/**
 * Embedding Service
 * Generates vector embeddings for semantic search and similarity analysis
 * Uses text-embedding-3-small model via OpenAI (or Ollama alternatives)
 */

import { getAIProvider } from '../factory'

export interface EmbeddingResult {
  embedding: number[]
  model: string
  dimension: number
}

/**
 * Service for generating and managing text embeddings
 */
export class EmbeddingService {
  private readonly EMBEDDING_MODEL = 'text-embedding-3-small'
  private readonly EMBEDDING_DIMENSION = 1536 // OpenAI's embedding dimension

  /**
   * Generate embedding for a single text
   */
  async generateEmbedding(text: string): Promise<EmbeddingResult> {
    if (!text || text.trim().length === 0) {
      throw new Error('Cannot generate embedding for empty text')
    }

    try {
      const provider = getAIProvider()

      // Use the existing getEmbeddings method from AIProvider
      const embedding = await provider.getEmbeddings(text)

      // Validate embedding dimension
      if (embedding.length !== this.EMBEDDING_DIMENSION) {
      }

      return {
        embedding,
        model: this.EMBEDDING_MODEL,
        dimension: embedding.length
      }
    } catch (error) {
      console.error('Error generating embedding:', error)
      throw new Error(`Failed to generate embedding: ${error}`)
    }
  }

  /**
   * Generate embeddings for multiple texts in batch
   * More efficient than calling generateEmbedding multiple times
   */
  async generateBatchEmbeddings(texts: string[]): Promise<EmbeddingResult[]> {
    if (!texts || texts.length === 0) {
      return []
    }

    // Filter out empty texts
    const validTexts = texts.filter(t => t && t.trim().length > 0)

    if (validTexts.length === 0) {
      return []
    }

    try {
      const provider = getAIProvider()

      // Batch embedding using the existing getEmbeddings method
      const embeddings = await Promise.all(
        validTexts.map(text => provider.getEmbeddings(text))
      )

      return embeddings.map(embedding => ({
        embedding,
        model: this.EMBEDDING_MODEL,
        dimension: embedding.length
      }))
    } catch (error) {
      console.error('Error generating batch embeddings:', error)
      throw error
    }
  }

  /**
   * Calculate cosine similarity between two embeddings
   * Returns value between -1 and 1, where 1 is identical
   */
  calculateCosineSimilarity(embedding1: number[], embedding2: number[]): number {
    if (embedding1.length !== embedding2.length) {
      throw new Error('Embeddings must have the same dimension')
    }

    let dotProduct = 0
    let magnitude1 = 0
    let magnitude2 = 0

    for (let i = 0; i < embedding1.length; i++) {
      dotProduct += embedding1[i] * embedding2[i]
      magnitude1 += embedding1[i] * embedding1[i]
      magnitude2 += embedding2[i] * embedding2[i]
    }

    magnitude1 = Math.sqrt(magnitude1)
    magnitude2 = Math.sqrt(magnitude2)

    if (magnitude1 === 0 || magnitude2 === 0) {
      return 0
    }

    return dotProduct / (magnitude1 * magnitude2)
  }

  /**
   * Calculate similarity between an embedding and multiple other embeddings
   * Returns array of similarities
   */
  calculateSimilarities(
    queryEmbedding: number[],
    targetEmbeddings: number[][]
  ): number[] {
    return targetEmbeddings.map(embedding =>
      this.calculateCosineSimilarity(queryEmbedding, embedding)
    )
  }

  /**
   * Find most similar embeddings to a query
   * Returns top-k results with their similarities
   */
  findMostSimilar(
    queryEmbedding: number[],
    targetEmbeddings: Array<{ id: string; embedding: number[] }>,
    topK: number = 10
  ): Array<{ id: string; similarity: number }> {
    const similarities = targetEmbeddings.map(({ id, embedding }) => ({
      id,
      similarity: this.calculateCosineSimilarity(queryEmbedding, embedding)
    }))

    // Sort by similarity descending and return top-k
    return similarities
      .sort((a, b) => b.similarity - a.similarity)
      .slice(0, topK)
  }

  /**
   * Get average embedding from multiple embeddings
   * Useful for clustering or centroid calculation
   */
  averageEmbeddings(embeddings: number[][]): number[] {
    if (embeddings.length === 0) {
      throw new Error('Cannot average empty embeddings array')
    }

    const dimension = embeddings[0].length
    const average = new Array(dimension).fill(0)

    for (const embedding of embeddings) {
      if (embedding.length !== dimension) {
        throw new Error('All embeddings must have the same dimension')
      }

      for (let i = 0; i < dimension; i++) {
        average[i] += embedding[i]
      }
    }

    // Divide by number of embeddings
    return average.map(val => val / embeddings.length)
  }

  /**
   * Serialize embedding to JSON-safe format (for storage)
   */
  serialize(embedding: number[]): string {
    return JSON.stringify(embedding)
  }

  /**
   * Deserialize embedding from JSON string
   */
  deserialize(jsonString: string): number[] {
    try {
      const parsed = JSON.parse(jsonString)
      if (!Array.isArray(parsed)) {
        throw new Error('Invalid embedding format')
      }
      return parsed
    } catch (error) {
      console.error('Error deserializing embedding:', error)
      throw new Error('Failed to deserialize embedding')
    }
  }

  /**
   * Check if a note needs embedding regeneration
   * (e.g., if content has changed significantly)
   */
  shouldRegenerateEmbedding(
    noteContent: string,
    lastEmbeddingContent: string | null,
    lastAnalysis: Date | null
  ): boolean {
    // If no previous embedding, generate one
    if (!lastEmbeddingContent || !lastAnalysis) {
      return true
    }

    // If content has changed more than 20% (simple heuristic)
    const contentChanged =
      Math.abs(noteContent.length - lastEmbeddingContent.length) / lastEmbeddingContent.length > 0.2

    // If last analysis is more than 7 days old
    const daysSinceAnalysis = (Date.now() - lastAnalysis.getTime()) / (1000 * 60 * 60 * 24)
    const isStale = daysSinceAnalysis > 7

    return contentChanged || isStale
  }
}

// Singleton instance
export const embeddingService = new EmbeddingService()