Does

lex-semantic-priming maintains a weighted semantic network where priming one concept causes activation to spread to related concepts with depth-attenuated strength. The network learns from use: every edge traversal during spreading activation calls strengthen! on that connection, gradually building stronger pathways for frequently co-activated concepts.

• Nodes: typed concept nodes (:concept, :entity, :action, :property, :relation, :event, :context)
• Connections: directional weighted edges; weights grow on traversal (Hebbian), decay over time
• Spreading activation: BFS from a seed node across outgoing connections; activation decays by DEPTH_DECAY_FACTOR (0.5) per hop and by weight * SPREADING_FACTOR (0.6) per edge
• Decay: all node activations and connection weights decay each tick; connections pruned at MIN_WEIGHT (0.05)
• Primed nodes: activation >= 0.4; active nodes: activation > 0.1

Usage

require 'legion/extensions/semantic_priming'

client = Legion::Extensions::SemanticPriming::Client.new

# Add concept nodes
ruby_id   = client.add_node(label: 'ruby', node_type: :concept, domain: :programming)[:node_id]
oop_id    = client.add_node(label: 'object_oriented', node_type: :concept, domain: :programming)[:node_id]
class_id  = client.add_node(label: 'class', node_type: :concept, domain: :programming)[:node_id]

# Connect them
client.connect_nodes(source_id: ruby_id, target_id: oop_id)
client.connect_nodes(source_id: oop_id, target_id: class_id)

# Prime a seed and spread activation
result = client.prime_and_spread(node_id: ruby_id)
# => { success: true, node_id: ..., activation: 0.3, spread_count: 2 }

# See which nodes are primed (activation >= 0.4)
client.primed_nodes
# => { success: true, nodes: [...], count: 1 }

# Most activated nodes
client.most_primed(limit: 5)
# => { success: true, nodes: [{ label: 'ruby', activation: 0.3, ... }, ...], count: ... }

# Per-tick decay (call each cognitive cycle)
client.decay
# => { success: true, active_nodes: 2, pruned_connections: 0 }

# Network summary
client.priming_report
# => { node_count:, connection_count:, active_count:, primed_count:, average_activation:, density:, ... }

Development

bundle install
bundle exec rspec
bundle exec rubocop

License

MIT