agentjs-core/dist/ml/interfaces.d.ts

A comprehensive agent-based modeling framework with built-in p5.js visualization

import { Position } from '../types/core';
import { Agent } from '../core/agents/Agent';
/**
 * Core ML behavior model interface for any domain
 */
export interface MLBehaviorModel {
    /**
     * Predict next action based on current agent state
     * @param state Current state of the agent
     * @returns Promise resolving to the predicted action
     */
    predict(state: MLAgentState): Promise<AgentAction>;
    /**
     * Load model from specified path or URL
     * @param modelPath Path to the model file
     */
    load(modelPath: string): Promise<void>;
    /**
     * Get list of required input features for this model
     * @returns Array of required input field names
     */
    getRequiredInputs(): string[];
    /**
     * Get list of possible output actions this model can produce
     * @returns Array of possible action types
     */
    getOutputActions(): string[];
    /**
     * Dispose of model and free memory
     */
    dispose(): void;
}
/**
 * Agent state representation for ML models
 */
export interface MLAgentState {
    /** Current position of the agent */
    position: Position;
    /** Agent properties as key-value pairs */
    properties: Record<string, number>;
    /** Environmental context information */
    environment: EnvironmentContext;
    /** Neighboring agents within detection radius */
    neighbors: Agent[];
    /** Optional: Previous actions taken by this agent */
    actionHistory?: AgentAction[];
}
/**
 * Agent action that can be executed
 */
export interface AgentAction {
    /** Type of action to perform */
    type: 'MOVE' | 'INTERACT' | 'CHANGE_PROPERTY' | 'CUSTOM';
    /** Parameters specific to the action type */
    parameters: Record<string, any>;
    /** Optional: Confidence score from ML prediction */
    confidence?: number;
    /** Optional: Action priority for execution ordering */
    priority?: number;
}
/**
 * Environmental context for agent decision making
 */
export interface EnvironmentContext {
    /** Environment bounds */
    bounds?: {
        width: number;
        height: number;
    };
    /** Global environment properties */
    globalProperties?: Record<string, number>;
    /** Current simulation step/time */
    currentStep?: number;
    /** Agent density in local area */
    localDensity?: number;
    /** Distance to environment boundaries */
    boundaryDistances?: {
        north: number;
        south: number;
        east: number;
        west: number;
    };
}
/**
 * Training dataset structure for generic data collection
 */
export interface TrainingDataset {
    /** State-action pairs for supervised learning */
    stateActionPairs: Array<{
        state: MLAgentState;
        action: AgentAction;
        reward?: number;
        nextState?: MLAgentState;
    }>;
    /** Metadata about the dataset */
    metadata: {
        scenarioType: string;
        agentCount: number;
        episodeLength: number;
        collectionTimestamp: number;
    };
}
/**
 * Performance metrics for ML inference
 */
export interface MLPerformanceMetrics {
    /** Average prediction time in milliseconds */
    avgPredictionTime: number;
    /** Memory usage in MB */
    memoryUsage: number;
    /** Number of predictions made */
    predictionCount: number;
    /** Error rate (failed predictions) */
    errorRate: number;
    /** Batch processing efficiency */
    batchEfficiency: number;
}
/**
 * Model loading configuration
 */
export interface ModelLoadConfig {
    /** Path to model file */
    modelPath: string;
    /** Optional model metadata */
    metadata?: {
        version: string;
        trainingDate: string;
        domain: string;
        performance: Record<string, number>;
    };
    /** Loading preferences */
    preferences?: {
        cachingEnabled: boolean;
        validationEnabled: boolean;
        optimizationLevel: 'fast' | 'balanced' | 'memory';
    };
}
/**
 * Inference request for batch processing
 */
export interface InferenceRequest {
    /** Agent requesting prediction */
    agentId: string;
    /** Current state */
    state: MLAgentState;
    /** Model to use for prediction */
    modelName: string;
    /** Request timestamp */
    timestamp: number;
    /** Priority level */
    priority: 'high' | 'normal' | 'low';
}
//# sourceMappingURL=interfaces.d.ts.map