Definitions of domain-specific and technical terms used across the AEI Framework.
The Autonomous & Evolutive Intelligence Framework, a Rust library for building dynamic multi-agent neural networks. See README.md.
An approach to software development that models complex domains in terms of bounded contexts and ubiquitous language.
Abstraction for CRUD operations on memory items.
Component providing vector-based search over memory items.
Strategy determining whether memory items are kept, archived, or deleted.
Reduces storage usage by merging or removing memory items.
Plans recurring tasks executed on manual ticks.
In-process publish/subscribe mechanism for events.
An intent to change state in the system. Commands are handled by dedicated command handlers and produce events when successful. See src/application/commands.rs.
A read-only request for information handled separately from commands. See src/application/queries.rs.
An immutable record describing a state change that occurred as a result of handling a command. Events are persisted to the event store and can be replayed to reconstruct state. See src/domain/events.rs.
Label categorizing a memory entry, enabling queries for specific kinds of experiences.
Persistent storage responsible for appending and loading domain events. Implementations live under src/infrastructure/event_store.rs.
Generic event store persisting events as JSON Lines files. See src/infrastructure/jsonl_event_store.rs.
Architectural pattern in which state is derived from a log of events rather than being stored directly. The AEI Framework rebuilds aggregates by replaying events from the store.
Component that validates and executes a command, emitting one or more events. Examples include AddRandomNeuronHandler and RemoveRandomNeuronHandler.
Shared structure bundling an event store, a hydrated Network, and a random number generator for command handlers operating on networks.
Shared structure bundling a memory event store and hydrated AdaptiveMemory with helpers to persist events and prune entries.
Component that serves a query by reading from a projection or read model. See src/application/query_handler.rs.
Process that transforms events into a read model suited for queries. Projections reside under src/infrastructure/projection.
State optimized for serving queries, maintained by projections derived from the event stream.
A domain object that enforces invariants and rebuilds its state by applying events, such as Network.
Bounded buffer storing scored experiences. Managed through event sourcing and queried via projections. See src/domain/memory.
The non-linear function a neuron applies to its input to produce an output.
Basic processing unit in the network. Defined in src/domain/neuron.rs.
Connection between neurons that carries signals. Defined in src/domain/synapse.rs.
Command that introduces a new neuron into the network at a random location. Implemented in src/application/add_random_neuron.rs.
Command that removes a randomly selected neuron from the network. Implemented in src/application/remove_random_neuron.rs.
Command that creates a synapse between two randomly chosen neurons. Implemented in src/application/add_random_synapse.rs.
Command requesting the removal of a random synapse from the network. Implemented in src/application/add_random_synapse.rs.
Command that inserts a neuron with a specific identifier and activation into the network. Handled by CommandHandler.
Command that deletes a neuron by its identifier and prunes connected synapses. Handled by CommandHandler.
Command that mutates the weight of a randomly selected synapse by adding Gaussian noise. Implemented in src/application/mutate_random_synapse_weight.rs.
Domain event recording a change in a synapse's weight. Emitted by MutateRandomSynapseWeightHandler.
Command that assigns a specific weight to an existing synapse. Implemented in src/application/set_synapse_weight.rs.
Domain event recording an explicit update of a synapse's weight. Emitted by SetSynapseWeightHandler.
Domain event emitted when a neuron is added to the network. Result of CreateNeuron.
Domain event emitted when a neuron is removed from the network. Result of RemoveNeuron.
Metric representing the exploratory potential of a neuron or synapse. Recomputed via RecalculateCuriosityScoreCommand and stored through CuriosityScoreUpdated events.
Read model mapping identifiers to curiosity scores. See src/infrastructure/projection/curiosity.rs.
Query retrieving a curiosity score for a neuron or synapse via QueryHandler.