forensic-pathways/src/utils/embeddings.ts

// src/utils/embeddings.ts
import { promises as fs } from 'fs';
import path from 'path';
import { getCompressedToolsDataForAI } from './dataService.js';

interface EmbeddingData {
  id: string;
  type: 'tool' | 'concept';
  name: string;
  content: string;
  embedding: number[];
  metadata: {
    domains?: string[];
    phases?: string[];
    tags?: string[];
    skillLevel?: string;
    type?: string;
  };
}

interface EmbeddingsDatabase {
  version: string;
  lastUpdated: number;
  embeddings: EmbeddingData[];
}

interface EmbeddingSearchResult extends EmbeddingData {
  similarity: number;
}

class EmbeddingsService {
  private embeddings: EmbeddingData[] = [];
  private isInitialized = false;
  private readonly embeddingsPath = path.join(process.cwd(), 'data', 'embeddings.json');
  private readonly batchSize: number;
  private readonly batchDelay: number;
  private readonly enabled: boolean;

  constructor() {
    this.enabled = process.env.AI_EMBEDDINGS_ENABLED === 'true';
    this.batchSize = parseInt(process.env.AI_EMBEDDINGS_BATCH_SIZE || '20', 10);
    this.batchDelay = parseInt(process.env.AI_EMBEDDINGS_BATCH_DELAY_MS || '1000', 10);
  }

  async initialize(): Promise<void> {
    if (!this.enabled) {
      console.log('[EMBEDDINGS] Embeddings disabled, skipping initialization');
      return;
    }

    try {
      console.log('[EMBEDDINGS] Initializing embeddings system...');
      
      // Create data directory if it doesn't exist
      await fs.mkdir(path.dirname(this.embeddingsPath), { recursive: true });
      
      const toolsData = await getCompressedToolsDataForAI();
      const currentDataHash = this.hashData(toolsData);
      
      // Try to load existing embeddings
      const existingEmbeddings = await this.loadEmbeddings();
      
      if (existingEmbeddings && existingEmbeddings.version === currentDataHash) {
        console.log('[EMBEDDINGS] Using cached embeddings');
        this.embeddings = existingEmbeddings.embeddings;
      } else {
        console.log('[EMBEDDINGS] Generating new embeddings...');
        await this.generateEmbeddings(toolsData, currentDataHash);
      }
      
      this.isInitialized = true;
      console.log(`[EMBEDDINGS] Initialized with ${this.embeddings.length} embeddings`);
      
    } catch (error) {
      console.error('[EMBEDDINGS] Failed to initialize:', error);
      this.isInitialized = false;
    }
  }

  private hashData(data: any): string {
    return Buffer.from(JSON.stringify(data)).toString('base64').slice(0, 32);
  }

  private async loadEmbeddings(): Promise<EmbeddingsDatabase | null> {
    try {
      const data = await fs.readFile(this.embeddingsPath, 'utf8');
      return JSON.parse(data);
    } catch (error) {
      console.log('[EMBEDDINGS] No existing embeddings found');
      return null;
    }
  }

  private async saveEmbeddings(version: string): Promise<void> {
    const database: EmbeddingsDatabase = {
      version,
      lastUpdated: Date.now(),
      embeddings: this.embeddings
    };
    
    await fs.writeFile(this.embeddingsPath, JSON.stringify(database, null, 2));
    console.log(`[EMBEDDINGS] Saved ${this.embeddings.length} embeddings to disk`);
  }

  private createContentString(item: any): string {
    const parts = [
      item.name,
      item.description || '',
      ...(item.tags || []),
      ...(item.domains || []),
      ...(item.phases || [])
    ];
    
    return parts.filter(Boolean).join(' ').toLowerCase();
  }

  private async generateEmbeddingsBatch(contents: string[]): Promise<number[][]> {
    const endpoint = process.env.AI_EMBEDDINGS_ENDPOINT;
    const apiKey = process.env.AI_EMBEDDINGS_API_KEY;
    const model = process.env.AI_EMBEDDINGS_MODEL;

    if (!endpoint || !apiKey || !model) {
      throw new Error('Missing embeddings API configuration');
    }

    const response = await fetch(endpoint, {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
        'Authorization': `Bearer ${apiKey}`
      },
      body: JSON.stringify({
        model,
        input: contents
      })
    });

    if (!response.ok) {
      const error = await response.text();
      throw new Error(`Embeddings API error: ${response.status} - ${error}`);
    }

    const data = await response.json();
    return data.data.map((item: any) => item.embedding);
  }

  private async generateEmbeddings(toolsData: any, version: string): Promise<void> {
    const allItems = [
      ...toolsData.tools.map((tool: any) => ({ ...tool, type: 'tool' })),
      ...toolsData.concepts.map((concept: any) => ({ ...concept, type: 'concept' }))
    ];

    const contents = allItems.map(item => this.createContentString(item));
    this.embeddings = [];

    // Process in batches to respect rate limits
    for (let i = 0; i < contents.length; i += this.batchSize) {
      const batch = contents.slice(i, i + this.batchSize);
      const batchItems = allItems.slice(i, i + this.batchSize);
      
      console.log(`[EMBEDDINGS] Processing batch ${Math.ceil((i + 1) / this.batchSize)} of ${Math.ceil(contents.length / this.batchSize)}`);
      
      try {
        const embeddings = await this.generateEmbeddingsBatch(batch);
        
        embeddings.forEach((embedding, index) => {
          const item = batchItems[index];
          this.embeddings.push({
            id: `${item.type}_${item.name.replace(/[^a-zA-Z0-9]/g, '_')}`,
            type: item.type,
            name: item.name,
            content: batch[index],
            embedding,
            metadata: {
              domains: item.domains,
              phases: item.phases,
              tags: item.tags,
              skillLevel: item.skillLevel,
              type: item.type
            }
          });
        });
        
        // Rate limiting delay between batches
        if (i + this.batchSize < contents.length) {
          await new Promise(resolve => setTimeout(resolve, this.batchDelay));
        }
        
      } catch (error) {
        console.error(`[EMBEDDINGS] Failed to process batch ${Math.ceil((i + 1) / this.batchSize)}:`, error);
        throw error;
      }
    }

    await this.saveEmbeddings(version);
  }

  public async embedText(text: string): Promise<number[]> {
    // Re‑use the private batch helper to avoid auth duplication
    const [embedding] = await this.generateEmbeddingsBatch([text.toLowerCase()]);
    return embedding;
  }

  private cosineSimilarity(a: number[], b: number[]): number {
    let dotProduct = 0;
    let normA = 0;
    let normB = 0;
    
    for (let i = 0; i < a.length; i++) {
      dotProduct += a[i] * b[i];
      normA += a[i] * a[i];
      normB += b[i] * b[i];
    }
    
    return dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));
  }

  async findSimilar(query: string, maxResults: number = 30, threshold: number = 0.3): Promise<EmbeddingSearchResult[]> {
    if (!this.enabled || !this.isInitialized || this.embeddings.length === 0) {
      return [];
    }

    try {
      const queryEmbeddings = await this.generateEmbeddingsBatch([query.toLowerCase()]);
      const queryEmbedding = queryEmbeddings[0];

      const similarities = this.embeddings.map(item => ({
        ...item,
        similarity: this.cosineSimilarity(queryEmbedding, item.embedding)
      }));

      return similarities
        .filter(item => item.similarity >= threshold)
        .sort((a, b) => b.similarity - a.similarity)
        .slice(0, maxResults);

    } catch (error) {
      console.error('[EMBEDDINGS] Failed to find similar items:', error);
      return [];
    }
  }

  isEnabled(): boolean {
    return this.enabled && this.isInitialized;
  }

  getStats(): { enabled: boolean; initialized: boolean; count: number } {
    return {
      enabled: this.enabled,
      initialized: this.isInitialized,
      count: this.embeddings.length
    };
  }
}



const embeddingsService = new EmbeddingsService();

export { embeddingsService, type EmbeddingData, type EmbeddingSearchResult };

if (typeof window === 'undefined' && process.env.NODE_ENV !== 'test') {
  embeddingsService.initialize().catch(error => {
    console.error('[EMBEDDINGS] Auto-initialization failed:', error);
  });
}