@ninebix/nmt-system 1.0.0

package/dist/core/probabilistic-orchestrator.js

@@ -0,0 +1,798 @@
+/**
+ * Probabilistic Orchestrator - Unified control for probabilistic modules
+ *
+ * Coordinates between:
+ * - BidirectionalInferenceEngine (추론)
+ * - AttractorModel (끌개)
+ * - FourStageLearningSystem (학습)
+ * - ProbabilisticNeuronManager (뉴런)
+ * - DynamicEmbeddingManager (차원)
+ *
+ * @module core/probabilistic-orchestrator
+ */
+import { EvolutionScheduler } from './evolution-scheduler.js';
+import { cosineSimilarity } from '../utils/similarity.js';
+import { createLogger } from '../utils/logger.js';
+import { getMetrics, MetricNames } from '../utils/metrics.js';
+const logger = createLogger('orchestrator');
+/**
+ * Default configuration
+ */
+const DEFAULT_CONFIG = {
+    autoSaveInterval: 300000, // 5 minutes
+    enableAutoEvolution: true,
+    enableAutoSave: true,
+    neuronEvolveInterval: 60000, // 1 minute
+    attractorDecayInterval: 3600000, // 1 hour
+};
+/**
+ * Probabilistic Orchestrator
+ *
+ * Unified coordination layer for all probabilistic ontology modules.
+ * Manages the interaction between inference, attractors, learning,
+ * probabilistic neurons, and dynamic embeddings.
+ *
+ * @example
+ * ```typescript
+ * const orchestrator = new ProbabilisticOrchestrator({
+ *   neuronStore,
+ *   attractorModel,
+ *   learningSystem,
+ *   neuronManager,
+ *   embeddingManager,
+ *   config: { enableAutoEvolution: true }
+ * });
+ *
+ * await orchestrator.init();
+ *
+ * // Learn from user interaction
+ * await orchestrator.learnFromInteraction({
+ *   input: 'user question',
+ *   output: 'system response',
+ *   success: true
+ * });
+ *
+ * // Clean shutdown
+ * await orchestrator.shutdown();
+ * ```
+ *
+ * @see {@link OrchestratorOptions} for configuration options
+ * @see {@link UnifiedInferenceResult} for inference result structure
+ */
+export class ProbabilisticOrchestrator {
+    store;
+    inference;
+    attractors;
+    learning;
+    neurons;
+    dimensions;
+    probabilisticStore;
+    config;
+    scheduler;
+    // State management
+    initialized = false;
+    autoSaveTimer;
+    lastSaveTime;
+    constructor(options) {
+        this.store = options.neuronStore;
+        this.inference = options.inferenceEngine;
+        this.attractors = options.attractorModel;
+        this.learning = options.learningSystem;
+        this.neurons = options.neuronManager;
+        this.dimensions = options.embeddingManager;
+        this.probabilisticStore = options.probabilisticStore;
+        this.config = { ...DEFAULT_CONFIG, ...options.config };
+    }
+    // ==================== Lifecycle Methods ====================
+    /**
+     * Initialize the orchestrator
+     * - Initialize probabilistic store
+     * - Load saved state
+     * - Start auto-save timer
+     */
+    async init() {
+        if (this.initialized)
+            return;
+        const metrics = getMetrics();
+        const end = logger.time('init');
+        // Register health check
+        metrics.registerHealthCheck('orchestrator', async () => ({
+            name: 'orchestrator',
+            healthy: this.initialized,
+            message: this.initialized ? 'Orchestrator running' : 'Not initialized',
+        }));
+        // Initialize store if available
+        if (this.probabilisticStore) {
+            await this.probabilisticStore.init();
+            logger.info('Probabilistic store initialized');
+            // Load saved state
+            await this.loadState();
+        }
+        // Start auto-save timer
+        if (this.config.enableAutoSave && this.probabilisticStore) {
+            this.startAutoSave();
+            logger.info('Auto-save enabled', { interval: this.config.autoSaveInterval });
+        }
+        // Start evolution scheduler
+        if (this.config.enableAutoEvolution && (this.neurons || this.attractors)) {
+            this.scheduler = new EvolutionScheduler({
+                neuronManager: this.neurons,
+                attractorModel: this.attractors,
+                neuronEvolveInterval: this.config.neuronEvolveInterval,
+                attractorDecayInterval: this.config.attractorDecayInterval,
+                onEvolve: async (stats) => {
+                    logger.debug('Evolution completed', {
+                        neuronsEvolved: stats.neuronsEvolved,
+                        attractorsDecayed: stats.attractorsDecayed,
+                    });
+                    // Auto-save after evolution
+                    if (this.config.enableAutoSave && this.probabilisticStore) {
+                        await this.saveState();
+                    }
+                },
+            });
+            this.scheduler.start();
+            logger.info('Evolution scheduler started', {
+                neuronInterval: this.config.neuronEvolveInterval,
+                attractorInterval: this.config.attractorDecayInterval,
+            });
+        }
+        this.initialized = true;
+        end();
+    }
+    /**
+     * Shutdown the orchestrator
+     * - Save current state
+     * - Stop timers
+     * - Close store
+     */
+    async shutdown() {
+        if (!this.initialized)
+            return;
+        const metrics = getMetrics();
+        metrics.unregisterHealthCheck('orchestrator');
+        logger.info('Shutting down orchestrator');
+        // Stop evolution scheduler
+        if (this.scheduler) {
+            this.scheduler.stop();
+            this.scheduler = undefined;
+            logger.debug('Evolution scheduler stopped');
+        }
+        // Stop auto-save timer
+        this.stopAutoSave();
+        // Save final state
+        if (this.probabilisticStore) {
+            await this.saveState();
+            await this.probabilisticStore.close();
+            logger.debug('State saved and store closed');
+        }
+        this.initialized = false;
+        logger.info('Orchestrator shutdown complete');
+    }
+    /**
+     * Save all module states to persistent storage
+     */
+    async saveState() {
+        const result = {
+            saved: false,
+            timestamp: new Date().toISOString(),
+            modules: {
+                attractors: false,
+                learning: false,
+                neurons: false,
+                dimensions: false,
+            },
+        };
+        if (!this.probabilisticStore) {
+            return result;
+        }
+        const states = {};
+        // Serialize each module
+        if (this.attractors) {
+            states.attractors = this.attractors.serialize();
+            result.modules.attractors = true;
+        }
+        if (this.learning) {
+            states.learning = this.learning.serialize();
+            result.modules.learning = true;
+        }
+        if (this.neurons) {
+            states.neurons = this.neurons.serialize();
+            result.modules.neurons = true;
+        }
+        if (this.dimensions) {
+            states.dimensions = this.dimensions.serializeAll();
+            result.modules.dimensions = true;
+        }
+        // Save all at once
+        await this.probabilisticStore.saveAll(states);
+        this.lastSaveTime = result.timestamp;
+        result.saved = true;
+        return result;
+    }
+    /**
+     * Load all module states from persistent storage
+     */
+    async loadState() {
+        const result = {
+            loaded: false,
+            modules: {
+                attractors: false,
+                learning: false,
+                neurons: false,
+                dimensions: false,
+            },
+        };
+        if (!this.probabilisticStore) {
+            return result;
+        }
+        const states = await this.probabilisticStore.loadAll();
+        // Load each module
+        if (states.attractors && this.attractors) {
+            this.attractors.load(states.attractors);
+            result.modules.attractors = true;
+        }
+        if (states.learning && this.learning) {
+            this.learning.load(states.learning);
+            result.modules.learning = true;
+        }
+        if (states.neurons && this.neurons) {
+            this.neurons.load(states.neurons);
+            result.modules.neurons = true;
+        }
+        if (states.dimensions && this.dimensions) {
+            this.dimensions.loadAll(states.dimensions);
+            result.modules.dimensions = true;
+        }
+        result.loaded = true;
+        return result;
+    }
+    // ==================== Core Methods ====================
+    /**
+     * Perform unified inference combining all available modules.
+     *
+     * Combines bidirectional inference, attractor influences, and probabilistic
+     * state analysis to provide comprehensive reasoning about a neuron.
+     *
+     * @param neuronId - The ID of the neuron to analyze
+     * @param options - Configuration for the inference
+     * @param options.includeAttractors - Include attractor influence analysis (default: true)
+     * @param options.includeProbabilistic - Include probabilistic state analysis (default: true)
+     * @param options.maxDepth - Maximum inference depth (default: 3)
+     * @returns Unified inference result with paths, influences, and recommendations
+     *
+     * @example
+     * ```typescript
+     * const result = await orchestrator.unifiedInfer('neuron-123', {
+     *   includeAttractors: true,
+     *   includeProbabilistic: true,
+     *   maxDepth: 5
+     * });
+     * console.log(result.forwardPaths); // Cause -> Effect paths
+     * console.log(result.dominantAttractor); // Most influential goal
+     * ```
+     */
+    async unifiedInfer(neuronId, options = {}) {
+        const metrics = getMetrics();
+        const timer = metrics.startTimer(MetricNames.INFERENCE_DURATION_MS, { type: 'unified' });
+        metrics.increment(MetricNames.INFERENCE_REQUESTS, { type: 'unified' });
+        const result = {
+            forwardPaths: [],
+            backwardPaths: [],
+            attractorInfluences: new Map(),
+            recommendations: [],
+        };
+        const neuron = await this.store.getNeuron(neuronId);
+        if (!neuron) {
+            timer();
+            return result;
+        }
+        // 1. Bidirectional inference
+        if (this.inference) {
+            const forward = await this.inference.forwardInfer(neuron);
+            result.forwardPaths = forward.map(r => ({
+                neuronId: r.target?.id ?? r.source.id,
+                confidence: r.confidence,
+            }));
+            const backward = await this.inference.backwardInfer(neuron);
+            result.backwardPaths = backward.map(r => ({
+                neuronId: r.source.id,
+                confidence: r.confidence,
+            }));
+        }
+        // 2. Attractor influences
+        if (this.attractors && options.includeAttractors !== false) {
+            result.attractorInfluences = this.attractors.calculateInfluence(neuron.embedding);
+            const dominant = this.attractors.getDominantAttractor(neuron.embedding);
+            if (dominant) {
+                result.dominantAttractor = dominant.id;
+                // Find path to dominant attractor
+                const pathResult = await this.attractors.findPathToAttractor(neuronId, dominant.id);
+                if (pathResult) {
+                    result.pathToGoal = pathResult.path;
+                }
+            }
+        }
+        // 3. Probabilistic state analysis
+        if (this.neurons && options.includeProbabilistic !== false) {
+            const probNeuron = this.neurons.getNeuron(neuronId);
+            if (probNeuron) {
+                const states = probNeuron.superposition.states;
+                let mostProbableIndex = 0;
+                let maxAmplitude = 0;
+                for (let i = 0; i < states.length; i++) {
+                    if (states[i].probability > maxAmplitude) {
+                        maxAmplitude = states[i].probability;
+                        mostProbableIndex = i;
+                    }
+                }
+                result.stateDistribution = {
+                    entropy: this.neurons.getUncertainty(neuronId),
+                    mostProbableState: mostProbableIndex,
+                };
+            }
+        }
+        // 4. Generate recommendations
+        result.recommendations = this.generateRecommendations(result, neuron);
+        timer();
+        return result;
+    }
+    /**
+     * Learn from a user interaction using the full 4-stage learning system.
+     *
+     * Implements the complete learning pipeline:
+     * 1. Extract - Find meaningful content from input/output
+     * 2. Patterns - Identify recurring patterns
+     * 3. Process - Learn the transformation steps
+     * 4. Outcome - Record success/failure for future reference
+     *
+     * @param data - The interaction data to learn from
+     * @param data.input - The user's input text
+     * @param data.output - The system's output text
+     * @param data.success - Whether the interaction was successful
+     * @param data.feedback - Quality score (0-1, optional)
+     * @param data.inputNeuronId - Existing neuron ID for input (optional)
+     * @param data.outputNeuronId - Existing neuron ID for output (optional)
+     * @returns Learning result with statistics on what was learned
+     *
+     * @example
+     * ```typescript
+     * const result = await orchestrator.learnFromInteraction({
+     *   input: "How do I implement a binary search?",
+     *   output: "Here's how to implement binary search...",
+     *   success: true,
+     *   feedback: 0.9
+     * });
+     *
+     * console.log(`Learned ${result.patternsLearned} patterns`);
+     * ```
+     */
+    async learnFromInteraction(data) {
+        const metrics = getMetrics();
+        metrics.increment(MetricNames.LEARNING_SESSIONS, { type: 'interaction' });
+        const result = {
+            patternsLearned: 0,
+            attractorsUpdated: 0,
+            neuronsUpdated: 0,
+            extractsCreated: 0,
+            processLearned: false,
+            outcomeRecorded: false,
+        };
+        // 1. Get or create input/output neurons
+        let inputNeuronId = data.inputNeuronId;
+        let outputNeuronId = data.outputNeuronId;
+        // If no input neuron provided, try to find similar or note absence
+        if (!inputNeuronId && this.dimensions) {
+            // Create a simple embedding from input text for matching
+            const inputEmbedding = this.createSimpleEmbedding(data.input);
+            const similar = await this.store.findSimilar(inputEmbedding, 1);
+            if (similar.length > 0) {
+                const similarity = cosineSimilarity(inputEmbedding, similar[0].embedding);
+                if (similarity > 0.8) {
+                    inputNeuronId = similar[0].id;
+                }
+            }
+        }
+        // 2. Full learning system integration
+        if (this.learning && inputNeuronId) {
+            const learningResult = await this.learning.learnFromInteraction(inputNeuronId, data.input, data.processSteps ?? [], outputNeuronId, data.output, data.success, data.feedback ?? (data.success ? 0.8 : 0.2));
+            result.extractsCreated = learningResult.extracts.length;
+            result.patternsLearned = learningResult.patterns.length;
+            result.processLearned = learningResult.process !== null;
+            result.outcomeRecorded = learningResult.outcome !== null;
+        }
+        // 3. Update attractors based on success/failure
+        if (this.attractors && data.success !== undefined) {
+            if (data.success && data.feedback !== undefined && data.feedback > 0.7) {
+                // Create or strengthen attractor for successful pattern
+                const outputEmbedding = outputNeuronId
+                    ? (await this.store.getNeuron(outputNeuronId))?.embedding
+                    : this.createSimpleEmbedding(data.output);
+                if (outputEmbedding) {
+                    // Find or create attractor for this successful outcome
+                    const dominant = this.attractors.getDominantAttractor(outputEmbedding);
+                    if (dominant) {
+                        // Strengthen existing attractor
+                        result.attractorsUpdated = 1;
+                    }
+                    else {
+                        // Create new attractor for highly successful outcomes
+                        if (data.feedback >= 0.9) {
+                            const attractorId = `attr_success_${Date.now()}`;
+                            this.attractors.createAttractor(attractorId, `Success Pattern`, `Auto-created from successful interaction`, outputEmbedding, { strength: data.feedback });
+                            result.attractorsUpdated = 1;
+                        }
+                    }
+                }
+            }
+        }
+        // 4. Create probabilistic neurons for new patterns
+        if (this.neurons && inputNeuronId) {
+            const inputNeuron = await this.store.getNeuron(inputNeuronId);
+            if (inputNeuron) {
+                const existingProb = this.neurons.getNeuron(inputNeuronId);
+                if (!existingProb) {
+                    await this.neurons.createProbabilisticNeuron(inputNeuron);
+                    result.neuronsUpdated++;
+                }
+            }
+        }
+        if (this.neurons && outputNeuronId) {
+            const outputNeuron = await this.store.getNeuron(outputNeuronId);
+            if (outputNeuron) {
+                const existingProb = this.neurons.getNeuron(outputNeuronId);
+                if (!existingProb) {
+                    await this.neurons.createProbabilisticNeuron(outputNeuron);
+                    result.neuronsUpdated++;
+                }
+            }
+        }
+        // 5. Store neuron IDs in result
+        result.inputNeuronId = inputNeuronId;
+        result.outputNeuronId = outputNeuronId;
+        // 6. Auto-save after learning
+        if (this.config.enableAutoSave && this.probabilisticStore) {
+            await this.saveState();
+        }
+        // 7. Record metrics
+        if (result.extractsCreated > 0) {
+            metrics.increment(MetricNames.EXTRACTS_CREATED, undefined, result.extractsCreated);
+        }
+        if (result.patternsLearned > 0) {
+            metrics.increment(MetricNames.PATTERNS_LEARNED, undefined, result.patternsLearned);
+        }
+        if (result.attractorsUpdated > 0) {
+            metrics.increment(MetricNames.ATTRACTOR_ACTIVATIONS, undefined, result.attractorsUpdated);
+        }
+        return result;
+    }
+    /**
+     * Provide feedback for a previous interaction.
+     *
+     * Updates the learning system with quality feedback, which helps
+     * improve future pattern matching and process selection.
+     *
+     * @param inputNeuronId - The input neuron from the original interaction
+     * @param outputNeuronId - The output neuron from the original interaction
+     * @param quality - Quality score between 0 (poor) and 1 (excellent)
+     * @param feedbackText - Optional textual feedback description
+     * @returns Whether the feedback was successfully recorded
+     *
+     * @example
+     * ```typescript
+     * await orchestrator.provideFeedback(
+     *   'input-neuron-id',
+     *   'output-neuron-id',
+     *   0.85,
+     *   'Good answer but could be more concise'
+     * );
+     * ```
+     */
+    async provideFeedback(inputNeuronId, outputNeuronId, quality, feedbackText) {
+        if (!this.learning) {
+            return { updated: false };
+        }
+        // Find the process that was used
+        const inputNeuron = await this.store.getNeuron(inputNeuronId);
+        if (!inputNeuron) {
+            return { updated: false };
+        }
+        // Get relevant patterns to find the process
+        const relevantPatterns = await this.learning.findRelevantPatterns(inputNeuron.embedding, 1);
+        if (relevantPatterns.length > 0) {
+            // Record outcome with feedback
+            await this.learning.learnOutcome(relevantPatterns[0].id, // Use pattern ID as process reference
+            inputNeuronId, outputNeuronId, quality >= 0.5, quality, feedbackText);
+            // Update attractors based on feedback
+            if (this.attractors && quality >= 0.8) {
+                const outputNeuron = await this.store.getNeuron(outputNeuronId);
+                if (outputNeuron) {
+                    const dominant = this.attractors.getDominantAttractor(outputNeuron.embedding);
+                    if (!dominant) {
+                        const attractorId = `attr_feedback_${Date.now()}`;
+                        this.attractors.createAttractor(attractorId, `Feedback Pattern`, `Auto-created from positive feedback`, outputNeuron.embedding, { strength: quality });
+                    }
+                }
+            }
+            return { updated: true };
+        }
+        return { updated: false };
+    }
+    /**
+     * Reinforce a successful path between neurons.
+     *
+     * Strengthens the connection between two neurons that produced a
+     * successful outcome, making similar paths more likely in the future.
+     *
+     * @param fromId - Source neuron ID
+     * @param toId - Target neuron ID
+     * @param strength - Reinforcement strength (default: 0.1)
+     * @returns Whether the path was successfully reinforced
+     *
+     * @example
+     * ```typescript
+     * // After a successful interaction, reinforce the path
+     * await orchestrator.reinforceSuccessfulPath(
+     *   'input-neuron',
+     *   'output-neuron',
+     *   0.2
+     * );
+     * ```
+     */
+    async reinforceSuccessfulPath(fromId, toId, strength = 0.1) {
+        // Update probabilistic neurons if they exist
+        if (this.neurons) {
+            const fromProb = this.neurons.getNeuron(fromId);
+            const toProb = this.neurons.getNeuron(toId);
+            if (fromProb && toProb) {
+                // Entangle successful pairs
+                this.neurons.entangle(fromId, toId);
+            }
+        }
+        // Update attractor model - strengthen path
+        if (this.attractors) {
+            const fromNeuron = await this.store.getNeuron(fromId);
+            const toNeuron = await this.store.getNeuron(toId);
+            if (fromNeuron && toNeuron) {
+                // The path is successful, so the destination could be an attractor
+                const dominant = this.attractors.getDominantAttractor(toNeuron.embedding);
+                if (dominant) {
+                    // Path leads to a known goal - this is good
+                    return { reinforced: true };
+                }
+            }
+        }
+        return { reinforced: true };
+    }
+    /**
+     * Create a goal-driven neuron biased toward an attractor.
+     *
+     * Creates a probabilistic neuron with states weighted toward
+     * the specified attractor (goal), implementing teleological
+     * determinism where future goals influence present state.
+     *
+     * @param baseNeuronId - The base neuron to enhance with goal direction
+     * @param attractorId - The target attractor (goal) to move toward
+     * @returns Goal-driven neuron info or null if creation failed
+     */
+    async createGoalDrivenNeuron(baseNeuronId, attractorId) {
+        if (!this.neurons || !this.attractors) {
+            return null;
+        }
+        const baseNeuron = await this.store.getNeuron(baseNeuronId);
+        if (!baseNeuron)
+            return null;
+        // Create probabilistic neuron
+        const probNeuron = await this.neurons.createProbabilisticNeuron(baseNeuron);
+        // Get path to attractor
+        const pathResult = await this.attractors.findPathToAttractor(baseNeuronId, attractorId);
+        return {
+            neuronId: baseNeuronId,
+            initialState: 0,
+            pathProbability: pathResult?.probability ?? 0,
+        };
+    }
+    /**
+     * Expand embedding dimensions for a new concept.
+     *
+     * Registers a new semantic dimension, allowing the system to
+     * capture novel concepts that weren't in the original embedding space.
+     *
+     * @param concept - The concept name to register
+     * @param category - Semantic category for organization (default: 'custom')
+     * @returns The new dimension name and total dimension count
+     */
+    async expandForConcept(concept, category = 'custom') {
+        if (!this.dimensions) {
+            return { dimensionName: concept, totalDimensions: 384 };
+        }
+        this.dimensions.registerDimension(concept, {
+            name: concept,
+            semanticCategory: category,
+            description: `Dimension for concept: ${concept}`,
+            createdAt: new Date().toISOString(),
+            usageCount: 0,
+        });
+        const stats = this.dimensions.getStats();
+        return {
+            dimensionName: concept,
+            totalDimensions: stats.totalDimensions,
+        };
+    }
+    /**
+     * Get comprehensive statistics from all modules.
+     *
+     * Provides a unified view of the system's current state including
+     * inference availability, attractor counts, learning progress,
+     * neuron statistics, and persistence status.
+     *
+     * @returns Aggregated statistics from all modules
+     *
+     * @example
+     * ```typescript
+     * const stats = orchestrator.getStats();
+     * console.log(`Active attractors: ${stats.attractors.active}`);
+     * console.log(`Patterns learned: ${stats.learning.patterns}`);
+     * console.log(`Last saved: ${stats.persistence.lastSave}`);
+     * ```
+     */
+    getStats() {
+        return {
+            inference: {
+                available: !!this.inference,
+            },
+            attractors: this.attractors ? {
+                count: this.attractors.getStats().totalAttractors,
+                active: this.attractors.getStats().activeAttractors,
+            } : { count: 0, active: 0 },
+            learning: this.learning ? {
+                patterns: this.learning.getStats().patterns,
+                processes: this.learning.getStats().processes,
+            } : { patterns: 0, processes: 0 },
+            neurons: this.neurons ? {
+                count: this.neurons.getStats().totalNeurons,
+                avgEntropy: this.neurons.getStats().averageEntropy,
+            } : { count: 0, avgEntropy: 0 },
+            dimensions: this.dimensions ? {
+                total: this.dimensions.getStats().totalDimensions,
+                expanded: this.dimensions.getStats().expandedDimensions,
+            } : { total: 384, expanded: 0 },
+            persistence: {
+                enabled: !!this.probabilisticStore,
+                lastSave: this.lastSaveTime,
+            },
+        };
+    }
+    /**
+     * Check if orchestrator is initialized
+     */
+    isInitialized() {
+        return this.initialized;
+    }
+    /**
+     * Get configuration
+     */
+    getConfig() {
+        return { ...this.config };
+    }
+    /**
+     * Get evolution scheduler status
+     */
+    getEvolutionStatus() {
+        if (!this.scheduler)
+            return null;
+        return this.scheduler.getStatus();
+    }
+    /**
+     * Trigger manual evolution of all modules
+     */
+    async triggerEvolution() {
+        if (!this.scheduler)
+            return null;
+        return this.scheduler.evolveNow();
+    }
+    /**
+     * Update evolution intervals (applies immediately)
+     */
+    setEvolutionIntervals(intervals) {
+        if (intervals.neuronEvolveInterval !== undefined) {
+            this.config.neuronEvolveInterval = intervals.neuronEvolveInterval;
+        }
+        if (intervals.attractorDecayInterval !== undefined) {
+            this.config.attractorDecayInterval = intervals.attractorDecayInterval;
+        }
+        if (this.scheduler) {
+            this.scheduler.setIntervals({
+                neuronEvolve: intervals.neuronEvolveInterval,
+                attractorDecay: intervals.attractorDecayInterval,
+            });
+        }
+    }
+    // ==================== Auto-Save Management ====================
+    startAutoSave() {
+        if (this.autoSaveTimer) {
+            clearInterval(this.autoSaveTimer);
+        }
+        this.autoSaveTimer = setInterval(async () => {
+            try {
+                await this.saveState();
+            }
+            catch (error) {
+                console.error('[Orchestrator] Auto-save failed:', error);
+            }
+        }, this.config.autoSaveInterval);
+    }
+    stopAutoSave() {
+        if (this.autoSaveTimer) {
+            clearInterval(this.autoSaveTimer);
+            this.autoSaveTimer = undefined;
+        }
+    }
+    // ==================== Private Helpers ====================
+    generateRecommendations(result, neuron) {
+        const recommendations = [];
+        // Forward paths with high confidence
+        for (const path of result.forwardPaths.slice(0, 3)) {
+            if (path.confidence > 0.5) {
+                recommendations.push({
+                    neuronId: path.neuronId,
+                    reason: 'Strong forward causal connection',
+                    score: path.confidence,
+                });
+            }
+        }
+        // Backward paths (abductive reasoning)
+        for (const path of result.backwardPaths.slice(0, 2)) {
+            if (path.confidence > 0.5) {
+                recommendations.push({
+                    neuronId: path.neuronId,
+                    reason: 'Likely cause (abductive)',
+                    score: path.confidence * 0.9,
+                });
+            }
+        }
+        // Attractor-influenced paths
+        if (result.pathToGoal && result.pathToGoal.length > 1) {
+            const nextStep = result.pathToGoal[1];
+            recommendations.push({
+                neuronId: nextStep,
+                reason: `Step toward goal attractor`,
+                score: 0.8,
+            });
+        }
+        // Sort by score
+        recommendations.sort((a, b) => b.score - a.score);
+        return recommendations.slice(0, 5);
+    }
+    /**
+     * Create a simple embedding from text
+     * Uses a deterministic hash-based approach for text without external embedding service
+     */
+    createSimpleEmbedding(text) {
+        const embedding = new Float32Array(384);
+        const words = text.toLowerCase().split(/\s+/);
+        // Simple bag-of-words style embedding
+        for (let i = 0; i < words.length; i++) {
+            const word = words[i];
+            for (let j = 0; j < word.length; j++) {
+                const charCode = word.charCodeAt(j);
+                const index = (charCode * (i + 1) * (j + 1)) % 384;
+                embedding[index] += 1 / (1 + Math.log(1 + i));
+            }
+        }
+        // Normalize
+        let norm = 0;
+        for (let i = 0; i < 384; i++) {
+            norm += embedding[i] * embedding[i];
+        }
+        norm = Math.sqrt(norm);
+        if (norm > 0) {
+            for (let i = 0; i < 384; i++) {
+                embedding[i] /= norm;
+            }
+        }
+        return embedding;
+    }
+}
+export default ProbabilisticOrchestrator;
+//# sourceMappingURL=probabilistic-orchestrator.js.map