agentdb 2.0.0-alpha.2.11 → 2.0.0-alpha.2.13

package/dist/agentdb.browser.js

@@ -1,4 +1,4 @@
-/*! AgentDB Browser Bundle v2.0.0-alpha.2.10 | MIT License | https://agentdb.ruv.io */
+/*! AgentDB Browser Bundle v2.0.0-alpha.2.13 | MIT License | https://agentdb.ruv.io */
 var __defProp = Object.defineProperty;
 var __getOwnPropNames = Object.getOwnPropertyNames;
 var __esm = (fn, res) => function __init() {

package/dist/agentdb.browser.js.map

@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["agentdb.wasm-loader.js", "../src/browser/ProductQuantization.ts", "../src/browser/HNSWIndex.ts", "../src/browser/AdvancedFeatures.ts", "../src/browser/AttentionBrowser.ts", "../src/browser/index.ts"],
-  "sourcesContent": ["/**\n * AgentDB WASM Attention Module Loader\n * Lazy-loaded high-performance attention mechanisms\n *\n * Features:\n * - Flash Attention\n * - Hyperbolic Attention\n * - Memory Consolidation\n */\n\nlet wasmModule = null;\nlet wasmLoading = null;\nlet wasmLoadError = null;\n\n/**\n * Initialize WASM module (lazy loaded on first use)\n */\nexport async function initWASM() {\n  if (wasmModule) return wasmModule;\n  if (wasmLoading) return wasmLoading;\n\n  wasmLoading = (async () => {\n    try {\n      // Check for WASM support\n      if (typeof WebAssembly === 'undefined') {\n        throw new Error('WebAssembly not supported in this browser');\n      }\n\n      // Check for SIMD support\n      const simdSupported = await detectWasmSIMD();\n      console.log(`WASM SIMD support: ${simdSupported}`);\n\n      // In a real implementation, this would load the actual WASM binary\n      // For now, we create a mock implementation\n      wasmModule = {\n        flashAttention: createFlashAttentionMock(),\n        hyperbolicAttention: createHyperbolicAttentionMock(),\n        memoryConsolidation: createMemoryConsolidationMock(),\n        simdSupported\n      };\n\n      console.log('\u2705 WASM attention module loaded');\n      return wasmModule;\n    } catch (error) {\n      wasmLoadError = error;\n      console.warn('\u26A0\uFE0F  WASM loading failed, using fallback:', error.message);\n\n      // Return fallback implementations\n      wasmModule = {\n        flashAttention: createFlashAttentionMock(),\n        hyperbolicAttention: createHyperbolicAttentionMock(),\n        memoryConsolidation: createMemoryConsolidationMock(),\n        simdSupported: false\n      };\n\n      return wasmModule;\n    } finally {\n      wasmLoading = null;\n    }\n  })();\n\n  return wasmLoading;\n}\n\n/**\n * Detect WASM SIMD support\n */\nasync function detectWasmSIMD() {\n  try {\n    const simdTest = new Uint8Array([\n      0x00, 0x61, 0x73, 0x6d, 0x01, 0x00, 0x00, 0x00,\n      0x01, 0x05, 0x01, 0x60, 0x00, 0x01, 0x7b, 0x03,\n      0x02, 0x01, 0x00, 0x0a, 0x0a, 0x01, 0x08, 0x00,\n      0xfd, 0x0c, 0xfd, 0x0c, 0xfd, 0x54, 0x0b\n    ]);\n\n    const module = await WebAssembly.instantiate(simdTest);\n    return module instanceof WebAssembly.Instance;\n  } catch {\n    return false;\n  }\n}\n\n/**\n * Mock implementations (replaced by actual WASM in production)\n */\nfunction createFlashAttentionMock() {\n  return (query, keys, values, options = {}) => {\n    const { dim = 384, numHeads = 4, blockSize = 64 } = options;\n    const seqLen = keys.length / dim;\n    const output = new Float32Array(query.length);\n\n    // Simple attention for demonstration\n    for (let i = 0; i < query.length; i += dim) {\n      const q = query.slice(i, i + dim);\n      let sumWeights = 0;\n      const weights = new Float32Array(seqLen);\n\n      // Compute attention weights\n      for (let j = 0; j < seqLen; j++) {\n        const k = keys.slice(j * dim, (j + 1) * dim);\n        let dot = 0;\n        for (let d = 0; d < dim; d++) {\n          dot += q[d] * k[d];\n        }\n        weights[j] = Math.exp(dot / Math.sqrt(dim));\n        sumWeights += weights[j];\n      }\n\n      // Normalize and apply to values\n      for (let j = 0; j < seqLen; j++) {\n        weights[j] /= sumWeights;\n        const v = values.slice(j * dim, (j + 1) * dim);\n        for (let d = 0; d < dim; d++) {\n          output[i + d] += weights[j] * v[d];\n        }\n      }\n    }\n\n    return output;\n  };\n}\n\nfunction createHyperbolicAttentionMock() {\n  return (query, keys, options = {}) => {\n    const { curvature = -1.0 } = options;\n    const k = Math.abs(curvature);\n    const similarities = new Float32Array(keys.length / query.length);\n\n    // Hyperbolic distance computation\n    for (let i = 0; i < similarities.length; i++) {\n      const offset = i * query.length;\n      let dotProduct = 0;\n      let normQ = 0;\n      let normK = 0;\n\n      for (let j = 0; j < query.length; j++) {\n        dotProduct += query[j] * keys[offset + j];\n        normQ += query[j] * query[j];\n        normK += keys[offset + j] * keys[offset + j];\n      }\n\n      // Poincar\u00E9 distance approximation\n      const euclidean = Math.sqrt(normQ + normK - 2 * dotProduct);\n      const poincare = Math.acosh(1 + 2 * k * euclidean * euclidean);\n      similarities[i] = 1 / (1 + poincare);\n    }\n\n    return similarities;\n  };\n}\n\nfunction createMemoryConsolidationMock() {\n  return (memories, options = {}) => {\n    const { threshold = 0.8, maxClusters = 10 } = options;\n    const consolidated = [];\n    const used = new Set();\n\n    // Simple clustering by similarity\n    for (let i = 0; i < memories.length; i++) {\n      if (used.has(i)) continue;\n\n      const cluster = [memories[i]];\n      used.add(i);\n\n      for (let j = i + 1; j < memories.length; j++) {\n        if (used.has(j)) continue;\n\n        // Compute similarity\n        let dot = 0;\n        let norm1 = 0;\n        let norm2 = 0;\n        for (let k = 0; k < memories[i].length; k++) {\n          dot += memories[i][k] * memories[j][k];\n          norm1 += memories[i][k] * memories[i][k];\n          norm2 += memories[j][k] * memories[j][k];\n        }\n        const similarity = dot / (Math.sqrt(norm1 * norm2) || 1);\n\n        if (similarity > threshold) {\n          cluster.push(memories[j]);\n          used.add(j);\n        }\n      }\n\n      // Average cluster members\n      const avg = new Float32Array(memories[i].length);\n      for (const mem of cluster) {\n        for (let k = 0; k < avg.length; k++) {\n          avg[k] += mem[k] / cluster.length;\n        }\n      }\n\n      consolidated.push({\n        memory: avg,\n        count: cluster.size,\n        members: cluster\n      });\n\n      if (consolidated.length >= maxClusters) break;\n    }\n\n    return consolidated;\n  };\n}\n\nexport { wasmModule, wasmLoadError };\n", "/**\n * Product Quantization for Browser\n *\n * Compresses high-dimensional vectors using product quantization.\n * Achieves 4-32x memory reduction with minimal accuracy loss.\n *\n * Features:\n * - PQ8: 8 subvectors, 256 centroids each (4x compression)\n * - PQ16: 16 subvectors, 256 centroids each (8x compression)\n * - Asymmetric distance computation (ADC)\n * - K-means clustering for codebook training\n *\n * Performance:\n * - Memory: Float32 (4 bytes) \u2192 uint8 (1 byte) per subvector\n * - Speed: ~1.5x slower search vs uncompressed\n * - Accuracy: 95-99% recall@10\n */\n\nexport interface PQConfig {\n  dimension: number;\n  numSubvectors: number;      // 8, 16, 32, or 64\n  numCentroids: number;        // Usually 256 (uint8)\n  maxIterations?: number;      // K-means iterations\n  convergenceThreshold?: number;\n}\n\nexport interface PQCodebook {\n  subvectorDim: number;\n  numSubvectors: number;\n  numCentroids: number;\n  centroids: Float32Array[];   // [numSubvectors][numCentroids][subvectorDim]\n}\n\nexport interface CompressedVector {\n  codes: Uint8Array;           // [numSubvectors] - indices into centroids\n  norm: number;                // Original vector norm (for normalization)\n}\n\nexport class ProductQuantization {\n  private config: Required<PQConfig>;\n  private codebook: PQCodebook | null = null;\n  private trained = false;\n\n  constructor(config: PQConfig) {\n    this.config = {\n      dimension: config.dimension,\n      numSubvectors: config.numSubvectors,\n      numCentroids: config.numCentroids,\n      maxIterations: config.maxIterations || 50,\n      convergenceThreshold: config.convergenceThreshold || 1e-4\n    };\n\n    // Validate config\n    if (this.config.dimension % this.config.numSubvectors !== 0) {\n      throw new Error(`Dimension ${this.config.dimension} must be divisible by numSubvectors ${this.config.numSubvectors}`);\n    }\n  }\n\n  /**\n   * Train codebook using k-means on training vectors\n   */\n  async train(vectors: Float32Array[]): Promise<void> {\n    if (vectors.length === 0) {\n      throw new Error('Training requires at least one vector');\n    }\n\n    const subvectorDim = this.config.dimension / this.config.numSubvectors;\n    const centroids: Float32Array[] = [];\n\n    console.log(`[PQ] Training ${this.config.numSubvectors} subvectors with ${this.config.numCentroids} centroids each...`);\n\n    // Train each subvector independently\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const startDim = s * subvectorDim;\n      const endDim = startDim + subvectorDim;\n\n      // Extract subvectors\n      const subvectors = vectors.map(v => v.slice(startDim, endDim));\n\n      // Run k-means\n      const subCentroids = await this.kMeans(subvectors, this.config.numCentroids);\n      centroids.push(...subCentroids);\n\n      if ((s + 1) % 4 === 0 || s === this.config.numSubvectors - 1) {\n        console.log(`[PQ] Trained ${s + 1}/${this.config.numSubvectors} subvectors`);\n      }\n    }\n\n    this.codebook = {\n      subvectorDim,\n      numSubvectors: this.config.numSubvectors,\n      numCentroids: this.config.numCentroids,\n      centroids\n    };\n\n    this.trained = true;\n    console.log('[PQ] Training complete');\n  }\n\n  /**\n   * K-means clustering for centroids\n   */\n  private async kMeans(vectors: Float32Array[], k: number): Promise<Float32Array[]> {\n    const dim = vectors[0].length;\n    const n = vectors.length;\n\n    // Initialize centroids with k-means++\n    const centroids = this.kMeansPlusPlus(vectors, k);\n    const assignments = new Uint32Array(n);\n    let prevInertia = Infinity;\n\n    for (let iter = 0; iter < this.config.maxIterations; iter++) {\n      // Assign vectors to nearest centroid\n      let inertia = 0;\n      for (let i = 0; i < n; i++) {\n        let minDist = Infinity;\n        let minIdx = 0;\n\n        for (let j = 0; j < k; j++) {\n          const dist = this.squaredDistance(vectors[i], centroids[j]);\n          if (dist < minDist) {\n            minDist = dist;\n            minIdx = j;\n          }\n        }\n\n        assignments[i] = minIdx;\n        inertia += minDist;\n      }\n\n      // Check convergence\n      if (Math.abs(prevInertia - inertia) < this.config.convergenceThreshold) {\n        break;\n      }\n      prevInertia = inertia;\n\n      // Update centroids\n      const counts = new Uint32Array(k);\n      const sums = Array.from({ length: k }, () => new Float32Array(dim));\n\n      for (let i = 0; i < n; i++) {\n        const cluster = assignments[i];\n        counts[cluster]++;\n        for (let d = 0; d < dim; d++) {\n          sums[cluster][d] += vectors[i][d];\n        }\n      }\n\n      for (let j = 0; j < k; j++) {\n        if (counts[j] > 0) {\n          for (let d = 0; d < dim; d++) {\n            centroids[j][d] = sums[j][d] / counts[j];\n          }\n        }\n      }\n    }\n\n    return centroids;\n  }\n\n  /**\n   * K-means++ initialization for better centroid selection\n   */\n  private kMeansPlusPlus(vectors: Float32Array[], k: number): Float32Array[] {\n    const n = vectors.length;\n    const dim = vectors[0].length;\n    const centroids: Float32Array[] = [];\n\n    // Choose first centroid randomly\n    const firstIdx = Math.floor(Math.random() * n);\n    centroids.push(new Float32Array(vectors[firstIdx]));\n\n    // Choose remaining centroids\n    for (let i = 1; i < k; i++) {\n      const distances = new Float32Array(n);\n      let sumDistances = 0;\n\n      // Calculate distances to nearest centroid\n      for (let j = 0; j < n; j++) {\n        let minDist = Infinity;\n        for (const centroid of centroids) {\n          const dist = this.squaredDistance(vectors[j], centroid);\n          minDist = Math.min(minDist, dist);\n        }\n        distances[j] = minDist;\n        sumDistances += minDist;\n      }\n\n      // Choose next centroid with probability proportional to distance\u00B2\n      let r = Math.random() * sumDistances;\n      for (let j = 0; j < n; j++) {\n        r -= distances[j];\n        if (r <= 0) {\n          centroids.push(new Float32Array(vectors[j]));\n          break;\n        }\n      }\n    }\n\n    return centroids;\n  }\n\n  /**\n   * Compress a vector using trained codebook\n   */\n  compress(vector: Float32Array): CompressedVector {\n    if (!this.trained || !this.codebook) {\n      throw new Error('Codebook must be trained before compression');\n    }\n\n    const codes = new Uint8Array(this.config.numSubvectors);\n    const subvectorDim = this.codebook.subvectorDim;\n\n    // Compute norm for later reconstruction\n    let norm = 0;\n    for (let i = 0; i < vector.length; i++) {\n      norm += vector[i] * vector[i];\n    }\n    norm = Math.sqrt(norm);\n\n    // Encode each subvector\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const startDim = s * subvectorDim;\n      const subvector = vector.slice(startDim, startDim + subvectorDim);\n\n      // Find nearest centroid\n      let minDist = Infinity;\n      let minIdx = 0;\n\n      const centroidOffset = s * this.config.numCentroids;\n      for (let c = 0; c < this.config.numCentroids; c++) {\n        const centroid = this.codebook.centroids[centroidOffset + c];\n        const dist = this.squaredDistance(subvector, centroid);\n        if (dist < minDist) {\n          minDist = dist;\n          minIdx = c;\n        }\n      }\n\n      codes[s] = minIdx;\n    }\n\n    return { codes, norm };\n  }\n\n  /**\n   * Decompress a vector (approximate reconstruction)\n   */\n  decompress(compressed: CompressedVector): Float32Array {\n    if (!this.codebook) {\n      throw new Error('Codebook not available');\n    }\n\n    const vector = new Float32Array(this.config.dimension);\n    const subvectorDim = this.codebook.subvectorDim;\n\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const code = compressed.codes[s];\n      const centroidOffset = s * this.config.numCentroids;\n      const centroid = this.codebook.centroids[centroidOffset + code];\n\n      const startDim = s * subvectorDim;\n      for (let d = 0; d < subvectorDim; d++) {\n        vector[startDim + d] = centroid[d];\n      }\n    }\n\n    return vector;\n  }\n\n  /**\n   * Asymmetric Distance Computation (ADC)\n   * Computes distance from query vector to compressed vector\n   */\n  asymmetricDistance(query: Float32Array, compressed: CompressedVector): number {\n    if (!this.codebook) {\n      throw new Error('Codebook not available');\n    }\n\n    let distance = 0;\n    const subvectorDim = this.codebook.subvectorDim;\n\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const code = compressed.codes[s];\n      const centroidOffset = s * this.config.numCentroids;\n      const centroid = this.codebook.centroids[centroidOffset + code];\n\n      const startDim = s * subvectorDim;\n      const querySubvector = query.slice(startDim, startDim + subvectorDim);\n\n      distance += this.squaredDistance(querySubvector, centroid);\n    }\n\n    return Math.sqrt(distance);\n  }\n\n  /**\n   * Batch compression for multiple vectors\n   */\n  batchCompress(vectors: Float32Array[]): CompressedVector[] {\n    return vectors.map(v => this.compress(v));\n  }\n\n  /**\n   * Get memory savings\n   */\n  getCompressionRatio(): number {\n    // Original: dimension * 4 bytes (Float32)\n    // Compressed: numSubvectors * 1 byte (Uint8) + 4 bytes (norm)\n    const originalBytes = this.config.dimension * 4;\n    const compressedBytes = this.config.numSubvectors + 4;\n    return originalBytes / compressedBytes;\n  }\n\n  /**\n   * Export codebook for persistence\n   */\n  exportCodebook(): string {\n    if (!this.codebook) {\n      throw new Error('No codebook to export');\n    }\n\n    return JSON.stringify({\n      config: this.config,\n      codebook: {\n        subvectorDim: this.codebook.subvectorDim,\n        numSubvectors: this.codebook.numSubvectors,\n        numCentroids: this.codebook.numCentroids,\n        centroids: this.codebook.centroids.map(c => Array.from(c))\n      }\n    });\n  }\n\n  /**\n   * Import codebook\n   */\n  importCodebook(json: string): void {\n    const data = JSON.parse(json);\n    this.config = data.config;\n    this.codebook = {\n      subvectorDim: data.codebook.subvectorDim,\n      numSubvectors: data.codebook.numSubvectors,\n      numCentroids: data.codebook.numCentroids,\n      centroids: data.codebook.centroids.map((c: number[]) => new Float32Array(c))\n    };\n    this.trained = true;\n  }\n\n  /**\n   * Utility: Squared Euclidean distance\n   */\n  private squaredDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      const diff = a[i] - b[i];\n      sum += diff * diff;\n    }\n    return sum;\n  }\n\n  /**\n   * Get statistics\n   */\n  getStats(): {\n    trained: boolean;\n    compressionRatio: number;\n    memoryPerVector: number;\n    codebookSize: number;\n  } {\n    const compressionRatio = this.getCompressionRatio();\n    const memoryPerVector = this.config.numSubvectors + 4; // codes + norm\n    const codebookSize = this.codebook\n      ? this.config.numSubvectors * this.config.numCentroids * (this.config.dimension / this.config.numSubvectors) * 4\n      : 0;\n\n    return {\n      trained: this.trained,\n      compressionRatio,\n      memoryPerVector,\n      codebookSize\n    };\n  }\n}\n\n/**\n * Helper function to create PQ8 (8 subvectors, 4x compression)\n */\nexport function createPQ8(dimension: number): ProductQuantization {\n  return new ProductQuantization({\n    dimension,\n    numSubvectors: 8,\n    numCentroids: 256,\n    maxIterations: 50\n  });\n}\n\n/**\n * Helper function to create PQ16 (16 subvectors, 8x compression)\n */\nexport function createPQ16(dimension: number): ProductQuantization {\n  return new ProductQuantization({\n    dimension,\n    numSubvectors: 16,\n    numCentroids: 256,\n    maxIterations: 50\n  });\n}\n\n/**\n * Helper function to create PQ32 (32 subvectors, 16x compression)\n */\nexport function createPQ32(dimension: number): ProductQuantization {\n  return new ProductQuantization({\n    dimension,\n    numSubvectors: 32,\n    numCentroids: 256,\n    maxIterations: 50\n  });\n}\n", "/**\n * HNSW (Hierarchical Navigable Small World) Index for Browser\n *\n * JavaScript implementation of HNSW algorithm for fast approximate nearest neighbor search.\n * Achieves O(log n) search complexity vs O(n) for linear scan.\n *\n * Features:\n * - Multi-layer graph structure\n * - Probabilistic layer assignment\n * - Greedy search algorithm\n * - Dynamic insertion\n * - Configurable M (connections per node)\n * - Configurable efConstruction and efSearch\n *\n * Performance:\n * - 10-20x faster than linear scan (vs 150x for native HNSW)\n * - Memory: ~16 bytes per edge + vector storage\n * - Suitable for datasets up to 100K vectors in browser\n */\n\nexport interface HNSWConfig {\n  dimension: number;\n  M: number;                    // Max connections per node (default: 16)\n  efConstruction: number;       // Size of dynamic candidate list (default: 200)\n  efSearch: number;             // Size of search candidate list (default: 50)\n  ml: number;                   // Layer assignment multiplier (default: 1/ln(2))\n  maxLayers: number;            // Maximum number of layers (default: 16)\n  distanceFunction?: 'cosine' | 'euclidean' | 'manhattan';\n}\n\nexport interface HNSWNode {\n  id: number;\n  vector: Float32Array;\n  level: number;\n  connections: Map<number, number[]>; // layer -> [neighbor ids]\n}\n\nexport interface SearchResult {\n  id: number;\n  distance: number;\n  vector: Float32Array;\n}\n\nclass MinHeap<T> {\n  private items: Array<{ item: T; priority: number }> = [];\n\n  push(item: T, priority: number): void {\n    this.items.push({ item, priority });\n    this.bubbleUp(this.items.length - 1);\n  }\n\n  pop(): T | undefined {\n    if (this.items.length === 0) return undefined;\n    const result = this.items[0].item;\n    const last = this.items.pop()!;\n    if (this.items.length > 0) {\n      this.items[0] = last;\n      this.bubbleDown(0);\n    }\n    return result;\n  }\n\n  peek(): T | undefined {\n    return this.items[0]?.item;\n  }\n\n  size(): number {\n    return this.items.length;\n  }\n\n  private bubbleUp(index: number): void {\n    while (index > 0) {\n      const parentIndex = Math.floor((index - 1) / 2);\n      if (this.items[index].priority >= this.items[parentIndex].priority) break;\n      [this.items[index], this.items[parentIndex]] = [this.items[parentIndex], this.items[index]];\n      index = parentIndex;\n    }\n  }\n\n  private bubbleDown(index: number): void {\n    while (true) {\n      const leftChild = 2 * index + 1;\n      const rightChild = 2 * index + 2;\n      let smallest = index;\n\n      if (leftChild < this.items.length && this.items[leftChild].priority < this.items[smallest].priority) {\n        smallest = leftChild;\n      }\n      if (rightChild < this.items.length && this.items[rightChild].priority < this.items[smallest].priority) {\n        smallest = rightChild;\n      }\n      if (smallest === index) break;\n\n      [this.items[index], this.items[smallest]] = [this.items[smallest], this.items[index]];\n      index = smallest;\n    }\n  }\n}\n\nexport class HNSWIndex {\n  private config: Required<HNSWConfig>;\n  private nodes: Map<number, HNSWNode> = new Map();\n  private entryPoint: number | null = null;\n  private currentId = 0;\n  private ml: number;\n\n  constructor(config: Partial<HNSWConfig> = {}) {\n    this.config = {\n      dimension: config.dimension || 384,\n      M: config.M || 16,\n      efConstruction: config.efConstruction || 200,\n      efSearch: config.efSearch || 50,\n      ml: config.ml || 1 / Math.log(2),\n      maxLayers: config.maxLayers || 16,\n      distanceFunction: config.distanceFunction || 'cosine'\n    };\n\n    this.ml = this.config.ml;\n  }\n\n  /**\n   * Add vector to index\n   */\n  add(vector: Float32Array, id?: number): number {\n    const nodeId = id !== undefined ? id : this.currentId++;\n    const level = this.randomLevel();\n\n    const node: HNSWNode = {\n      id: nodeId,\n      vector,\n      level,\n      connections: new Map()\n    };\n\n    // Initialize connections for each layer\n    for (let l = 0; l <= level; l++) {\n      node.connections.set(l, []);\n    }\n\n    if (this.entryPoint === null) {\n      // First node\n      this.entryPoint = nodeId;\n      this.nodes.set(nodeId, node);\n      return nodeId;\n    }\n\n    // Find nearest neighbors at each layer\n    const ep = this.entryPoint;\n    let nearest = ep;\n\n    // Search from top layer to target layer + 1\n    for (let lc = this.nodes.get(ep)!.level; lc > level; lc--) {\n      nearest = this.searchLayer(vector, nearest, 1, lc)[0];\n    }\n\n    // Insert node at layers 0 to level\n    for (let lc = Math.min(level, this.nodes.get(ep)!.level); lc >= 0; lc--) {\n      const candidates = this.searchLayer(vector, nearest, this.config.efConstruction, lc);\n\n      // Select M neighbors\n      const M = lc === 0 ? this.config.M * 2 : this.config.M;\n      const neighbors = this.selectNeighbors(vector, candidates, M);\n\n      // Add bidirectional connections\n      for (const neighbor of neighbors) {\n        this.connect(nodeId, neighbor, lc);\n        this.connect(neighbor, nodeId, lc);\n\n        // Prune connections if necessary\n        const neighborNode = this.nodes.get(neighbor)!;\n        const neighborConnections = neighborNode.connections.get(lc)!;\n        if (neighborConnections.length > M) {\n          const newNeighbors = this.selectNeighbors(\n            neighborNode.vector,\n            neighborConnections,\n            M\n          );\n          neighborNode.connections.set(lc, newNeighbors);\n        }\n      }\n\n      nearest = candidates[0];\n    }\n\n    // Update entry point if necessary\n    if (level > this.nodes.get(this.entryPoint)!.level) {\n      this.entryPoint = nodeId;\n    }\n\n    this.nodes.set(nodeId, node);\n    return nodeId;\n  }\n\n  /**\n   * Search for k nearest neighbors\n   */\n  search(query: Float32Array, k: number, ef?: number): SearchResult[] {\n    if (this.entryPoint === null) return [];\n\n    ef = ef || Math.max(this.config.efSearch, k);\n\n    let ep = this.entryPoint;\n    let nearest = ep;\n\n    // Search from top to layer 1\n    for (let lc = this.nodes.get(ep)!.level; lc > 0; lc--) {\n      nearest = this.searchLayer(query, nearest, 1, lc)[0];\n    }\n\n    // Search at layer 0\n    const candidates = this.searchLayer(query, nearest, ef, 0);\n\n    // Convert to SearchResult and return top k\n    return candidates\n      .slice(0, k)\n      .map(id => ({\n        id,\n        distance: this.distance(query, this.nodes.get(id)!.vector),\n        vector: this.nodes.get(id)!.vector\n      }));\n  }\n\n  /**\n   * Search at specific layer\n   */\n  private searchLayer(query: Float32Array, ep: number, ef: number, layer: number): number[] {\n    const visited = new Set<number>();\n    const candidates = new MinHeap<number>();\n    const w = new MinHeap<number>();\n\n    const dist = this.distance(query, this.nodes.get(ep)!.vector);\n    candidates.push(ep, dist);\n    w.push(ep, -dist); // Max heap (negate for min heap)\n    visited.add(ep);\n\n    while (candidates.size() > 0) {\n      const c = candidates.pop()!;\n      const fDist = -w.peek()!; // Furthest point distance\n\n      const cDist = this.distance(query, this.nodes.get(c)!.vector);\n      if (cDist > fDist) break;\n\n      const neighbors = this.nodes.get(c)!.connections.get(layer) || [];\n      for (const e of neighbors) {\n        if (visited.has(e)) continue;\n        visited.add(e);\n\n        const eDist = this.distance(query, this.nodes.get(e)!.vector);\n        const fDist = -w.peek()!;\n\n        if (eDist < fDist || w.size() < ef) {\n          candidates.push(e, eDist);\n          w.push(e, -eDist);\n\n          if (w.size() > ef) {\n            w.pop();\n          }\n        }\n      }\n    }\n\n    // Return ef nearest neighbors\n    const result: number[] = [];\n    while (w.size() > 0) {\n      result.unshift(w.pop()!);\n    }\n    return result;\n  }\n\n  /**\n   * Select best neighbors using heuristic\n   */\n  private selectNeighbors(base: Float32Array, candidates: number[], M: number): number[] {\n    if (candidates.length <= M) return candidates;\n\n    // Sort by distance\n    const sorted = candidates\n      .map(id => ({\n        id,\n        distance: this.distance(base, this.nodes.get(id)!.vector)\n      }))\n      .sort((a, b) => a.distance - b.distance);\n\n    return sorted.slice(0, M).map(x => x.id);\n  }\n\n  /**\n   * Connect two nodes at layer\n   */\n  private connect(from: number, to: number, layer: number): void {\n    const node = this.nodes.get(from)!;\n    const connections = node.connections.get(layer)!;\n    if (!connections.includes(to)) {\n      connections.push(to);\n    }\n  }\n\n  /**\n   * Random level assignment\n   */\n  private randomLevel(): number {\n    let level = 0;\n    while (Math.random() < this.ml && level < this.config.maxLayers - 1) {\n      level++;\n    }\n    return level;\n  }\n\n  /**\n   * Distance function\n   */\n  private distance(a: Float32Array, b: Float32Array): number {\n    switch (this.config.distanceFunction) {\n      case 'cosine':\n        return 1 - this.cosineSimilarity(a, b);\n      case 'euclidean':\n        return this.euclideanDistance(a, b);\n      case 'manhattan':\n        return this.manhattanDistance(a, b);\n      default:\n        return 1 - this.cosineSimilarity(a, b);\n    }\n  }\n\n  private cosineSimilarity(a: Float32Array, b: Float32Array): number {\n    let dotProduct = 0;\n    let normA = 0;\n    let normB = 0;\n\n    for (let i = 0; i < a.length; i++) {\n      dotProduct += a[i] * b[i];\n      normA += a[i] * a[i];\n      normB += b[i] * b[i];\n    }\n\n    return dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));\n  }\n\n  private euclideanDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      const diff = a[i] - b[i];\n      sum += diff * diff;\n    }\n    return Math.sqrt(sum);\n  }\n\n  private manhattanDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      sum += Math.abs(a[i] - b[i]);\n    }\n    return sum;\n  }\n\n  /**\n   * Get index statistics\n   */\n  getStats(): {\n    numNodes: number;\n    numLayers: number;\n    avgConnections: number;\n    entryPointLevel: number;\n    memoryBytes: number;\n  } {\n    if (this.nodes.size === 0) {\n      return {\n        numNodes: 0,\n        numLayers: 0,\n        avgConnections: 0,\n        entryPointLevel: 0,\n        memoryBytes: 0\n      };\n    }\n\n    const maxLevel = Math.max(...Array.from(this.nodes.values()).map(n => n.level));\n    let totalConnections = 0;\n\n    for (const node of this.nodes.values()) {\n      for (const connections of node.connections.values()) {\n        totalConnections += connections.length;\n      }\n    }\n\n    const avgConnections = totalConnections / this.nodes.size;\n\n    // Estimate memory: vector + connections + metadata\n    const vectorBytes = this.config.dimension * 4; // Float32Array\n    const connectionBytes = avgConnections * 4; // number array\n    const metadataBytes = 100; // rough estimate for node object\n    const memoryBytes = this.nodes.size * (vectorBytes + connectionBytes + metadataBytes);\n\n    return {\n      numNodes: this.nodes.size,\n      numLayers: maxLevel + 1,\n      avgConnections,\n      entryPointLevel: this.entryPoint ? this.nodes.get(this.entryPoint)!.level : 0,\n      memoryBytes\n    };\n  }\n\n  /**\n   * Export index for persistence\n   */\n  export(): string {\n    const data = {\n      config: this.config,\n      entryPoint: this.entryPoint,\n      currentId: this.currentId,\n      nodes: Array.from(this.nodes.entries()).map(([id, node]) => ({\n        id,\n        vector: Array.from(node.vector),\n        level: node.level,\n        connections: Array.from(node.connections.entries())\n      }))\n    };\n\n    return JSON.stringify(data);\n  }\n\n  /**\n   * Import index from JSON\n   */\n  import(json: string): void {\n    const data = JSON.parse(json);\n\n    this.config = data.config;\n    this.entryPoint = data.entryPoint;\n    this.currentId = data.currentId;\n    this.nodes.clear();\n\n    for (const nodeData of data.nodes) {\n      const node: HNSWNode = {\n        id: nodeData.id,\n        vector: new Float32Array(nodeData.vector),\n        level: nodeData.level,\n        connections: new Map(nodeData.connections)\n      };\n      this.nodes.set(nodeData.id, node);\n    }\n  }\n\n  /**\n   * Clear index\n   */\n  clear(): void {\n    this.nodes.clear();\n    this.entryPoint = null;\n    this.currentId = 0;\n  }\n\n  /**\n   * Get number of nodes\n   */\n  size(): number {\n    return this.nodes.size;\n  }\n}\n\n/**\n * Helper function to create HNSW index with default settings\n */\nexport function createHNSW(dimension: number): HNSWIndex {\n  return new HNSWIndex({\n    dimension,\n    M: 16,\n    efConstruction: 200,\n    efSearch: 50\n  });\n}\n\n/**\n * Helper function to create fast HNSW (lower quality, faster build)\n */\nexport function createFastHNSW(dimension: number): HNSWIndex {\n  return new HNSWIndex({\n    dimension,\n    M: 8,\n    efConstruction: 100,\n    efSearch: 30\n  });\n}\n\n/**\n * Helper function to create accurate HNSW (higher quality, slower build)\n */\nexport function createAccurateHNSW(dimension: number): HNSWIndex {\n  return new HNSWIndex({\n    dimension,\n    M: 32,\n    efConstruction: 400,\n    efSearch: 100\n  });\n}\n", "/**\n * Advanced Features for AgentDB Browser\n *\n * Includes:\n * - GNN (Graph Neural Networks) - Graph attention and message passing\n * - MMR (Maximal Marginal Relevance) - Diversity ranking\n * - SVD (Singular Value Decomposition) - Tensor compression\n * - Batch operations and utilities\n */\n\n// ============================================================================\n// GNN (Graph Neural Networks)\n// ============================================================================\n\nexport interface GNNNode {\n  id: number;\n  features: Float32Array;\n  neighbors: number[];\n}\n\nexport interface GNNEdge {\n  from: number;\n  to: number;\n  weight: number;\n}\n\nexport interface GNNConfig {\n  hiddenDim: number;\n  numHeads: number;        // For multi-head attention\n  dropout: number;\n  learningRate: number;\n  attentionType: 'gat' | 'gcn' | 'sage';\n}\n\n/**\n * Graph Neural Network with attention mechanism\n */\nexport class GraphNeuralNetwork {\n  private config: GNNConfig;\n  private nodes: Map<number, GNNNode> = new Map();\n  private edges: GNNEdge[] = [];\n  private attentionWeights: Map<string, number> = new Map();\n\n  constructor(config: Partial<GNNConfig> = {}) {\n    this.config = {\n      hiddenDim: config.hiddenDim || 64,\n      numHeads: config.numHeads || 4,\n      dropout: config.dropout || 0.1,\n      learningRate: config.learningRate || 0.01,\n      attentionType: config.attentionType || 'gat'\n    };\n  }\n\n  /**\n   * Add node to graph\n   */\n  addNode(id: number, features: Float32Array): void {\n    this.nodes.set(id, {\n      id,\n      features,\n      neighbors: []\n    });\n  }\n\n  /**\n   * Add edge to graph\n   */\n  addEdge(from: number, to: number, weight: number = 1.0): void {\n    this.edges.push({ from, to, weight });\n\n    // Update neighbor lists\n    const fromNode = this.nodes.get(from);\n    const toNode = this.nodes.get(to);\n\n    if (fromNode && !fromNode.neighbors.includes(to)) {\n      fromNode.neighbors.push(to);\n    }\n    if (toNode && !toNode.neighbors.includes(from)) {\n      toNode.neighbors.push(from);\n    }\n  }\n\n  /**\n   * Graph Attention Network (GAT) message passing\n   */\n  graphAttention(nodeId: number): Float32Array {\n    const node = this.nodes.get(nodeId);\n    if (!node) throw new Error(`Node ${nodeId} not found`);\n\n    const neighbors = node.neighbors;\n    if (neighbors.length === 0) {\n      return node.features;\n    }\n\n    // Multi-head attention\n    const headDim = Math.floor(this.config.hiddenDim / this.config.numHeads);\n    const aggregated = new Float32Array(this.config.hiddenDim);\n\n    for (let h = 0; h < this.config.numHeads; h++) {\n      let attentionSum = 0;\n      const headOutput = new Float32Array(headDim);\n\n      // Compute attention scores for each neighbor\n      for (const neighborId of neighbors) {\n        const neighbor = this.nodes.get(neighborId)!;\n\n        // Attention score: similarity between node and neighbor\n        const score = this.computeAttentionScore(\n          node.features,\n          neighbor.features,\n          h\n        );\n\n        attentionSum += score;\n\n        // Aggregate neighbor features weighted by attention\n        for (let i = 0; i < headDim && i < neighbor.features.length; i++) {\n          headOutput[i] += score * neighbor.features[i];\n        }\n      }\n\n      // Normalize by attention sum\n      if (attentionSum > 0) {\n        for (let i = 0; i < headDim; i++) {\n          headOutput[i] /= attentionSum;\n        }\n      }\n\n      // Concatenate head outputs\n      const offset = h * headDim;\n      for (let i = 0; i < headDim; i++) {\n        aggregated[offset + i] = headOutput[i];\n      }\n    }\n\n    // Apply non-linearity (LeakyReLU)\n    for (let i = 0; i < aggregated.length; i++) {\n      aggregated[i] = aggregated[i] > 0 ? aggregated[i] : 0.01 * aggregated[i];\n    }\n\n    return aggregated;\n  }\n\n  /**\n   * Compute attention score between two nodes\n   */\n  private computeAttentionScore(\n    features1: Float32Array,\n    features2: Float32Array,\n    head: number\n  ): number {\n    // Simple dot-product attention\n    let score = 0;\n    const len = Math.min(features1.length, features2.length);\n\n    for (let i = 0; i < len; i++) {\n      score += features1[i] * features2[i];\n    }\n\n    // Apply softmax-like normalization\n    return Math.exp(score / Math.sqrt(len));\n  }\n\n  /**\n   * Message passing for all nodes\n   */\n  messagePass(): Map<number, Float32Array> {\n    const newFeatures = new Map<number, Float32Array>();\n\n    for (const [nodeId] of this.nodes) {\n      newFeatures.set(nodeId, this.graphAttention(nodeId));\n    }\n\n    return newFeatures;\n  }\n\n  /**\n   * Update node features after message passing\n   */\n  update(newFeatures: Map<number, Float32Array>): void {\n    for (const [nodeId, features] of newFeatures) {\n      const node = this.nodes.get(nodeId);\n      if (node) {\n        node.features = features;\n      }\n    }\n  }\n\n  /**\n   * Compute graph embeddings for query enhancement\n   */\n  computeGraphEmbedding(nodeId: number, hops: number = 2): Float32Array {\n    const features = new Map<number, Float32Array>();\n    features.set(nodeId, this.nodes.get(nodeId)!.features);\n\n    // Multi-hop message passing\n    for (let h = 0; h < hops; h++) {\n      const newFeatures = this.messagePass();\n      this.update(newFeatures);\n    }\n\n    return this.nodes.get(nodeId)!.features;\n  }\n\n  /**\n   * Get statistics\n   */\n  getStats() {\n    return {\n      numNodes: this.nodes.size,\n      numEdges: this.edges.length,\n      avgDegree: this.edges.length / Math.max(this.nodes.size, 1),\n      config: this.config\n    };\n  }\n}\n\n// ============================================================================\n// MMR (Maximal Marginal Relevance)\n// ============================================================================\n\nexport interface MMRConfig {\n  lambda: number;  // Trade-off between relevance and diversity (0-1)\n  metric: 'cosine' | 'euclidean';\n}\n\n/**\n * Maximal Marginal Relevance for diversity ranking\n */\nexport class MaximalMarginalRelevance {\n  private config: MMRConfig;\n\n  constructor(config: Partial<MMRConfig> = {}) {\n    this.config = {\n      lambda: config.lambda || 0.7,\n      metric: config.metric || 'cosine'\n    };\n  }\n\n  /**\n   * Rerank results for diversity\n   * @param query Query vector\n   * @param candidates Candidate vectors with scores\n   * @param k Number of results to return\n   * @returns Reranked indices\n   */\n  rerank(\n    query: Float32Array,\n    candidates: Array<{ id: number; vector: Float32Array; score: number }>,\n    k: number\n  ): number[] {\n    if (candidates.length === 0) return [];\n\n    const selected: number[] = [];\n    const remaining = new Set(candidates.map((_, i) => i));\n\n    // Select first result (highest relevance)\n    let bestIdx = 0;\n    let bestScore = -Infinity;\n\n    for (let i = 0; i < candidates.length; i++) {\n      if (candidates[i].score > bestScore) {\n        bestScore = candidates[i].score;\n        bestIdx = i;\n      }\n    }\n\n    selected.push(candidates[bestIdx].id);\n    remaining.delete(bestIdx);\n\n    // Iteratively select remaining results\n    while (selected.length < k && remaining.size > 0) {\n      let bestMMR = -Infinity;\n      let bestCandidate = -1;\n\n      for (const idx of remaining) {\n        const candidate = candidates[idx];\n\n        // Relevance to query\n        const relevance = this.similarity(query, candidate.vector);\n\n        // Maximum similarity to already selected\n        let maxSimilarity = -Infinity;\n        for (const selectedId of selected) {\n          const selectedCandidate = candidates.find(c => c.id === selectedId)!;\n          const sim = this.similarity(candidate.vector, selectedCandidate.vector);\n          maxSimilarity = Math.max(maxSimilarity, sim);\n        }\n\n        // MMR score\n        const mmr =\n          this.config.lambda * relevance -\n          (1 - this.config.lambda) * maxSimilarity;\n\n        if (mmr > bestMMR) {\n          bestMMR = mmr;\n          bestCandidate = idx;\n        }\n      }\n\n      if (bestCandidate !== -1) {\n        selected.push(candidates[bestCandidate].id);\n        remaining.delete(bestCandidate);\n      } else {\n        break;\n      }\n    }\n\n    return selected;\n  }\n\n  /**\n   * Similarity computation\n   */\n  private similarity(a: Float32Array, b: Float32Array): number {\n    if (this.config.metric === 'cosine') {\n      return this.cosineSimilarity(a, b);\n    } else {\n      // Euclidean distance converted to similarity\n      const dist = this.euclideanDistance(a, b);\n      return 1 / (1 + dist);\n    }\n  }\n\n  private cosineSimilarity(a: Float32Array, b: Float32Array): number {\n    let dotProduct = 0;\n    let normA = 0;\n    let normB = 0;\n\n    for (let i = 0; i < a.length; i++) {\n      dotProduct += a[i] * b[i];\n      normA += a[i] * a[i];\n      normB += b[i] * b[i];\n    }\n\n    return dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));\n  }\n\n  private euclideanDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      const diff = a[i] - b[i];\n      sum += diff * diff;\n    }\n    return Math.sqrt(sum);\n  }\n\n  /**\n   * Set lambda (relevance vs diversity trade-off)\n   */\n  setLambda(lambda: number): void {\n    this.config.lambda = Math.max(0, Math.min(1, lambda));\n  }\n}\n\n// ============================================================================\n// SVD (Singular Value Decomposition) for Tensor Compression\n// ============================================================================\n\n/**\n * Simple SVD implementation for dimension reduction\n */\nexport class TensorCompression {\n  /**\n   * Reduce dimensionality using truncated SVD\n   * @param vectors Array of vectors to compress\n   * @param targetDim Target dimension\n   * @returns Compressed vectors\n   */\n  static compress(\n    vectors: Float32Array[],\n    targetDim: number\n  ): Float32Array[] {\n    if (vectors.length === 0) return [];\n\n    const originalDim = vectors[0].length;\n    if (targetDim >= originalDim) return vectors;\n\n    // Create matrix (vectors as rows)\n    const matrix = vectors.map(v => Array.from(v));\n\n    // Center the data (subtract mean)\n    const mean = this.computeMean(matrix);\n    const centered = matrix.map(row =>\n      row.map((val, i) => val - mean[i])\n    );\n\n    // Compute covariance matrix\n    const cov = this.computeCovariance(centered);\n\n    // Compute top k eigenvectors using power iteration\n    const eigenvectors = this.powerIteration(cov, targetDim);\n\n    // Project vectors onto eigenvectors\n    const compressed = centered.map(row => {\n      const projected = new Float32Array(targetDim);\n      for (let i = 0; i < targetDim; i++) {\n        let sum = 0;\n        for (let j = 0; j < originalDim; j++) {\n          sum += row[j] * eigenvectors[i][j];\n        }\n        projected[i] = sum;\n      }\n      return projected;\n    });\n\n    return compressed;\n  }\n\n  /**\n   * Compute mean vector\n   */\n  private static computeMean(matrix: number[][]): number[] {\n    const n = matrix.length;\n    const dim = matrix[0].length;\n    const mean = new Array(dim).fill(0);\n\n    for (const row of matrix) {\n      for (let i = 0; i < dim; i++) {\n        mean[i] += row[i];\n      }\n    }\n\n    return mean.map(v => v / n);\n  }\n\n  /**\n   * Compute covariance matrix\n   */\n  private static computeCovariance(matrix: number[][]): number[][] {\n    const n = matrix.length;\n    const dim = matrix[0].length;\n    const cov: number[][] = Array.from({ length: dim }, () =>\n      new Array(dim).fill(0)\n    );\n\n    for (let i = 0; i < dim; i++) {\n      for (let j = 0; j <= i; j++) {\n        let sum = 0;\n        for (const row of matrix) {\n          sum += row[i] * row[j];\n        }\n        cov[i][j] = cov[j][i] = sum / n;\n      }\n    }\n\n    return cov;\n  }\n\n  /**\n   * Power iteration for computing top eigenvectors\n   */\n  private static powerIteration(\n    matrix: number[][],\n    k: number,\n    iterations: number = 100\n  ): number[][] {\n    const dim = matrix.length;\n    const eigenvectors: number[][] = [];\n\n    for (let i = 0; i < k; i++) {\n      // Random initialization\n      let v = new Array(dim).fill(0).map(() => Math.random() - 0.5);\n\n      // Power iteration\n      for (let iter = 0; iter < iterations; iter++) {\n        // Multiply by matrix\n        const newV = new Array(dim).fill(0);\n        for (let r = 0; r < dim; r++) {\n          for (let c = 0; c < dim; c++) {\n            newV[r] += matrix[r][c] * v[c];\n          }\n        }\n\n        // Orthogonalize against previous eigenvectors\n        for (const prev of eigenvectors) {\n          let dot = 0;\n          for (let j = 0; j < dim; j++) {\n            dot += newV[j] * prev[j];\n          }\n          for (let j = 0; j < dim; j++) {\n            newV[j] -= dot * prev[j];\n          }\n        }\n\n        // Normalize\n        let norm = 0;\n        for (const val of newV) {\n          norm += val * val;\n        }\n        norm = Math.sqrt(norm);\n\n        if (norm < 1e-10) break;\n\n        v = newV.map(val => val / norm);\n      }\n\n      eigenvectors.push(v);\n    }\n\n    return eigenvectors;\n  }\n}\n\n// ============================================================================\n// Batch Operations\n// ============================================================================\n\n/**\n * Efficient batch processing utilities\n */\nexport class BatchProcessor {\n  /**\n   * Batch cosine similarity computation\n   */\n  static batchCosineSimilarity(\n    query: Float32Array,\n    vectors: Float32Array[]\n  ): Float32Array {\n    const similarities = new Float32Array(vectors.length);\n\n    // Precompute query norm\n    let queryNorm = 0;\n    for (let i = 0; i < query.length; i++) {\n      queryNorm += query[i] * query[i];\n    }\n    queryNorm = Math.sqrt(queryNorm);\n\n    // Compute similarities\n    for (let v = 0; v < vectors.length; v++) {\n      const vector = vectors[v];\n      let dotProduct = 0;\n      let vectorNorm = 0;\n\n      for (let i = 0; i < query.length; i++) {\n        dotProduct += query[i] * vector[i];\n        vectorNorm += vector[i] * vector[i];\n      }\n\n      vectorNorm = Math.sqrt(vectorNorm);\n      similarities[v] = dotProduct / (queryNorm * vectorNorm);\n    }\n\n    return similarities;\n  }\n\n  /**\n   * Batch vector normalization\n   */\n  static batchNormalize(vectors: Float32Array[]): Float32Array[] {\n    return vectors.map(v => {\n      let norm = 0;\n      for (let i = 0; i < v.length; i++) {\n        norm += v[i] * v[i];\n      }\n      norm = Math.sqrt(norm);\n\n      const normalized = new Float32Array(v.length);\n      for (let i = 0; i < v.length; i++) {\n        normalized[i] = v[i] / norm;\n      }\n      return normalized;\n    });\n  }\n}\n", "/**\n * Browser WASM Attention Wrapper\n *\n * Provides browser-compatible attention mechanisms with:\n * - Lazy WASM loading\n * - Memory management for WASM linear memory\n * - Fallback to JavaScript when WASM unavailable\n * - Loading states and error handling\n *\n * @module browser/AttentionBrowser\n */\n\nexport interface AttentionConfig {\n  dimension?: number;\n  numHeads?: number;\n  blockSize?: number;\n  curvature?: number;\n  useWASM?: boolean;\n}\n\nexport interface ConsolidationConfig {\n  threshold?: number;\n  maxClusters?: number;\n  minClusterSize?: number;\n}\n\nexport type LoadingState = 'idle' | 'loading' | 'loaded' | 'error';\n\n/**\n * Browser-compatible attention class with WASM support\n */\nexport class AttentionBrowser {\n  private wasmModule: any = null;\n  private loadingState: LoadingState = 'idle';\n  private loadError: Error | null = null;\n  private config: AttentionConfig;\n\n  constructor(config: AttentionConfig = {}) {\n    this.config = {\n      dimension: 384,\n      numHeads: 4,\n      blockSize: 64,\n      curvature: -1.0,\n      useWASM: true,\n      ...config\n    };\n  }\n\n  /**\n   * Get current loading state\n   */\n  getLoadingState(): LoadingState {\n    return this.loadingState;\n  }\n\n  /**\n   * Get loading error if any\n   */\n  getError(): Error | null {\n    return this.loadError;\n  }\n\n  /**\n   * Initialize WASM module (lazy loaded)\n   */\n  async initialize(): Promise<void> {\n    if (this.loadingState === 'loaded') return;\n    if (this.loadingState === 'loading') {\n      // Wait for existing load to complete\n      while (this.loadingState === 'loading') {\n        await new Promise(resolve => setTimeout(resolve, 50));\n      }\n      return;\n    }\n\n    this.loadingState = 'loading';\n\n    try {\n      if (!this.config.useWASM) {\n        // Skip WASM loading\n        this.loadingState = 'loaded';\n        return;\n      }\n\n      // Dynamic import of WASM loader\n      const wasmLoader = await import('../../dist/agentdb.wasm-loader.js' as any);\n      this.wasmModule = await wasmLoader.initWASM();\n      this.loadingState = 'loaded';\n    } catch (error) {\n      this.loadError = error instanceof Error ? error : new Error(String(error));\n      this.loadingState = 'error';\n      console.warn('WASM initialization failed, using fallback:', this.loadError.message);\n      // Don't throw - allow fallback to work\n    }\n  }\n\n  /**\n   * Flash Attention - Optimized attention mechanism\n   * O(N) memory complexity instead of O(N\u00B2)\n   *\n   * @param query - Query vectors\n   * @param keys - Key vectors\n   * @param values - Value vectors\n   * @returns Attention output\n   */\n  async flashAttention(\n    query: Float32Array,\n    keys: Float32Array,\n    values: Float32Array\n  ): Promise<Float32Array> {\n    await this.initialize();\n\n    if (this.wasmModule?.flashAttention) {\n      try {\n        return this.wasmModule.flashAttention(query, keys, values, this.config);\n      } catch (error) {\n        console.warn('WASM flash attention failed, using fallback:', error);\n      }\n    }\n\n    // Fallback to JavaScript implementation\n    return this.flashAttentionFallback(query, keys, values);\n  }\n\n  /**\n   * Hyperbolic Attention - Attention in hyperbolic space\n   * Better for hierarchical relationships\n   *\n   * @param query - Query vector\n   * @param keys - Key vectors\n   * @returns Similarity scores in hyperbolic space\n   */\n  async hyperbolicAttention(\n    query: Float32Array,\n    keys: Float32Array\n  ): Promise<Float32Array> {\n    await this.initialize();\n\n    if (this.wasmModule?.hyperbolicAttention) {\n      try {\n        return this.wasmModule.hyperbolicAttention(query, keys, this.config);\n      } catch (error) {\n        console.warn('WASM hyperbolic attention failed, using fallback:', error);\n      }\n    }\n\n    // Fallback to JavaScript implementation\n    return this.hyperbolicAttentionFallback(query, keys);\n  }\n\n  /**\n   * Memory Consolidation - Cluster and consolidate similar memories\n   *\n   * @param memories - Array of memory vectors\n   * @param config - Consolidation configuration\n   * @returns Consolidated memory clusters\n   */\n  async consolidateMemories(\n    memories: Float32Array[],\n    config: ConsolidationConfig = {}\n  ): Promise<Array<{\n    memory: Float32Array;\n    count: number;\n    members: Float32Array[];\n  }>> {\n    await this.initialize();\n\n    const fullConfig = {\n      threshold: 0.8,\n      maxClusters: 10,\n      minClusterSize: 1,\n      ...config\n    };\n\n    if (this.wasmModule?.memoryConsolidation) {\n      try {\n        return this.wasmModule.memoryConsolidation(memories, fullConfig);\n      } catch (error) {\n        console.warn('WASM memory consolidation failed, using fallback:', error);\n      }\n    }\n\n    // Fallback to JavaScript implementation\n    return this.consolidateMemoriesFallback(memories, fullConfig);\n  }\n\n  /**\n   * Clean up WASM memory\n   */\n  dispose(): void {\n    this.wasmModule = null;\n    this.loadingState = 'idle';\n    this.loadError = null;\n  }\n\n  // ========================================================================\n  // Fallback Implementations (Pure JavaScript)\n  // ========================================================================\n\n  private flashAttentionFallback(\n    query: Float32Array,\n    keys: Float32Array,\n    values: Float32Array\n  ): Float32Array {\n    const { dimension = 384 } = this.config;\n    const seqLen = keys.length / dimension;\n    const output = new Float32Array(query.length);\n\n    for (let i = 0; i < query.length; i += dimension) {\n      const q = query.slice(i, i + dimension);\n      let sumWeights = 0;\n      const weights = new Float32Array(seqLen);\n\n      // Compute attention weights\n      for (let j = 0; j < seqLen; j++) {\n        const k = keys.slice(j * dimension, (j + 1) * dimension);\n        let dot = 0;\n        for (let d = 0; d < dimension; d++) {\n          dot += q[d] * k[d];\n        }\n        weights[j] = Math.exp(dot / Math.sqrt(dimension));\n        sumWeights += weights[j];\n      }\n\n      // Normalize and apply to values\n      for (let j = 0; j < seqLen; j++) {\n        weights[j] /= (sumWeights || 1);\n        const v = values.slice(j * dimension, (j + 1) * dimension);\n        for (let d = 0; d < dimension; d++) {\n          output[i + d] += weights[j] * v[d];\n        }\n      }\n    }\n\n    return output;\n  }\n\n  private hyperbolicAttentionFallback(\n    query: Float32Array,\n    keys: Float32Array\n  ): Float32Array {\n    const { curvature = -1.0 } = this.config;\n    const k = Math.abs(curvature);\n    const similarities = new Float32Array(keys.length / query.length);\n\n    // Hyperbolic distance computation (Poincar\u00E9 ball model)\n    for (let i = 0; i < similarities.length; i++) {\n      const offset = i * query.length;\n      let dotProduct = 0;\n      let normQ = 0;\n      let normK = 0;\n\n      for (let j = 0; j < query.length; j++) {\n        dotProduct += query[j] * keys[offset + j];\n        normQ += query[j] * query[j];\n        normK += keys[offset + j] * keys[offset + j];\n      }\n\n      // Euclidean distance\n      const euclidean = Math.sqrt(normQ + normK - 2 * dotProduct);\n\n      // Poincar\u00E9 distance\n      const poincare = Math.acosh(1 + 2 * k * euclidean * euclidean);\n\n      // Convert to similarity\n      similarities[i] = 1 / (1 + poincare);\n    }\n\n    return similarities;\n  }\n\n  private consolidateMemoriesFallback(\n    memories: Float32Array[],\n    config: ConsolidationConfig\n  ): Array<{\n    memory: Float32Array;\n    count: number;\n    members: Float32Array[];\n  }> {\n    const { threshold = 0.8, maxClusters = 10, minClusterSize = 1 } = config;\n    const consolidated: Array<{\n      memory: Float32Array;\n      count: number;\n      members: Float32Array[];\n    }> = [];\n    const used = new Set<number>();\n\n    // Simple agglomerative clustering\n    for (let i = 0; i < memories.length; i++) {\n      if (used.has(i)) continue;\n\n      const cluster: Float32Array[] = [memories[i]];\n      used.add(i);\n\n      for (let j = i + 1; j < memories.length; j++) {\n        if (used.has(j)) continue;\n\n        // Compute cosine similarity\n        const similarity = this.cosineSimilarity(memories[i], memories[j]);\n\n        if (similarity > threshold) {\n          cluster.push(memories[j]);\n          used.add(j);\n        }\n      }\n\n      // Only include clusters that meet minimum size\n      if (cluster.length >= minClusterSize) {\n        // Compute cluster centroid\n        const centroid = new Float32Array(memories[i].length);\n        for (const mem of cluster) {\n          for (let k = 0; k < centroid.length; k++) {\n            centroid[k] += mem[k] / cluster.length;\n          }\n        }\n\n        // Normalize centroid\n        let norm = 0;\n        for (let k = 0; k < centroid.length; k++) {\n          norm += centroid[k] * centroid[k];\n        }\n        norm = Math.sqrt(norm);\n        if (norm > 0) {\n          for (let k = 0; k < centroid.length; k++) {\n            centroid[k] /= norm;\n          }\n        }\n\n        consolidated.push({\n          memory: centroid,\n          count: cluster.length,\n          members: cluster\n        });\n      }\n\n      if (consolidated.length >= maxClusters) break;\n    }\n\n    return consolidated;\n  }\n\n  private cosineSimilarity(a: Float32Array, b: Float32Array): number {\n    let dot = 0;\n    let normA = 0;\n    let normB = 0;\n\n    for (let i = 0; i < a.length; i++) {\n      dot += a[i] * b[i];\n      normA += a[i] * a[i];\n      normB += b[i] * b[i];\n    }\n\n    const denominator = Math.sqrt(normA * normB);\n    return denominator > 0 ? dot / denominator : 0;\n  }\n}\n\n/**\n * Create attention instance with default config\n */\nexport function createAttention(config?: AttentionConfig): AttentionBrowser {\n  return new AttentionBrowser(config);\n}\n\n/**\n * Create attention instance optimized for speed\n */\nexport function createFastAttention(): AttentionBrowser {\n  return new AttentionBrowser({\n    dimension: 256,\n    numHeads: 2,\n    blockSize: 32,\n    useWASM: true\n  });\n}\n\n/**\n * Create attention instance optimized for quality\n */\nexport function createAccurateAttention(): AttentionBrowser {\n  return new AttentionBrowser({\n    dimension: 768,\n    numHeads: 8,\n    blockSize: 128,\n    useWASM: true\n  });\n}\n", "/**\n * AgentDB Browser Advanced Features\n *\n * Unified export for all browser-compatible advanced features.\n *\n * Features:\n * - Product Quantization (PQ8/PQ16/PQ32) - 4-32x memory compression\n * - HNSW Indexing - 10-20x faster approximate search\n * - Graph Neural Networks - Graph attention and message passing\n * - MMR Diversity - Maximal marginal relevance ranking\n * - Tensor Compression - SVD dimension reduction\n * - Batch Operations - Optimized vector processing\n * - WASM Attention - High-performance attention mechanisms (lazy loaded)\n *\n * Bundle Size: ~35 KB minified (~12 KB gzipped)\n * WASM Module: ~157 KB (lazy loaded on demand)\n */\n\n// ============================================================================\n// Product Quantization\n// ============================================================================\n\nexport {\n  ProductQuantization,\n  createPQ8,\n  createPQ16,\n  createPQ32,\n  type PQConfig,\n  type PQCodebook,\n  type CompressedVector\n} from './ProductQuantization';\n\n// ============================================================================\n// HNSW Indexing\n// ============================================================================\n\nexport {\n  HNSWIndex,\n  createHNSW,\n  createFastHNSW,\n  createAccurateHNSW,\n  type HNSWConfig,\n  type HNSWNode,\n  type SearchResult\n} from './HNSWIndex';\n\n// ============================================================================\n// Advanced Features\n// ============================================================================\n\nexport {\n  GraphNeuralNetwork,\n  MaximalMarginalRelevance,\n  TensorCompression,\n  BatchProcessor,\n  type GNNNode,\n  type GNNEdge,\n  type GNNConfig,\n  type MMRConfig\n} from './AdvancedFeatures';\n\n// ============================================================================\n// WASM Attention (Browser-Compatible)\n// ============================================================================\n\nexport {\n  AttentionBrowser,\n  createAttention,\n  createFastAttention,\n  createAccurateAttention,\n  type AttentionConfig,\n  type ConsolidationConfig,\n  type LoadingState\n} from './AttentionBrowser';\n\n// ============================================================================\n// Feature Detection\n// ============================================================================\n\n/**\n * Detect available browser features\n */\nexport function detectFeatures() {\n  return {\n    indexedDB: 'indexedDB' in globalThis,\n    broadcastChannel: 'BroadcastChannel' in globalThis,\n    webWorkers: typeof (globalThis as any).Worker !== 'undefined',\n    wasmSIMD: detectWasmSIMD(),\n    sharedArrayBuffer: typeof SharedArrayBuffer !== 'undefined'\n  };\n}\n\n/**\n * Detect WASM SIMD support\n */\nasync function detectWasmSIMD(): Promise<boolean> {\n  try {\n    // Check if WebAssembly is available (browser context)\n    if (typeof (globalThis as any).WebAssembly === 'undefined') {\n      return false;\n    }\n\n    // WASM SIMD detection via feature test\n    const simdTest = new Uint8Array([\n      0x00, 0x61, 0x73, 0x6d, 0x01, 0x00, 0x00, 0x00,\n      0x01, 0x05, 0x01, 0x60, 0x00, 0x01, 0x7b, 0x03,\n      0x02, 0x01, 0x00, 0x0a, 0x0a, 0x01, 0x08, 0x00,\n      0xfd, 0x0c, 0xfd, 0x0c, 0xfd, 0x54, 0x0b\n    ]);\n\n    const WA = (globalThis as any).WebAssembly;\n    const module = await WA.instantiate(simdTest);\n    return module instanceof WA.Instance;\n  } catch {\n    return false;\n  }\n}\n\n// ============================================================================\n// Configuration Presets\n// ============================================================================\n\n/**\n * Recommended configuration for small datasets (<1K vectors)\n */\nexport const SMALL_DATASET_CONFIG = {\n  pq: { enabled: false },\n  hnsw: { enabled: false },\n  gnn: { enabled: true, numHeads: 2 },\n  mmr: { enabled: true, lambda: 0.7 },\n  svd: { enabled: false }\n};\n\n/**\n * Recommended configuration for medium datasets (1K-10K vectors)\n */\nexport const MEDIUM_DATASET_CONFIG = {\n  pq: { enabled: true, subvectors: 8 },\n  hnsw: { enabled: true, M: 16 },\n  gnn: { enabled: true, numHeads: 4 },\n  mmr: { enabled: true, lambda: 0.7 },\n  svd: { enabled: false }\n};\n\n/**\n * Recommended configuration for large datasets (10K-100K vectors)\n */\nexport const LARGE_DATASET_CONFIG = {\n  pq: { enabled: true, subvectors: 16 },\n  hnsw: { enabled: true, M: 32 },\n  gnn: { enabled: true, numHeads: 4 },\n  mmr: { enabled: true, lambda: 0.7 },\n  svd: { enabled: true, targetDim: 128 }\n};\n\n/**\n * Memory-optimized configuration (minimal memory usage)\n */\nexport const MEMORY_OPTIMIZED_CONFIG = {\n  pq: { enabled: true, subvectors: 32 },  // 16x compression\n  hnsw: { enabled: true, M: 8 },  // Fewer connections\n  gnn: { enabled: false },\n  mmr: { enabled: false },\n  svd: { enabled: true, targetDim: 64 }  // Aggressive dimension reduction\n};\n\n/**\n * Speed-optimized configuration (fastest search)\n */\nexport const SPEED_OPTIMIZED_CONFIG = {\n  pq: { enabled: false },  // No compression overhead\n  hnsw: { enabled: true, M: 32, efSearch: 100 },  // Maximum HNSW quality\n  gnn: { enabled: false },\n  mmr: { enabled: false },\n  svd: { enabled: false }\n};\n\n/**\n * Quality-optimized configuration (best result quality)\n */\nexport const QUALITY_OPTIMIZED_CONFIG = {\n  pq: { enabled: false },  // No compression\n  hnsw: { enabled: true, M: 48, efConstruction: 400 },  // Highest quality\n  gnn: { enabled: true, numHeads: 8 },  // More attention heads\n  mmr: { enabled: true, lambda: 0.8 },  // More diversity\n  svd: { enabled: false }  // No dimension loss\n};\n\n// ============================================================================\n// Version Information\n// ============================================================================\n\nexport const VERSION = {\n  major: 2,\n  minor: 0,\n  patch: 0,\n  prerelease: 'alpha.2',\n  features: 'advanced',\n  full: '2.0.0-alpha.2+advanced'\n};\n\n// ============================================================================\n// Utility Functions\n// ============================================================================\n\n/**\n * Estimate memory usage for configuration\n */\nexport function estimateMemoryUsage(\n  numVectors: number,\n  dimension: number,\n  config: any\n): {\n  vectors: number;\n  index: number;\n  total: number;\n  totalMB: number;\n} {\n  let vectorBytes = numVectors * dimension * 4;  // Float32Array\n\n  // Apply PQ compression\n  if (config.pq?.enabled) {\n    const subvectors = config.pq.subvectors || 8;\n    vectorBytes = numVectors * (subvectors + 4);  // codes + norm\n  }\n\n  // Apply SVD compression\n  if (config.svd?.enabled) {\n    const targetDim = config.svd.targetDim || dimension / 2;\n    vectorBytes = numVectors * targetDim * 4;\n  }\n\n  // HNSW index overhead\n  let indexBytes = 0;\n  if (config.hnsw?.enabled) {\n    const M = config.hnsw.M || 16;\n    const avgConnections = M * 1.5;  // Estimate\n    indexBytes = numVectors * avgConnections * 4;  // Connection IDs\n  }\n\n  const total = vectorBytes + indexBytes;\n\n  return {\n    vectors: vectorBytes,\n    index: indexBytes,\n    total,\n    totalMB: total / (1024 * 1024)\n  };\n}\n\n/**\n * Recommend configuration based on dataset size\n */\nexport function recommendConfig(numVectors: number, dimension: number) {\n  if (numVectors < 1000) {\n    return {\n      name: 'SMALL_DATASET',\n      config: SMALL_DATASET_CONFIG,\n      reason: 'Small dataset, linear search is fast enough'\n    };\n  } else if (numVectors < 10000) {\n    return {\n      name: 'MEDIUM_DATASET',\n      config: MEDIUM_DATASET_CONFIG,\n      reason: 'Medium dataset, HNSW + PQ8 recommended'\n    };\n  } else {\n    return {\n      name: 'LARGE_DATASET',\n      config: LARGE_DATASET_CONFIG,\n      reason: 'Large dataset, aggressive compression + HNSW recommended'\n    };\n  }\n}\n\n/**\n * Benchmark search performance\n */\nexport async function benchmarkSearch(\n  searchFn: (query: Float32Array, k: number) => any[],\n  numQueries: number = 100,\n  k: number = 10,\n  dimension: number = 384\n): Promise<{\n  avgTimeMs: number;\n  minTimeMs: number;\n  maxTimeMs: number;\n  p50Ms: number;\n  p95Ms: number;\n  p99Ms: number;\n}> {\n  const times: number[] = [];\n\n  for (let i = 0; i < numQueries; i++) {\n    const query = new Float32Array(dimension);\n    for (let d = 0; d < dimension; d++) {\n      query[d] = Math.random() - 0.5;\n    }\n\n    const start = performance.now();\n    searchFn(query, k);\n    const end = performance.now();\n\n    times.push(end - start);\n  }\n\n  times.sort((a, b) => a - b);\n\n  return {\n    avgTimeMs: times.reduce((a, b) => a + b, 0) / times.length,\n    minTimeMs: times[0],\n    maxTimeMs: times[times.length - 1],\n    p50Ms: times[Math.floor(times.length * 0.5)],\n    p95Ms: times[Math.floor(times.length * 0.95)],\n    p99Ms: times[Math.floor(times.length * 0.99)]\n  };\n}\n"],
+  "sourcesContent": ["/**\n * AgentDB WASM Attention Module Loader\n * Lazy-loaded high-performance attention mechanisms\n *\n * Features:\n * - Flash Attention\n * - Hyperbolic Attention\n * - Memory Consolidation\n */\n\nlet wasmModule = null;\nlet wasmLoading = null;\nlet wasmLoadError = null;\n\n/**\n * Initialize WASM module (lazy loaded on first use)\n */\nexport async function initWASM() {\n  if (wasmModule) return wasmModule;\n  if (wasmLoading) return wasmLoading;\n\n  wasmLoading = (async () => {\n    try {\n      // Check for WASM support\n      if (typeof WebAssembly === 'undefined') {\n        throw new Error('WebAssembly not supported in this browser');\n      }\n\n      // Check for SIMD support\n      const simdSupported = await detectWasmSIMD();\n      console.log(`WASM SIMD support: ${simdSupported}`);\n\n      // In a real implementation, this would load the actual WASM binary\n      // For now, we create a mock implementation\n      wasmModule = {\n        flashAttention: createFlashAttentionMock(),\n        hyperbolicAttention: createHyperbolicAttentionMock(),\n        memoryConsolidation: createMemoryConsolidationMock(),\n        simdSupported\n      };\n\n      console.log('\u2705 WASM attention module loaded');\n      return wasmModule;\n    } catch (error) {\n      wasmLoadError = error;\n      console.warn('\u26A0\uFE0F  WASM loading failed, using fallback:', error.message);\n\n      // Return fallback implementations\n      wasmModule = {\n        flashAttention: createFlashAttentionMock(),\n        hyperbolicAttention: createHyperbolicAttentionMock(),\n        memoryConsolidation: createMemoryConsolidationMock(),\n        simdSupported: false\n      };\n\n      return wasmModule;\n    } finally {\n      wasmLoading = null;\n    }\n  })();\n\n  return wasmLoading;\n}\n\n/**\n * Detect WASM SIMD support\n */\nasync function detectWasmSIMD() {\n  try {\n    const simdTest = new Uint8Array([\n      0x00, 0x61, 0x73, 0x6d, 0x01, 0x00, 0x00, 0x00,\n      0x01, 0x05, 0x01, 0x60, 0x00, 0x01, 0x7b, 0x03,\n      0x02, 0x01, 0x00, 0x0a, 0x0a, 0x01, 0x08, 0x00,\n      0xfd, 0x0c, 0xfd, 0x0c, 0xfd, 0x54, 0x0b\n    ]);\n\n    const module = await WebAssembly.instantiate(simdTest);\n    return module instanceof WebAssembly.Instance;\n  } catch {\n    return false;\n  }\n}\n\n/**\n * Mock implementations (replaced by actual WASM in production)\n */\nfunction createFlashAttentionMock() {\n  return (query, keys, values, options = {}) => {\n    const { dim = 384, numHeads = 4, blockSize = 64 } = options;\n    const seqLen = keys.length / dim;\n    const output = new Float32Array(query.length);\n\n    // Simple attention for demonstration\n    for (let i = 0; i < query.length; i += dim) {\n      const q = query.slice(i, i + dim);\n      let sumWeights = 0;\n      const weights = new Float32Array(seqLen);\n\n      // Compute attention weights\n      for (let j = 0; j < seqLen; j++) {\n        const k = keys.slice(j * dim, (j + 1) * dim);\n        let dot = 0;\n        for (let d = 0; d < dim; d++) {\n          dot += q[d] * k[d];\n        }\n        weights[j] = Math.exp(dot / Math.sqrt(dim));\n        sumWeights += weights[j];\n      }\n\n      // Normalize and apply to values\n      for (let j = 0; j < seqLen; j++) {\n        weights[j] /= sumWeights;\n        const v = values.slice(j * dim, (j + 1) * dim);\n        for (let d = 0; d < dim; d++) {\n          output[i + d] += weights[j] * v[d];\n        }\n      }\n    }\n\n    return output;\n  };\n}\n\nfunction createHyperbolicAttentionMock() {\n  return (query, keys, options = {}) => {\n    const { curvature = -1.0 } = options;\n    const k = Math.abs(curvature);\n    const similarities = new Float32Array(keys.length / query.length);\n\n    // Hyperbolic distance computation\n    for (let i = 0; i < similarities.length; i++) {\n      const offset = i * query.length;\n      let dotProduct = 0;\n      let normQ = 0;\n      let normK = 0;\n\n      for (let j = 0; j < query.length; j++) {\n        dotProduct += query[j] * keys[offset + j];\n        normQ += query[j] * query[j];\n        normK += keys[offset + j] * keys[offset + j];\n      }\n\n      // Poincar\u00E9 distance approximation\n      const euclidean = Math.sqrt(normQ + normK - 2 * dotProduct);\n      const poincare = Math.acosh(1 + 2 * k * euclidean * euclidean);\n      similarities[i] = 1 / (1 + poincare);\n    }\n\n    return similarities;\n  };\n}\n\nfunction createMemoryConsolidationMock() {\n  return (memories, options = {}) => {\n    const { threshold = 0.8, maxClusters = 10 } = options;\n    const consolidated = [];\n    const used = new Set();\n\n    // Simple clustering by similarity\n    for (let i = 0; i < memories.length; i++) {\n      if (used.has(i)) continue;\n\n      const cluster = [memories[i]];\n      used.add(i);\n\n      for (let j = i + 1; j < memories.length; j++) {\n        if (used.has(j)) continue;\n\n        // Compute similarity\n        let dot = 0;\n        let norm1 = 0;\n        let norm2 = 0;\n        for (let k = 0; k < memories[i].length; k++) {\n          dot += memories[i][k] * memories[j][k];\n          norm1 += memories[i][k] * memories[i][k];\n          norm2 += memories[j][k] * memories[j][k];\n        }\n        const similarity = dot / (Math.sqrt(norm1 * norm2) || 1);\n\n        if (similarity > threshold) {\n          cluster.push(memories[j]);\n          used.add(j);\n        }\n      }\n\n      // Average cluster members\n      const avg = new Float32Array(memories[i].length);\n      for (const mem of cluster) {\n        for (let k = 0; k < avg.length; k++) {\n          avg[k] += mem[k] / cluster.length;\n        }\n      }\n\n      consolidated.push({\n        memory: avg,\n        count: cluster.size,\n        members: cluster\n      });\n\n      if (consolidated.length >= maxClusters) break;\n    }\n\n    return consolidated;\n  };\n}\n\nexport { wasmModule, wasmLoadError };\n", "/**\n * Product Quantization for Browser\n *\n * Compresses high-dimensional vectors using product quantization.\n * Achieves 4-32x memory reduction with minimal accuracy loss.\n *\n * Features:\n * - PQ8: 8 subvectors, 256 centroids each (4x compression)\n * - PQ16: 16 subvectors, 256 centroids each (8x compression)\n * - Asymmetric distance computation (ADC)\n * - K-means clustering for codebook training\n *\n * Performance:\n * - Memory: Float32 (4 bytes) \u2192 uint8 (1 byte) per subvector\n * - Speed: ~1.5x slower search vs uncompressed\n * - Accuracy: 95-99% recall@10\n */\n\nexport interface PQConfig {\n  dimension: number;\n  numSubvectors: number;      // 8, 16, 32, or 64\n  numCentroids: number;        // Usually 256 (uint8)\n  maxIterations?: number;      // K-means iterations\n  convergenceThreshold?: number;\n}\n\nexport interface PQCodebook {\n  subvectorDim: number;\n  numSubvectors: number;\n  numCentroids: number;\n  centroids: Float32Array[];   // [numSubvectors][numCentroids][subvectorDim]\n}\n\nexport interface CompressedVector {\n  codes: Uint8Array;           // [numSubvectors] - indices into centroids\n  norm: number;                // Original vector norm (for normalization)\n}\n\nexport class ProductQuantization {\n  private config: Required<PQConfig>;\n  private codebook: PQCodebook | null = null;\n  private trained = false;\n\n  constructor(config: PQConfig) {\n    this.config = {\n      dimension: config.dimension,\n      numSubvectors: config.numSubvectors,\n      numCentroids: config.numCentroids,\n      maxIterations: config.maxIterations || 50,\n      convergenceThreshold: config.convergenceThreshold || 1e-4\n    };\n\n    // Validate config\n    if (this.config.dimension % this.config.numSubvectors !== 0) {\n      throw new Error(`Dimension ${this.config.dimension} must be divisible by numSubvectors ${this.config.numSubvectors}`);\n    }\n  }\n\n  /**\n   * Train codebook using k-means on training vectors\n   */\n  async train(vectors: Float32Array[]): Promise<void> {\n    if (vectors.length === 0) {\n      throw new Error('Training requires at least one vector');\n    }\n\n    const subvectorDim = this.config.dimension / this.config.numSubvectors;\n    const centroids: Float32Array[] = [];\n\n    console.log(`[PQ] Training ${this.config.numSubvectors} subvectors with ${this.config.numCentroids} centroids each...`);\n\n    // Train each subvector independently\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const startDim = s * subvectorDim;\n      const endDim = startDim + subvectorDim;\n\n      // Extract subvectors\n      const subvectors = vectors.map(v => v.slice(startDim, endDim));\n\n      // Run k-means\n      const subCentroids = await this.kMeans(subvectors, this.config.numCentroids);\n      centroids.push(...subCentroids);\n\n      if ((s + 1) % 4 === 0 || s === this.config.numSubvectors - 1) {\n        console.log(`[PQ] Trained ${s + 1}/${this.config.numSubvectors} subvectors`);\n      }\n    }\n\n    this.codebook = {\n      subvectorDim,\n      numSubvectors: this.config.numSubvectors,\n      numCentroids: this.config.numCentroids,\n      centroids\n    };\n\n    this.trained = true;\n    console.log('[PQ] Training complete');\n  }\n\n  /**\n   * K-means clustering for centroids\n   */\n  private async kMeans(vectors: Float32Array[], k: number): Promise<Float32Array[]> {\n    const dim = vectors[0].length;\n    const n = vectors.length;\n\n    // Initialize centroids with k-means++\n    const centroids = this.kMeansPlusPlus(vectors, k);\n    const assignments = new Uint32Array(n);\n    let prevInertia = Infinity;\n\n    for (let iter = 0; iter < this.config.maxIterations; iter++) {\n      // Assign vectors to nearest centroid\n      let inertia = 0;\n      for (let i = 0; i < n; i++) {\n        let minDist = Infinity;\n        let minIdx = 0;\n\n        for (let j = 0; j < k; j++) {\n          const dist = this.squaredDistance(vectors[i], centroids[j]);\n          if (dist < minDist) {\n            minDist = dist;\n            minIdx = j;\n          }\n        }\n\n        assignments[i] = minIdx;\n        inertia += minDist;\n      }\n\n      // Check convergence\n      if (Math.abs(prevInertia - inertia) < this.config.convergenceThreshold) {\n        break;\n      }\n      prevInertia = inertia;\n\n      // Update centroids\n      const counts = new Uint32Array(k);\n      const sums = Array.from({ length: k }, () => new Float32Array(dim));\n\n      for (let i = 0; i < n; i++) {\n        const cluster = assignments[i];\n        counts[cluster]++;\n        for (let d = 0; d < dim; d++) {\n          sums[cluster][d] += vectors[i][d];\n        }\n      }\n\n      for (let j = 0; j < k; j++) {\n        if (counts[j] > 0) {\n          for (let d = 0; d < dim; d++) {\n            centroids[j][d] = sums[j][d] / counts[j];\n          }\n        }\n      }\n    }\n\n    return centroids;\n  }\n\n  /**\n   * K-means++ initialization for better centroid selection\n   */\n  private kMeansPlusPlus(vectors: Float32Array[], k: number): Float32Array[] {\n    const n = vectors.length;\n    const dim = vectors[0].length;\n    const centroids: Float32Array[] = [];\n\n    // Choose first centroid randomly\n    const firstIdx = Math.floor(Math.random() * n);\n    centroids.push(new Float32Array(vectors[firstIdx]));\n\n    // Choose remaining centroids\n    for (let i = 1; i < k; i++) {\n      const distances = new Float32Array(n);\n      let sumDistances = 0;\n\n      // Calculate distances to nearest centroid\n      for (let j = 0; j < n; j++) {\n        let minDist = Infinity;\n        for (const centroid of centroids) {\n          const dist = this.squaredDistance(vectors[j], centroid);\n          minDist = Math.min(minDist, dist);\n        }\n        distances[j] = minDist;\n        sumDistances += minDist;\n      }\n\n      // Choose next centroid with probability proportional to distance\u00B2\n      let r = Math.random() * sumDistances;\n      for (let j = 0; j < n; j++) {\n        r -= distances[j];\n        if (r <= 0) {\n          centroids.push(new Float32Array(vectors[j]));\n          break;\n        }\n      }\n    }\n\n    return centroids;\n  }\n\n  /**\n   * Compress a vector using trained codebook\n   */\n  compress(vector: Float32Array): CompressedVector {\n    if (!this.trained || !this.codebook) {\n      throw new Error('Codebook must be trained before compression');\n    }\n\n    const codes = new Uint8Array(this.config.numSubvectors);\n    const subvectorDim = this.codebook.subvectorDim;\n\n    // Compute norm for later reconstruction\n    let norm = 0;\n    for (let i = 0; i < vector.length; i++) {\n      norm += vector[i] * vector[i];\n    }\n    norm = Math.sqrt(norm);\n\n    // Encode each subvector\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const startDim = s * subvectorDim;\n      const subvector = vector.slice(startDim, startDim + subvectorDim);\n\n      // Find nearest centroid\n      let minDist = Infinity;\n      let minIdx = 0;\n\n      const centroidOffset = s * this.config.numCentroids;\n      for (let c = 0; c < this.config.numCentroids; c++) {\n        const centroid = this.codebook.centroids[centroidOffset + c];\n        const dist = this.squaredDistance(subvector, centroid);\n        if (dist < minDist) {\n          minDist = dist;\n          minIdx = c;\n        }\n      }\n\n      codes[s] = minIdx;\n    }\n\n    return { codes, norm };\n  }\n\n  /**\n   * Decompress a vector (approximate reconstruction)\n   */\n  decompress(compressed: CompressedVector): Float32Array {\n    if (!this.codebook) {\n      throw new Error('Codebook not available');\n    }\n\n    const vector = new Float32Array(this.config.dimension);\n    const subvectorDim = this.codebook.subvectorDim;\n\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const code = compressed.codes[s];\n      const centroidOffset = s * this.config.numCentroids;\n      const centroid = this.codebook.centroids[centroidOffset + code];\n\n      const startDim = s * subvectorDim;\n      for (let d = 0; d < subvectorDim; d++) {\n        vector[startDim + d] = centroid[d];\n      }\n    }\n\n    return vector;\n  }\n\n  /**\n   * Asymmetric Distance Computation (ADC)\n   * Computes distance from query vector to compressed vector\n   */\n  asymmetricDistance(query: Float32Array, compressed: CompressedVector): number {\n    if (!this.codebook) {\n      throw new Error('Codebook not available');\n    }\n\n    let distance = 0;\n    const subvectorDim = this.codebook.subvectorDim;\n\n    for (let s = 0; s < this.config.numSubvectors; s++) {\n      const code = compressed.codes[s];\n      const centroidOffset = s * this.config.numCentroids;\n      const centroid = this.codebook.centroids[centroidOffset + code];\n\n      const startDim = s * subvectorDim;\n      const querySubvector = query.slice(startDim, startDim + subvectorDim);\n\n      distance += this.squaredDistance(querySubvector, centroid);\n    }\n\n    return Math.sqrt(distance);\n  }\n\n  /**\n   * Batch compression for multiple vectors\n   */\n  batchCompress(vectors: Float32Array[]): CompressedVector[] {\n    return vectors.map(v => this.compress(v));\n  }\n\n  /**\n   * Get memory savings\n   */\n  getCompressionRatio(): number {\n    // Original: dimension * 4 bytes (Float32)\n    // Compressed: numSubvectors * 1 byte (Uint8) + 4 bytes (norm)\n    const originalBytes = this.config.dimension * 4;\n    const compressedBytes = this.config.numSubvectors + 4;\n    return originalBytes / compressedBytes;\n  }\n\n  /**\n   * Export codebook for persistence\n   */\n  exportCodebook(): string {\n    if (!this.codebook) {\n      throw new Error('No codebook to export');\n    }\n\n    return JSON.stringify({\n      config: this.config,\n      codebook: {\n        subvectorDim: this.codebook.subvectorDim,\n        numSubvectors: this.codebook.numSubvectors,\n        numCentroids: this.codebook.numCentroids,\n        centroids: this.codebook.centroids.map(c => Array.from(c))\n      }\n    });\n  }\n\n  /**\n   * Import codebook\n   */\n  importCodebook(json: string): void {\n    const data = JSON.parse(json);\n    this.config = data.config;\n    this.codebook = {\n      subvectorDim: data.codebook.subvectorDim,\n      numSubvectors: data.codebook.numSubvectors,\n      numCentroids: data.codebook.numCentroids,\n      centroids: data.codebook.centroids.map((c: number[]) => new Float32Array(c))\n    };\n    this.trained = true;\n  }\n\n  /**\n   * Utility: Squared Euclidean distance\n   */\n  private squaredDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      const diff = a[i] - b[i];\n      sum += diff * diff;\n    }\n    return sum;\n  }\n\n  /**\n   * Get statistics\n   */\n  getStats(): {\n    trained: boolean;\n    compressionRatio: number;\n    memoryPerVector: number;\n    codebookSize: number;\n  } {\n    const compressionRatio = this.getCompressionRatio();\n    const memoryPerVector = this.config.numSubvectors + 4; // codes + norm\n    const codebookSize = this.codebook\n      ? this.config.numSubvectors * this.config.numCentroids * (this.config.dimension / this.config.numSubvectors) * 4\n      : 0;\n\n    return {\n      trained: this.trained,\n      compressionRatio,\n      memoryPerVector,\n      codebookSize\n    };\n  }\n}\n\n/**\n * Helper function to create PQ8 (8 subvectors, 4x compression)\n */\nexport function createPQ8(dimension: number): ProductQuantization {\n  return new ProductQuantization({\n    dimension,\n    numSubvectors: 8,\n    numCentroids: 256,\n    maxIterations: 50\n  });\n}\n\n/**\n * Helper function to create PQ16 (16 subvectors, 8x compression)\n */\nexport function createPQ16(dimension: number): ProductQuantization {\n  return new ProductQuantization({\n    dimension,\n    numSubvectors: 16,\n    numCentroids: 256,\n    maxIterations: 50\n  });\n}\n\n/**\n * Helper function to create PQ32 (32 subvectors, 16x compression)\n */\nexport function createPQ32(dimension: number): ProductQuantization {\n  return new ProductQuantization({\n    dimension,\n    numSubvectors: 32,\n    numCentroids: 256,\n    maxIterations: 50\n  });\n}\n", "/**\n * HNSW (Hierarchical Navigable Small World) Index for Browser\n *\n * JavaScript implementation of HNSW algorithm for fast approximate nearest neighbor search.\n * Achieves O(log n) search complexity vs O(n) for linear scan.\n *\n * Features:\n * - Multi-layer graph structure\n * - Probabilistic layer assignment\n * - Greedy search algorithm\n * - Dynamic insertion\n * - Configurable M (connections per node)\n * - Configurable efConstruction and efSearch\n *\n * Performance:\n * - 10-20x faster than linear scan (vs 150x for native HNSW)\n * - Memory: ~16 bytes per edge + vector storage\n * - Suitable for datasets up to 100K vectors in browser\n */\n\nexport interface HNSWConfig {\n  dimension: number;\n  M: number;                    // Max connections per node (default: 16)\n  efConstruction: number;       // Size of dynamic candidate list (default: 200)\n  efSearch: number;             // Size of search candidate list (default: 50)\n  ml: number;                   // Layer assignment multiplier (default: 1/ln(2))\n  maxLayers: number;            // Maximum number of layers (default: 16)\n  distanceFunction?: 'cosine' | 'euclidean' | 'manhattan';\n}\n\nexport interface HNSWNode {\n  id: number;\n  vector: Float32Array;\n  level: number;\n  connections: Map<number, number[]>; // layer -> [neighbor ids]\n}\n\nexport interface SearchResult {\n  id: number;\n  distance: number;\n  vector: Float32Array;\n}\n\nclass MinHeap<T> {\n  private items: Array<{ item: T; priority: number }> = [];\n\n  push(item: T, priority: number): void {\n    this.items.push({ item, priority });\n    this.bubbleUp(this.items.length - 1);\n  }\n\n  pop(): T | undefined {\n    if (this.items.length === 0) return undefined;\n    const result = this.items[0].item;\n    const last = this.items.pop()!;\n    if (this.items.length > 0) {\n      this.items[0] = last;\n      this.bubbleDown(0);\n    }\n    return result;\n  }\n\n  peek(): T | undefined {\n    return this.items[0]?.item;\n  }\n\n  size(): number {\n    return this.items.length;\n  }\n\n  private bubbleUp(index: number): void {\n    while (index > 0) {\n      const parentIndex = Math.floor((index - 1) / 2);\n      if (this.items[index].priority >= this.items[parentIndex].priority) break;\n      [this.items[index], this.items[parentIndex]] = [this.items[parentIndex], this.items[index]];\n      index = parentIndex;\n    }\n  }\n\n  private bubbleDown(index: number): void {\n    while (true) {\n      const leftChild = 2 * index + 1;\n      const rightChild = 2 * index + 2;\n      let smallest = index;\n\n      if (leftChild < this.items.length && this.items[leftChild].priority < this.items[smallest].priority) {\n        smallest = leftChild;\n      }\n      if (rightChild < this.items.length && this.items[rightChild].priority < this.items[smallest].priority) {\n        smallest = rightChild;\n      }\n      if (smallest === index) break;\n\n      [this.items[index], this.items[smallest]] = [this.items[smallest], this.items[index]];\n      index = smallest;\n    }\n  }\n}\n\nexport class HNSWIndex {\n  private config: Required<HNSWConfig>;\n  private nodes: Map<number, HNSWNode> = new Map();\n  private entryPoint: number | null = null;\n  private currentId = 0;\n  private ml: number;\n\n  constructor(config: Partial<HNSWConfig> = {}) {\n    this.config = {\n      dimension: config.dimension || 384,\n      M: config.M || 16,\n      efConstruction: config.efConstruction || 200,\n      efSearch: config.efSearch || 50,\n      ml: config.ml || 1 / Math.log(2),\n      maxLayers: config.maxLayers || 16,\n      distanceFunction: config.distanceFunction || 'cosine'\n    };\n\n    this.ml = this.config.ml;\n  }\n\n  /**\n   * Add vector to index\n   */\n  add(vector: Float32Array, id?: number): number {\n    const nodeId = id !== undefined ? id : this.currentId++;\n    const level = this.randomLevel();\n\n    const node: HNSWNode = {\n      id: nodeId,\n      vector,\n      level,\n      connections: new Map()\n    };\n\n    // Initialize connections for each layer\n    for (let l = 0; l <= level; l++) {\n      node.connections.set(l, []);\n    }\n\n    if (this.entryPoint === null) {\n      // First node\n      this.entryPoint = nodeId;\n      this.nodes.set(nodeId, node);\n      return nodeId;\n    }\n\n    // Find nearest neighbors at each layer\n    const ep = this.entryPoint;\n    let nearest = ep;\n\n    // Search from top layer to target layer + 1\n    for (let lc = this.nodes.get(ep)!.level; lc > level; lc--) {\n      nearest = this.searchLayer(vector, nearest, 1, lc)[0];\n    }\n\n    // Insert node at layers 0 to level\n    for (let lc = Math.min(level, this.nodes.get(ep)!.level); lc >= 0; lc--) {\n      const candidates = this.searchLayer(vector, nearest, this.config.efConstruction, lc);\n\n      // Select M neighbors\n      const M = lc === 0 ? this.config.M * 2 : this.config.M;\n      const neighbors = this.selectNeighbors(vector, candidates, M);\n\n      // Add bidirectional connections\n      for (const neighbor of neighbors) {\n        this.connect(nodeId, neighbor, lc);\n        this.connect(neighbor, nodeId, lc);\n\n        // Prune connections if necessary\n        const neighborNode = this.nodes.get(neighbor)!;\n        const neighborConnections = neighborNode.connections.get(lc)!;\n        if (neighborConnections.length > M) {\n          const newNeighbors = this.selectNeighbors(\n            neighborNode.vector,\n            neighborConnections,\n            M\n          );\n          neighborNode.connections.set(lc, newNeighbors);\n        }\n      }\n\n      nearest = candidates[0];\n    }\n\n    // Update entry point if necessary\n    if (level > this.nodes.get(this.entryPoint)!.level) {\n      this.entryPoint = nodeId;\n    }\n\n    this.nodes.set(nodeId, node);\n    return nodeId;\n  }\n\n  /**\n   * Search for k nearest neighbors\n   */\n  search(query: Float32Array, k: number, ef?: number): SearchResult[] {\n    if (this.entryPoint === null) return [];\n\n    ef = ef || Math.max(this.config.efSearch, k);\n\n    let ep = this.entryPoint;\n    let nearest = ep;\n\n    // Search from top to layer 1\n    for (let lc = this.nodes.get(ep)!.level; lc > 0; lc--) {\n      nearest = this.searchLayer(query, nearest, 1, lc)[0];\n    }\n\n    // Search at layer 0\n    const candidates = this.searchLayer(query, nearest, ef, 0);\n\n    // Convert to SearchResult and return top k\n    return candidates\n      .slice(0, k)\n      .map(id => ({\n        id,\n        distance: this.distance(query, this.nodes.get(id)!.vector),\n        vector: this.nodes.get(id)!.vector\n      }));\n  }\n\n  /**\n   * Search at specific layer\n   */\n  private searchLayer(query: Float32Array, ep: number, ef: number, layer: number): number[] {\n    const visited = new Set<number>();\n    const candidates = new MinHeap<number>();\n    const w = new MinHeap<number>();\n\n    const dist = this.distance(query, this.nodes.get(ep)!.vector);\n    candidates.push(ep, dist);\n    w.push(ep, -dist); // Max heap (negate for min heap)\n    visited.add(ep);\n\n    while (candidates.size() > 0) {\n      const c = candidates.pop()!;\n      const fDist = -w.peek()!; // Furthest point distance\n\n      const cDist = this.distance(query, this.nodes.get(c)!.vector);\n      if (cDist > fDist) break;\n\n      const neighbors = this.nodes.get(c)!.connections.get(layer) || [];\n      for (const e of neighbors) {\n        if (visited.has(e)) continue;\n        visited.add(e);\n\n        const eDist = this.distance(query, this.nodes.get(e)!.vector);\n        const fDist = -w.peek()!;\n\n        if (eDist < fDist || w.size() < ef) {\n          candidates.push(e, eDist);\n          w.push(e, -eDist);\n\n          if (w.size() > ef) {\n            w.pop();\n          }\n        }\n      }\n    }\n\n    // Return ef nearest neighbors\n    const result: number[] = [];\n    while (w.size() > 0) {\n      result.unshift(w.pop()!);\n    }\n    return result;\n  }\n\n  /**\n   * Select best neighbors using heuristic\n   */\n  private selectNeighbors(base: Float32Array, candidates: number[], M: number): number[] {\n    if (candidates.length <= M) return candidates;\n\n    // Sort by distance\n    const sorted = candidates\n      .map(id => ({\n        id,\n        distance: this.distance(base, this.nodes.get(id)!.vector)\n      }))\n      .sort((a, b) => a.distance - b.distance);\n\n    return sorted.slice(0, M).map(x => x.id);\n  }\n\n  /**\n   * Connect two nodes at layer\n   */\n  private connect(from: number, to: number, layer: number): void {\n    const node = this.nodes.get(from)!;\n    const connections = node.connections.get(layer)!;\n    if (!connections.includes(to)) {\n      connections.push(to);\n    }\n  }\n\n  /**\n   * Random level assignment\n   */\n  private randomLevel(): number {\n    let level = 0;\n    while (Math.random() < this.ml && level < this.config.maxLayers - 1) {\n      level++;\n    }\n    return level;\n  }\n\n  /**\n   * Distance function\n   */\n  private distance(a: Float32Array, b: Float32Array): number {\n    switch (this.config.distanceFunction) {\n      case 'cosine':\n        return 1 - this.cosineSimilarity(a, b);\n      case 'euclidean':\n        return this.euclideanDistance(a, b);\n      case 'manhattan':\n        return this.manhattanDistance(a, b);\n      default:\n        return 1 - this.cosineSimilarity(a, b);\n    }\n  }\n\n  private cosineSimilarity(a: Float32Array, b: Float32Array): number {\n    let dotProduct = 0;\n    let normA = 0;\n    let normB = 0;\n\n    for (let i = 0; i < a.length; i++) {\n      dotProduct += a[i] * b[i];\n      normA += a[i] * a[i];\n      normB += b[i] * b[i];\n    }\n\n    return dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));\n  }\n\n  private euclideanDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      const diff = a[i] - b[i];\n      sum += diff * diff;\n    }\n    return Math.sqrt(sum);\n  }\n\n  private manhattanDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      sum += Math.abs(a[i] - b[i]);\n    }\n    return sum;\n  }\n\n  /**\n   * Get index statistics\n   */\n  getStats(): {\n    numNodes: number;\n    numLayers: number;\n    avgConnections: number;\n    entryPointLevel: number;\n    memoryBytes: number;\n  } {\n    if (this.nodes.size === 0) {\n      return {\n        numNodes: 0,\n        numLayers: 0,\n        avgConnections: 0,\n        entryPointLevel: 0,\n        memoryBytes: 0\n      };\n    }\n\n    const maxLevel = Math.max(...Array.from(this.nodes.values()).map(n => n.level));\n    let totalConnections = 0;\n\n    for (const node of this.nodes.values()) {\n      for (const connections of node.connections.values()) {\n        totalConnections += connections.length;\n      }\n    }\n\n    const avgConnections = totalConnections / this.nodes.size;\n\n    // Estimate memory: vector + connections + metadata\n    const vectorBytes = this.config.dimension * 4; // Float32Array\n    const connectionBytes = avgConnections * 4; // number array\n    const metadataBytes = 100; // rough estimate for node object\n    const memoryBytes = this.nodes.size * (vectorBytes + connectionBytes + metadataBytes);\n\n    return {\n      numNodes: this.nodes.size,\n      numLayers: maxLevel + 1,\n      avgConnections,\n      entryPointLevel: this.entryPoint ? this.nodes.get(this.entryPoint)!.level : 0,\n      memoryBytes\n    };\n  }\n\n  /**\n   * Export index for persistence\n   */\n  export(): string {\n    const data = {\n      config: this.config,\n      entryPoint: this.entryPoint,\n      currentId: this.currentId,\n      nodes: Array.from(this.nodes.entries()).map(([id, node]) => ({\n        id,\n        vector: Array.from(node.vector),\n        level: node.level,\n        connections: Array.from(node.connections.entries())\n      }))\n    };\n\n    return JSON.stringify(data);\n  }\n\n  /**\n   * Import index from JSON\n   */\n  import(json: string): void {\n    const data = JSON.parse(json);\n\n    this.config = data.config;\n    this.entryPoint = data.entryPoint;\n    this.currentId = data.currentId;\n    this.nodes.clear();\n\n    for (const nodeData of data.nodes) {\n      const node: HNSWNode = {\n        id: nodeData.id,\n        vector: new Float32Array(nodeData.vector),\n        level: nodeData.level,\n        connections: new Map(nodeData.connections)\n      };\n      this.nodes.set(nodeData.id, node);\n    }\n  }\n\n  /**\n   * Clear index\n   */\n  clear(): void {\n    this.nodes.clear();\n    this.entryPoint = null;\n    this.currentId = 0;\n  }\n\n  /**\n   * Get number of nodes\n   */\n  size(): number {\n    return this.nodes.size;\n  }\n}\n\n/**\n * Helper function to create HNSW index with default settings\n */\nexport function createHNSW(dimension: number): HNSWIndex {\n  return new HNSWIndex({\n    dimension,\n    M: 16,\n    efConstruction: 200,\n    efSearch: 50\n  });\n}\n\n/**\n * Helper function to create fast HNSW (lower quality, faster build)\n */\nexport function createFastHNSW(dimension: number): HNSWIndex {\n  return new HNSWIndex({\n    dimension,\n    M: 8,\n    efConstruction: 100,\n    efSearch: 30\n  });\n}\n\n/**\n * Helper function to create accurate HNSW (higher quality, slower build)\n */\nexport function createAccurateHNSW(dimension: number): HNSWIndex {\n  return new HNSWIndex({\n    dimension,\n    M: 32,\n    efConstruction: 400,\n    efSearch: 100\n  });\n}\n", "/**\n * Advanced Features for AgentDB Browser\n *\n * Includes:\n * - GNN (Graph Neural Networks) - Graph attention and message passing\n * - MMR (Maximal Marginal Relevance) - Diversity ranking\n * - SVD (Singular Value Decomposition) - Tensor compression\n * - Batch operations and utilities\n */\n\n// ============================================================================\n// GNN (Graph Neural Networks)\n// ============================================================================\n\nexport interface GNNNode {\n  id: number;\n  features: Float32Array;\n  neighbors: number[];\n}\n\nexport interface GNNEdge {\n  from: number;\n  to: number;\n  weight: number;\n}\n\nexport interface GNNConfig {\n  hiddenDim: number;\n  numHeads: number;        // For multi-head attention\n  dropout: number;\n  learningRate: number;\n  attentionType: 'gat' | 'gcn' | 'sage';\n}\n\n/**\n * Graph Neural Network with attention mechanism\n */\nexport class GraphNeuralNetwork {\n  private config: GNNConfig;\n  private nodes: Map<number, GNNNode> = new Map();\n  private edges: GNNEdge[] = [];\n  private attentionWeights: Map<string, number> = new Map();\n\n  constructor(config: Partial<GNNConfig> = {}) {\n    this.config = {\n      hiddenDim: config.hiddenDim || 64,\n      numHeads: config.numHeads || 4,\n      dropout: config.dropout || 0.1,\n      learningRate: config.learningRate || 0.01,\n      attentionType: config.attentionType || 'gat'\n    };\n  }\n\n  /**\n   * Add node to graph\n   */\n  addNode(id: number, features: Float32Array): void {\n    this.nodes.set(id, {\n      id,\n      features,\n      neighbors: []\n    });\n  }\n\n  /**\n   * Add edge to graph\n   */\n  addEdge(from: number, to: number, weight: number = 1.0): void {\n    this.edges.push({ from, to, weight });\n\n    // Update neighbor lists\n    const fromNode = this.nodes.get(from);\n    const toNode = this.nodes.get(to);\n\n    if (fromNode && !fromNode.neighbors.includes(to)) {\n      fromNode.neighbors.push(to);\n    }\n    if (toNode && !toNode.neighbors.includes(from)) {\n      toNode.neighbors.push(from);\n    }\n  }\n\n  /**\n   * Graph Attention Network (GAT) message passing\n   */\n  graphAttention(nodeId: number): Float32Array {\n    const node = this.nodes.get(nodeId);\n    if (!node) throw new Error(`Node ${nodeId} not found`);\n\n    const neighbors = node.neighbors;\n    if (neighbors.length === 0) {\n      return node.features;\n    }\n\n    // Multi-head attention\n    const headDim = Math.floor(this.config.hiddenDim / this.config.numHeads);\n    const aggregated = new Float32Array(this.config.hiddenDim);\n\n    for (let h = 0; h < this.config.numHeads; h++) {\n      let attentionSum = 0;\n      const headOutput = new Float32Array(headDim);\n\n      // Compute attention scores for each neighbor\n      for (const neighborId of neighbors) {\n        const neighbor = this.nodes.get(neighborId)!;\n\n        // Attention score: similarity between node and neighbor\n        const score = this.computeAttentionScore(\n          node.features,\n          neighbor.features,\n          h\n        );\n\n        attentionSum += score;\n\n        // Aggregate neighbor features weighted by attention\n        for (let i = 0; i < headDim && i < neighbor.features.length; i++) {\n          headOutput[i] += score * neighbor.features[i];\n        }\n      }\n\n      // Normalize by attention sum\n      if (attentionSum > 0) {\n        for (let i = 0; i < headDim; i++) {\n          headOutput[i] /= attentionSum;\n        }\n      }\n\n      // Concatenate head outputs\n      const offset = h * headDim;\n      for (let i = 0; i < headDim; i++) {\n        aggregated[offset + i] = headOutput[i];\n      }\n    }\n\n    // Apply non-linearity (LeakyReLU)\n    for (let i = 0; i < aggregated.length; i++) {\n      aggregated[i] = aggregated[i] > 0 ? aggregated[i] : 0.01 * aggregated[i];\n    }\n\n    return aggregated;\n  }\n\n  /**\n   * Compute attention score between two nodes\n   */\n  private computeAttentionScore(\n    features1: Float32Array,\n    features2: Float32Array,\n    head: number\n  ): number {\n    // Simple dot-product attention\n    let score = 0;\n    const len = Math.min(features1.length, features2.length);\n\n    for (let i = 0; i < len; i++) {\n      score += features1[i] * features2[i];\n    }\n\n    // Apply softmax-like normalization\n    return Math.exp(score / Math.sqrt(len));\n  }\n\n  /**\n   * Message passing for all nodes\n   */\n  messagePass(): Map<number, Float32Array> {\n    const newFeatures = new Map<number, Float32Array>();\n\n    for (const [nodeId] of this.nodes) {\n      newFeatures.set(nodeId, this.graphAttention(nodeId));\n    }\n\n    return newFeatures;\n  }\n\n  /**\n   * Update node features after message passing\n   */\n  update(newFeatures: Map<number, Float32Array>): void {\n    for (const [nodeId, features] of newFeatures) {\n      const node = this.nodes.get(nodeId);\n      if (node) {\n        node.features = features;\n      }\n    }\n  }\n\n  /**\n   * Compute graph embeddings for query enhancement\n   */\n  computeGraphEmbedding(nodeId: number, hops: number = 2): Float32Array {\n    const features = new Map<number, Float32Array>();\n    features.set(nodeId, this.nodes.get(nodeId)!.features);\n\n    // Multi-hop message passing\n    for (let h = 0; h < hops; h++) {\n      const newFeatures = this.messagePass();\n      this.update(newFeatures);\n    }\n\n    return this.nodes.get(nodeId)!.features;\n  }\n\n  /**\n   * Get statistics\n   */\n  getStats() {\n    return {\n      numNodes: this.nodes.size,\n      numEdges: this.edges.length,\n      avgDegree: this.edges.length / Math.max(this.nodes.size, 1),\n      config: this.config\n    };\n  }\n}\n\n// ============================================================================\n// MMR (Maximal Marginal Relevance)\n// ============================================================================\n\nexport interface MMRConfig {\n  lambda: number;  // Trade-off between relevance and diversity (0-1)\n  metric: 'cosine' | 'euclidean';\n}\n\n/**\n * Maximal Marginal Relevance for diversity ranking\n */\nexport class MaximalMarginalRelevance {\n  private config: MMRConfig;\n\n  constructor(config: Partial<MMRConfig> = {}) {\n    this.config = {\n      lambda: config.lambda || 0.7,\n      metric: config.metric || 'cosine'\n    };\n  }\n\n  /**\n   * Rerank results for diversity\n   * @param query Query vector\n   * @param candidates Candidate vectors with scores\n   * @param k Number of results to return\n   * @returns Reranked indices\n   */\n  rerank(\n    query: Float32Array,\n    candidates: Array<{ id: number; vector: Float32Array; score: number }>,\n    k: number\n  ): number[] {\n    if (candidates.length === 0) return [];\n\n    const selected: number[] = [];\n    const remaining = new Set(candidates.map((_, i) => i));\n\n    // Select first result (highest relevance)\n    let bestIdx = 0;\n    let bestScore = -Infinity;\n\n    for (let i = 0; i < candidates.length; i++) {\n      if (candidates[i].score > bestScore) {\n        bestScore = candidates[i].score;\n        bestIdx = i;\n      }\n    }\n\n    selected.push(candidates[bestIdx].id);\n    remaining.delete(bestIdx);\n\n    // Iteratively select remaining results\n    while (selected.length < k && remaining.size > 0) {\n      let bestMMR = -Infinity;\n      let bestCandidate = -1;\n\n      for (const idx of remaining) {\n        const candidate = candidates[idx];\n\n        // Relevance to query\n        const relevance = this.similarity(query, candidate.vector);\n\n        // Maximum similarity to already selected\n        let maxSimilarity = -Infinity;\n        for (const selectedId of selected) {\n          const selectedCandidate = candidates.find(c => c.id === selectedId)!;\n          const sim = this.similarity(candidate.vector, selectedCandidate.vector);\n          maxSimilarity = Math.max(maxSimilarity, sim);\n        }\n\n        // MMR score\n        const mmr =\n          this.config.lambda * relevance -\n          (1 - this.config.lambda) * maxSimilarity;\n\n        if (mmr > bestMMR) {\n          bestMMR = mmr;\n          bestCandidate = idx;\n        }\n      }\n\n      if (bestCandidate !== -1) {\n        selected.push(candidates[bestCandidate].id);\n        remaining.delete(bestCandidate);\n      } else {\n        break;\n      }\n    }\n\n    return selected;\n  }\n\n  /**\n   * Similarity computation\n   */\n  private similarity(a: Float32Array, b: Float32Array): number {\n    if (this.config.metric === 'cosine') {\n      return this.cosineSimilarity(a, b);\n    } else {\n      // Euclidean distance converted to similarity\n      const dist = this.euclideanDistance(a, b);\n      return 1 / (1 + dist);\n    }\n  }\n\n  private cosineSimilarity(a: Float32Array, b: Float32Array): number {\n    let dotProduct = 0;\n    let normA = 0;\n    let normB = 0;\n\n    for (let i = 0; i < a.length; i++) {\n      dotProduct += a[i] * b[i];\n      normA += a[i] * a[i];\n      normB += b[i] * b[i];\n    }\n\n    return dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));\n  }\n\n  private euclideanDistance(a: Float32Array, b: Float32Array): number {\n    let sum = 0;\n    for (let i = 0; i < a.length; i++) {\n      const diff = a[i] - b[i];\n      sum += diff * diff;\n    }\n    return Math.sqrt(sum);\n  }\n\n  /**\n   * Set lambda (relevance vs diversity trade-off)\n   */\n  setLambda(lambda: number): void {\n    this.config.lambda = Math.max(0, Math.min(1, lambda));\n  }\n}\n\n// ============================================================================\n// SVD (Singular Value Decomposition) for Tensor Compression\n// ============================================================================\n\n/**\n * Simple SVD implementation for dimension reduction\n */\nexport class TensorCompression {\n  /**\n   * Reduce dimensionality using truncated SVD\n   * @param vectors Array of vectors to compress\n   * @param targetDim Target dimension\n   * @returns Compressed vectors\n   */\n  static compress(\n    vectors: Float32Array[],\n    targetDim: number\n  ): Float32Array[] {\n    if (vectors.length === 0) return [];\n\n    const originalDim = vectors[0].length;\n    if (targetDim >= originalDim) return vectors;\n\n    // Create matrix (vectors as rows)\n    const matrix = vectors.map(v => Array.from(v));\n\n    // Center the data (subtract mean)\n    const mean = this.computeMean(matrix);\n    const centered = matrix.map(row =>\n      row.map((val, i) => val - mean[i])\n    );\n\n    // Compute covariance matrix\n    const cov = this.computeCovariance(centered);\n\n    // Compute top k eigenvectors using power iteration\n    const eigenvectors = this.powerIteration(cov, targetDim);\n\n    // Project vectors onto eigenvectors\n    const compressed = centered.map(row => {\n      const projected = new Float32Array(targetDim);\n      for (let i = 0; i < targetDim; i++) {\n        let sum = 0;\n        for (let j = 0; j < originalDim; j++) {\n          sum += row[j] * eigenvectors[i][j];\n        }\n        projected[i] = sum;\n      }\n      return projected;\n    });\n\n    return compressed;\n  }\n\n  /**\n   * Compute mean vector\n   */\n  private static computeMean(matrix: number[][]): number[] {\n    const n = matrix.length;\n    const dim = matrix[0].length;\n    const mean = new Array(dim).fill(0);\n\n    for (const row of matrix) {\n      for (let i = 0; i < dim; i++) {\n        mean[i] += row[i];\n      }\n    }\n\n    return mean.map(v => v / n);\n  }\n\n  /**\n   * Compute covariance matrix\n   */\n  private static computeCovariance(matrix: number[][]): number[][] {\n    const n = matrix.length;\n    const dim = matrix[0].length;\n    const cov: number[][] = Array.from({ length: dim }, () =>\n      new Array(dim).fill(0)\n    );\n\n    for (let i = 0; i < dim; i++) {\n      for (let j = 0; j <= i; j++) {\n        let sum = 0;\n        for (const row of matrix) {\n          sum += row[i] * row[j];\n        }\n        cov[i][j] = cov[j][i] = sum / n;\n      }\n    }\n\n    return cov;\n  }\n\n  /**\n   * Power iteration for computing top eigenvectors\n   */\n  private static powerIteration(\n    matrix: number[][],\n    k: number,\n    iterations: number = 100\n  ): number[][] {\n    const dim = matrix.length;\n    const eigenvectors: number[][] = [];\n\n    for (let i = 0; i < k; i++) {\n      // Random initialization\n      let v = new Array(dim).fill(0).map(() => Math.random() - 0.5);\n\n      // Power iteration\n      for (let iter = 0; iter < iterations; iter++) {\n        // Multiply by matrix\n        const newV = new Array(dim).fill(0);\n        for (let r = 0; r < dim; r++) {\n          for (let c = 0; c < dim; c++) {\n            newV[r] += matrix[r][c] * v[c];\n          }\n        }\n\n        // Orthogonalize against previous eigenvectors\n        for (const prev of eigenvectors) {\n          let dot = 0;\n          for (let j = 0; j < dim; j++) {\n            dot += newV[j] * prev[j];\n          }\n          for (let j = 0; j < dim; j++) {\n            newV[j] -= dot * prev[j];\n          }\n        }\n\n        // Normalize\n        let norm = 0;\n        for (const val of newV) {\n          norm += val * val;\n        }\n        norm = Math.sqrt(norm);\n\n        if (norm < 1e-10) break;\n\n        v = newV.map(val => val / norm);\n      }\n\n      eigenvectors.push(v);\n    }\n\n    return eigenvectors;\n  }\n}\n\n// ============================================================================\n// Batch Operations\n// ============================================================================\n\n/**\n * Efficient batch processing utilities\n */\nexport class BatchProcessor {\n  /**\n   * Batch cosine similarity computation\n   */\n  static batchCosineSimilarity(\n    query: Float32Array,\n    vectors: Float32Array[]\n  ): Float32Array {\n    const similarities = new Float32Array(vectors.length);\n\n    // Precompute query norm\n    let queryNorm = 0;\n    for (let i = 0; i < query.length; i++) {\n      queryNorm += query[i] * query[i];\n    }\n    queryNorm = Math.sqrt(queryNorm);\n\n    // Compute similarities\n    for (let v = 0; v < vectors.length; v++) {\n      const vector = vectors[v];\n      let dotProduct = 0;\n      let vectorNorm = 0;\n\n      for (let i = 0; i < query.length; i++) {\n        dotProduct += query[i] * vector[i];\n        vectorNorm += vector[i] * vector[i];\n      }\n\n      vectorNorm = Math.sqrt(vectorNorm);\n      similarities[v] = dotProduct / (queryNorm * vectorNorm);\n    }\n\n    return similarities;\n  }\n\n  /**\n   * Batch vector normalization\n   */\n  static batchNormalize(vectors: Float32Array[]): Float32Array[] {\n    return vectors.map(v => {\n      let norm = 0;\n      for (let i = 0; i < v.length; i++) {\n        norm += v[i] * v[i];\n      }\n      norm = Math.sqrt(norm);\n\n      const normalized = new Float32Array(v.length);\n      for (let i = 0; i < v.length; i++) {\n        normalized[i] = v[i] / norm;\n      }\n      return normalized;\n    });\n  }\n}\n", "/**\n * Browser WASM Attention Wrapper\n *\n * Provides browser-compatible attention mechanisms with:\n * - Lazy WASM loading\n * - Memory management for WASM linear memory\n * - Fallback to JavaScript when WASM unavailable\n * - Loading states and error handling\n *\n * @module browser/AttentionBrowser\n */\n\nexport interface AttentionConfig {\n  dimension?: number;\n  numHeads?: number;\n  blockSize?: number;\n  curvature?: number;\n  useWASM?: boolean;\n}\n\nexport interface ConsolidationConfig {\n  threshold?: number;\n  maxClusters?: number;\n  minClusterSize?: number;\n}\n\nexport type LoadingState = 'idle' | 'loading' | 'loaded' | 'error';\n\n/**\n * Browser-compatible attention class with WASM support\n */\nexport class AttentionBrowser {\n  private wasmModule: any = null;\n  private loadingState: LoadingState = 'idle';\n  private loadError: Error | null = null;\n  private config: AttentionConfig;\n\n  constructor(config: AttentionConfig = {}) {\n    this.config = {\n      dimension: 384,\n      numHeads: 4,\n      blockSize: 64,\n      curvature: -1.0,\n      useWASM: true,\n      ...config\n    };\n  }\n\n  /**\n   * Get current loading state\n   */\n  getLoadingState(): LoadingState {\n    return this.loadingState;\n  }\n\n  /**\n   * Get loading error if any\n   */\n  getError(): Error | null {\n    return this.loadError;\n  }\n\n  /**\n   * Initialize WASM module (lazy loaded)\n   */\n  async initialize(): Promise<void> {\n    if (this.loadingState === 'loaded') return;\n    if (this.loadingState === 'loading') {\n      // Wait for existing load to complete\n      while (this.loadingState === 'loading') {\n        await new Promise(resolve => setTimeout(resolve, 50));\n      }\n      return;\n    }\n\n    this.loadingState = 'loading';\n\n    try {\n      if (!this.config.useWASM) {\n        // Skip WASM loading\n        this.loadingState = 'loaded';\n        return;\n      }\n\n      // Dynamic import of WASM loader\n      const wasmLoader = await import('../../dist/agentdb.wasm-loader.js');\n      this.wasmModule = await wasmLoader.initWASM();\n      this.loadingState = 'loaded';\n    } catch (error) {\n      this.loadError = error instanceof Error ? error : new Error(String(error));\n      this.loadingState = 'error';\n      console.warn('WASM initialization failed, using fallback:', this.loadError.message);\n      // Don't throw - allow fallback to work\n    }\n  }\n\n  /**\n   * Flash Attention - Optimized attention mechanism\n   * O(N) memory complexity instead of O(N\u00B2)\n   *\n   * @param query - Query vectors\n   * @param keys - Key vectors\n   * @param values - Value vectors\n   * @returns Attention output\n   */\n  async flashAttention(\n    query: Float32Array,\n    keys: Float32Array,\n    values: Float32Array\n  ): Promise<Float32Array> {\n    await this.initialize();\n\n    if (this.wasmModule?.flashAttention) {\n      try {\n        return this.wasmModule.flashAttention(query, keys, values, this.config);\n      } catch (error) {\n        console.warn('WASM flash attention failed, using fallback:', error);\n      }\n    }\n\n    // Fallback to JavaScript implementation\n    return this.flashAttentionFallback(query, keys, values);\n  }\n\n  /**\n   * Hyperbolic Attention - Attention in hyperbolic space\n   * Better for hierarchical relationships\n   *\n   * @param query - Query vector\n   * @param keys - Key vectors\n   * @returns Similarity scores in hyperbolic space\n   */\n  async hyperbolicAttention(\n    query: Float32Array,\n    keys: Float32Array\n  ): Promise<Float32Array> {\n    await this.initialize();\n\n    if (this.wasmModule?.hyperbolicAttention) {\n      try {\n        return this.wasmModule.hyperbolicAttention(query, keys, this.config);\n      } catch (error) {\n        console.warn('WASM hyperbolic attention failed, using fallback:', error);\n      }\n    }\n\n    // Fallback to JavaScript implementation\n    return this.hyperbolicAttentionFallback(query, keys);\n  }\n\n  /**\n   * Memory Consolidation - Cluster and consolidate similar memories\n   *\n   * @param memories - Array of memory vectors\n   * @param config - Consolidation configuration\n   * @returns Consolidated memory clusters\n   */\n  async consolidateMemories(\n    memories: Float32Array[],\n    config: ConsolidationConfig = {}\n  ): Promise<Array<{\n    memory: Float32Array;\n    count: number;\n    members: Float32Array[];\n  }>> {\n    await this.initialize();\n\n    const fullConfig = {\n      threshold: 0.8,\n      maxClusters: 10,\n      minClusterSize: 1,\n      ...config\n    };\n\n    if (this.wasmModule?.memoryConsolidation) {\n      try {\n        return this.wasmModule.memoryConsolidation(memories, fullConfig);\n      } catch (error) {\n        console.warn('WASM memory consolidation failed, using fallback:', error);\n      }\n    }\n\n    // Fallback to JavaScript implementation\n    return this.consolidateMemoriesFallback(memories, fullConfig);\n  }\n\n  /**\n   * Clean up WASM memory\n   */\n  dispose(): void {\n    this.wasmModule = null;\n    this.loadingState = 'idle';\n    this.loadError = null;\n  }\n\n  // ========================================================================\n  // Fallback Implementations (Pure JavaScript)\n  // ========================================================================\n\n  private flashAttentionFallback(\n    query: Float32Array,\n    keys: Float32Array,\n    values: Float32Array\n  ): Float32Array {\n    const { dimension = 384 } = this.config;\n    const seqLen = keys.length / dimension;\n    const output = new Float32Array(query.length);\n\n    for (let i = 0; i < query.length; i += dimension) {\n      const q = query.slice(i, i + dimension);\n      let sumWeights = 0;\n      const weights = new Float32Array(seqLen);\n\n      // Compute attention weights\n      for (let j = 0; j < seqLen; j++) {\n        const k = keys.slice(j * dimension, (j + 1) * dimension);\n        let dot = 0;\n        for (let d = 0; d < dimension; d++) {\n          dot += q[d] * k[d];\n        }\n        weights[j] = Math.exp(dot / Math.sqrt(dimension));\n        sumWeights += weights[j];\n      }\n\n      // Normalize and apply to values\n      for (let j = 0; j < seqLen; j++) {\n        weights[j] /= (sumWeights || 1);\n        const v = values.slice(j * dimension, (j + 1) * dimension);\n        for (let d = 0; d < dimension; d++) {\n          output[i + d] += weights[j] * v[d];\n        }\n      }\n    }\n\n    return output;\n  }\n\n  private hyperbolicAttentionFallback(\n    query: Float32Array,\n    keys: Float32Array\n  ): Float32Array {\n    const { curvature = -1.0 } = this.config;\n    const k = Math.abs(curvature);\n    const similarities = new Float32Array(keys.length / query.length);\n\n    // Hyperbolic distance computation (Poincar\u00E9 ball model)\n    for (let i = 0; i < similarities.length; i++) {\n      const offset = i * query.length;\n      let dotProduct = 0;\n      let normQ = 0;\n      let normK = 0;\n\n      for (let j = 0; j < query.length; j++) {\n        dotProduct += query[j] * keys[offset + j];\n        normQ += query[j] * query[j];\n        normK += keys[offset + j] * keys[offset + j];\n      }\n\n      // Euclidean distance\n      const euclidean = Math.sqrt(normQ + normK - 2 * dotProduct);\n\n      // Poincar\u00E9 distance\n      const poincare = Math.acosh(1 + 2 * k * euclidean * euclidean);\n\n      // Convert to similarity\n      similarities[i] = 1 / (1 + poincare);\n    }\n\n    return similarities;\n  }\n\n  private consolidateMemoriesFallback(\n    memories: Float32Array[],\n    config: ConsolidationConfig\n  ): Array<{\n    memory: Float32Array;\n    count: number;\n    members: Float32Array[];\n  }> {\n    const { threshold = 0.8, maxClusters = 10, minClusterSize = 1 } = config;\n    const consolidated: Array<{\n      memory: Float32Array;\n      count: number;\n      members: Float32Array[];\n    }> = [];\n    const used = new Set<number>();\n\n    // Simple agglomerative clustering\n    for (let i = 0; i < memories.length; i++) {\n      if (used.has(i)) continue;\n\n      const cluster: Float32Array[] = [memories[i]];\n      used.add(i);\n\n      for (let j = i + 1; j < memories.length; j++) {\n        if (used.has(j)) continue;\n\n        // Compute cosine similarity\n        const similarity = this.cosineSimilarity(memories[i], memories[j]);\n\n        if (similarity > threshold) {\n          cluster.push(memories[j]);\n          used.add(j);\n        }\n      }\n\n      // Only include clusters that meet minimum size\n      if (cluster.length >= minClusterSize) {\n        // Compute cluster centroid\n        const centroid = new Float32Array(memories[i].length);\n        for (const mem of cluster) {\n          for (let k = 0; k < centroid.length; k++) {\n            centroid[k] += mem[k] / cluster.length;\n          }\n        }\n\n        // Normalize centroid\n        let norm = 0;\n        for (let k = 0; k < centroid.length; k++) {\n          norm += centroid[k] * centroid[k];\n        }\n        norm = Math.sqrt(norm);\n        if (norm > 0) {\n          for (let k = 0; k < centroid.length; k++) {\n            centroid[k] /= norm;\n          }\n        }\n\n        consolidated.push({\n          memory: centroid,\n          count: cluster.length,\n          members: cluster\n        });\n      }\n\n      if (consolidated.length >= maxClusters) break;\n    }\n\n    return consolidated;\n  }\n\n  private cosineSimilarity(a: Float32Array, b: Float32Array): number {\n    let dot = 0;\n    let normA = 0;\n    let normB = 0;\n\n    for (let i = 0; i < a.length; i++) {\n      dot += a[i] * b[i];\n      normA += a[i] * a[i];\n      normB += b[i] * b[i];\n    }\n\n    const denominator = Math.sqrt(normA * normB);\n    return denominator > 0 ? dot / denominator : 0;\n  }\n}\n\n/**\n * Create attention instance with default config\n */\nexport function createAttention(config?: AttentionConfig): AttentionBrowser {\n  return new AttentionBrowser(config);\n}\n\n/**\n * Create attention instance optimized for speed\n */\nexport function createFastAttention(): AttentionBrowser {\n  return new AttentionBrowser({\n    dimension: 256,\n    numHeads: 2,\n    blockSize: 32,\n    useWASM: true\n  });\n}\n\n/**\n * Create attention instance optimized for quality\n */\nexport function createAccurateAttention(): AttentionBrowser {\n  return new AttentionBrowser({\n    dimension: 768,\n    numHeads: 8,\n    blockSize: 128,\n    useWASM: true\n  });\n}\n", "/**\n * AgentDB Browser Advanced Features\n *\n * Unified export for all browser-compatible advanced features.\n *\n * Features:\n * - Product Quantization (PQ8/PQ16/PQ32) - 4-32x memory compression\n * - HNSW Indexing - 10-20x faster approximate search\n * - Graph Neural Networks - Graph attention and message passing\n * - MMR Diversity - Maximal marginal relevance ranking\n * - Tensor Compression - SVD dimension reduction\n * - Batch Operations - Optimized vector processing\n * - WASM Attention - High-performance attention mechanisms (lazy loaded)\n *\n * Bundle Size: ~35 KB minified (~12 KB gzipped)\n * WASM Module: ~157 KB (lazy loaded on demand)\n */\n\n// ============================================================================\n// Product Quantization\n// ============================================================================\n\nexport {\n  ProductQuantization,\n  createPQ8,\n  createPQ16,\n  createPQ32,\n  type PQConfig,\n  type PQCodebook,\n  type CompressedVector\n} from './ProductQuantization';\n\n// ============================================================================\n// HNSW Indexing\n// ============================================================================\n\nexport {\n  HNSWIndex,\n  createHNSW,\n  createFastHNSW,\n  createAccurateHNSW,\n  type HNSWConfig,\n  type HNSWNode,\n  type SearchResult\n} from './HNSWIndex';\n\n// ============================================================================\n// Advanced Features\n// ============================================================================\n\nexport {\n  GraphNeuralNetwork,\n  MaximalMarginalRelevance,\n  TensorCompression,\n  BatchProcessor,\n  type GNNNode,\n  type GNNEdge,\n  type GNNConfig,\n  type MMRConfig\n} from './AdvancedFeatures';\n\n// ============================================================================\n// WASM Attention (Browser-Compatible)\n// ============================================================================\n\nexport {\n  AttentionBrowser,\n  createAttention,\n  createFastAttention,\n  createAccurateAttention,\n  type AttentionConfig,\n  type ConsolidationConfig,\n  type LoadingState\n} from './AttentionBrowser';\n\n// ============================================================================\n// Feature Detection\n// ============================================================================\n\n/**\n * Detect available browser features\n */\nexport function detectFeatures() {\n  return {\n    indexedDB: 'indexedDB' in globalThis,\n    broadcastChannel: 'BroadcastChannel' in globalThis,\n    webWorkers: typeof (globalThis as any).Worker !== 'undefined',\n    wasmSIMD: detectWasmSIMD(),\n    sharedArrayBuffer: typeof SharedArrayBuffer !== 'undefined'\n  };\n}\n\n/**\n * Detect WASM SIMD support\n */\nasync function detectWasmSIMD(): Promise<boolean> {\n  try {\n    // Check if WebAssembly is available (browser context)\n    if (typeof (globalThis as any).WebAssembly === 'undefined') {\n      return false;\n    }\n\n    // WASM SIMD detection via feature test\n    const simdTest = new Uint8Array([\n      0x00, 0x61, 0x73, 0x6d, 0x01, 0x00, 0x00, 0x00,\n      0x01, 0x05, 0x01, 0x60, 0x00, 0x01, 0x7b, 0x03,\n      0x02, 0x01, 0x00, 0x0a, 0x0a, 0x01, 0x08, 0x00,\n      0xfd, 0x0c, 0xfd, 0x0c, 0xfd, 0x54, 0x0b\n    ]);\n\n    const WA = (globalThis as any).WebAssembly;\n    const module = await WA.instantiate(simdTest);\n    return module instanceof WA.Instance;\n  } catch {\n    return false;\n  }\n}\n\n// ============================================================================\n// Configuration Presets\n// ============================================================================\n\n/**\n * Recommended configuration for small datasets (<1K vectors)\n */\nexport const SMALL_DATASET_CONFIG = {\n  pq: { enabled: false },\n  hnsw: { enabled: false },\n  gnn: { enabled: true, numHeads: 2 },\n  mmr: { enabled: true, lambda: 0.7 },\n  svd: { enabled: false }\n};\n\n/**\n * Recommended configuration for medium datasets (1K-10K vectors)\n */\nexport const MEDIUM_DATASET_CONFIG = {\n  pq: { enabled: true, subvectors: 8 },\n  hnsw: { enabled: true, M: 16 },\n  gnn: { enabled: true, numHeads: 4 },\n  mmr: { enabled: true, lambda: 0.7 },\n  svd: { enabled: false }\n};\n\n/**\n * Recommended configuration for large datasets (10K-100K vectors)\n */\nexport const LARGE_DATASET_CONFIG = {\n  pq: { enabled: true, subvectors: 16 },\n  hnsw: { enabled: true, M: 32 },\n  gnn: { enabled: true, numHeads: 4 },\n  mmr: { enabled: true, lambda: 0.7 },\n  svd: { enabled: true, targetDim: 128 }\n};\n\n/**\n * Memory-optimized configuration (minimal memory usage)\n */\nexport const MEMORY_OPTIMIZED_CONFIG = {\n  pq: { enabled: true, subvectors: 32 },  // 16x compression\n  hnsw: { enabled: true, M: 8 },  // Fewer connections\n  gnn: { enabled: false },\n  mmr: { enabled: false },\n  svd: { enabled: true, targetDim: 64 }  // Aggressive dimension reduction\n};\n\n/**\n * Speed-optimized configuration (fastest search)\n */\nexport const SPEED_OPTIMIZED_CONFIG = {\n  pq: { enabled: false },  // No compression overhead\n  hnsw: { enabled: true, M: 32, efSearch: 100 },  // Maximum HNSW quality\n  gnn: { enabled: false },\n  mmr: { enabled: false },\n  svd: { enabled: false }\n};\n\n/**\n * Quality-optimized configuration (best result quality)\n */\nexport const QUALITY_OPTIMIZED_CONFIG = {\n  pq: { enabled: false },  // No compression\n  hnsw: { enabled: true, M: 48, efConstruction: 400 },  // Highest quality\n  gnn: { enabled: true, numHeads: 8 },  // More attention heads\n  mmr: { enabled: true, lambda: 0.8 },  // More diversity\n  svd: { enabled: false }  // No dimension loss\n};\n\n// ============================================================================\n// Version Information\n// ============================================================================\n\nexport const VERSION = {\n  major: 2,\n  minor: 0,\n  patch: 0,\n  prerelease: 'alpha.2',\n  features: 'advanced',\n  full: '2.0.0-alpha.2+advanced'\n};\n\n// ============================================================================\n// Utility Functions\n// ============================================================================\n\n/**\n * Estimate memory usage for configuration\n */\nexport function estimateMemoryUsage(\n  numVectors: number,\n  dimension: number,\n  config: any\n): {\n  vectors: number;\n  index: number;\n  total: number;\n  totalMB: number;\n} {\n  let vectorBytes = numVectors * dimension * 4;  // Float32Array\n\n  // Apply PQ compression\n  if (config.pq?.enabled) {\n    const subvectors = config.pq.subvectors || 8;\n    vectorBytes = numVectors * (subvectors + 4);  // codes + norm\n  }\n\n  // Apply SVD compression\n  if (config.svd?.enabled) {\n    const targetDim = config.svd.targetDim || dimension / 2;\n    vectorBytes = numVectors * targetDim * 4;\n  }\n\n  // HNSW index overhead\n  let indexBytes = 0;\n  if (config.hnsw?.enabled) {\n    const M = config.hnsw.M || 16;\n    const avgConnections = M * 1.5;  // Estimate\n    indexBytes = numVectors * avgConnections * 4;  // Connection IDs\n  }\n\n  const total = vectorBytes + indexBytes;\n\n  return {\n    vectors: vectorBytes,\n    index: indexBytes,\n    total,\n    totalMB: total / (1024 * 1024)\n  };\n}\n\n/**\n * Recommend configuration based on dataset size\n */\nexport function recommendConfig(numVectors: number, dimension: number) {\n  if (numVectors < 1000) {\n    return {\n      name: 'SMALL_DATASET',\n      config: SMALL_DATASET_CONFIG,\n      reason: 'Small dataset, linear search is fast enough'\n    };\n  } else if (numVectors < 10000) {\n    return {\n      name: 'MEDIUM_DATASET',\n      config: MEDIUM_DATASET_CONFIG,\n      reason: 'Medium dataset, HNSW + PQ8 recommended'\n    };\n  } else {\n    return {\n      name: 'LARGE_DATASET',\n      config: LARGE_DATASET_CONFIG,\n      reason: 'Large dataset, aggressive compression + HNSW recommended'\n    };\n  }\n}\n\n/**\n * Benchmark search performance\n */\nexport async function benchmarkSearch(\n  searchFn: (query: Float32Array, k: number) => any[],\n  numQueries: number = 100,\n  k: number = 10,\n  dimension: number = 384\n): Promise<{\n  avgTimeMs: number;\n  minTimeMs: number;\n  maxTimeMs: number;\n  p50Ms: number;\n  p95Ms: number;\n  p99Ms: number;\n}> {\n  const times: number[] = [];\n\n  for (let i = 0; i < numQueries; i++) {\n    const query = new Float32Array(dimension);\n    for (let d = 0; d < dimension; d++) {\n      query[d] = Math.random() - 0.5;\n    }\n\n    const start = performance.now();\n    searchFn(query, k);\n    const end = performance.now();\n\n    times.push(end - start);\n  }\n\n  times.sort((a, b) => a - b);\n\n  return {\n    avgTimeMs: times.reduce((a, b) => a + b, 0) / times.length,\n    minTimeMs: times[0],\n    maxTimeMs: times[times.length - 1],\n    p50Ms: times[Math.floor(times.length * 0.5)],\n    p95Ms: times[Math.floor(times.length * 0.95)],\n    p99Ms: times[Math.floor(times.length * 0.99)]\n  };\n}\n"],
   "mappings": ";;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAiBA,eAAsB,WAAW;AAC/B,MAAI,WAAY,QAAO;AACvB,MAAI,YAAa,QAAO;AAExB,iBAAe,YAAY;AACzB,QAAI;AAEF,UAAI,OAAO,gBAAgB,aAAa;AACtC,cAAM,IAAI,MAAM,2CAA2C;AAAA,MAC7D;AAGA,YAAM,gBAAgB,MAAM,eAAe;AAC3C,cAAQ,IAAI,sBAAsB,aAAa,EAAE;AAIjD,mBAAa;AAAA,QACX,gBAAgB,yBAAyB;AAAA,QACzC,qBAAqB,8BAA8B;AAAA,QACnD,qBAAqB,8BAA8B;AAAA,QACnD;AAAA,MACF;AAEA,cAAQ,IAAI,qCAAgC;AAC5C,aAAO;AAAA,IACT,SAAS,OAAO;AACd,sBAAgB;AAChB,cAAQ,KAAK,sDAA4C,MAAM,OAAO;AAGtE,mBAAa;AAAA,QACX,gBAAgB,yBAAyB;AAAA,QACzC,qBAAqB,8BAA8B;AAAA,QACnD,qBAAqB,8BAA8B;AAAA,QACnD,eAAe;AAAA,MACjB;AAEA,aAAO;AAAA,IACT,UAAE;AACA,oBAAc;AAAA,IAChB;AAAA,EACF,GAAG;AAEH,SAAO;AACT;AAKA,eAAe,iBAAiB;AAC9B,MAAI;AACF,UAAM,WAAW,IAAI,WAAW;AAAA,MAC9B;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAC1C;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAC1C;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAC1C;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,IACtC,CAAC;AAED,UAAM,SAAS,MAAM,YAAY,YAAY,QAAQ;AACrD,WAAO,kBAAkB,YAAY;AAAA,EACvC,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAKA,SAAS,2BAA2B;AAClC,SAAO,CAAC,OAAO,MAAM,QAAQ,UAAU,CAAC,MAAM;AAC5C,UAAM,EAAE,MAAM,KAAK,WAAW,GAAG,YAAY,GAAG,IAAI;AACpD,UAAM,SAAS,KAAK,SAAS;AAC7B,UAAM,SAAS,IAAI,aAAa,MAAM,MAAM;AAG5C,aAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK,KAAK;AAC1C,YAAM,IAAI,MAAM,MAAM,GAAG,IAAI,GAAG;AAChC,UAAI,aAAa;AACjB,YAAM,UAAU,IAAI,aAAa,MAAM;AAGvC,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,cAAM,IAAI,KAAK,MAAM,IAAI,MAAM,IAAI,KAAK,GAAG;AAC3C,YAAI,MAAM;AACV,iBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,iBAAO,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,QACnB;AACA,gBAAQ,CAAC,IAAI,KAAK,IAAI,MAAM,KAAK,KAAK,GAAG,CAAC;AAC1C,sBAAc,QAAQ,CAAC;AAAA,MACzB;AAGA,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,gBAAQ,CAAC,KAAK;AACd,cAAM,IAAI,OAAO,MAAM,IAAI,MAAM,IAAI,KAAK,GAAG;AAC7C,iBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,iBAAO,IAAI,CAAC,KAAK,QAAQ,CAAC,IAAI,EAAE,CAAC;AAAA,QACnC;AAAA,MACF;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AACF;AAEA,SAAS,gCAAgC;AACvC,SAAO,CAAC,OAAO,MAAM,UAAU,CAAC,MAAM;AACpC,UAAM,EAAE,YAAY,GAAK,IAAI;AAC7B,UAAM,IAAI,KAAK,IAAI,SAAS;AAC5B,UAAM,eAAe,IAAI,aAAa,KAAK,SAAS,MAAM,MAAM;AAGhE,aAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,KAAK;AAC5C,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,aAAa;AACjB,UAAI,QAAQ;AACZ,UAAI,QAAQ;AAEZ,eAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,sBAAc,MAAM,CAAC,IAAI,KAAK,SAAS,CAAC;AACxC,iBAAS,MAAM,CAAC,IAAI,MAAM,CAAC;AAC3B,iBAAS,KAAK,SAAS,CAAC,IAAI,KAAK,SAAS,CAAC;AAAA,MAC7C;AAGA,YAAM,YAAY,KAAK,KAAK,QAAQ,QAAQ,IAAI,UAAU;AAC1D,YAAM,WAAW,KAAK,MAAM,IAAI,IAAI,IAAI,YAAY,SAAS;AAC7D,mBAAa,CAAC,IAAI,KAAK,IAAI;AAAA,IAC7B;AAEA,WAAO;AAAA,EACT;AACF;AAEA,SAAS,gCAAgC;AACvC,SAAO,CAAC,UAAU,UAAU,CAAC,MAAM;AACjC,UAAM,EAAE,YAAY,KAAK,cAAc,GAAG,IAAI;AAC9C,UAAM,eAAe,CAAC;AACtB,UAAM,OAAO,oBAAI,IAAI;AAGrB,aAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,UAAI,KAAK,IAAI,CAAC,EAAG;AAEjB,YAAM,UAAU,CAAC,SAAS,CAAC,CAAC;AAC5B,WAAK,IAAI,CAAC;AAEV,eAAS,IAAI,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AAC5C,YAAI,KAAK,IAAI,CAAC,EAAG;AAGjB,YAAI,MAAM;AACV,YAAI,QAAQ;AACZ,YAAI,QAAQ;AACZ,iBAAS,IAAI,GAAG,IAAI,SAAS,CAAC,EAAE,QAAQ,KAAK;AAC3C,iBAAO,SAAS,CAAC,EAAE,CAAC,IAAI,SAAS,CAAC,EAAE,CAAC;AACrC,mBAAS,SAAS,CAAC,EAAE,CAAC,IAAI,SAAS,CAAC,EAAE,CAAC;AACvC,mBAAS,SAAS,CAAC,EAAE,CAAC,IAAI,SAAS,CAAC,EAAE,CAAC;AAAA,QACzC;AACA,cAAM,aAAa,OAAO,KAAK,KAAK,QAAQ,KAAK,KAAK;AAEtD,YAAI,aAAa,WAAW;AAC1B,kBAAQ,KAAK,SAAS,CAAC,CAAC;AACxB,eAAK,IAAI,CAAC;AAAA,QACZ;AAAA,MACF;AAGA,YAAM,MAAM,IAAI,aAAa,SAAS,CAAC,EAAE,MAAM;AAC/C,iBAAW,OAAO,SAAS;AACzB,iBAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK;AACnC,cAAI,CAAC,KAAK,IAAI,CAAC,IAAI,QAAQ;AAAA,QAC7B;AAAA,MACF;AAEA,mBAAa,KAAK;AAAA,QAChB,QAAQ;AAAA,QACR,OAAO,QAAQ;AAAA,QACf,SAAS;AAAA,MACX,CAAC;AAED,UAAI,aAAa,UAAU,YAAa;AAAA,IAC1C;AAEA,WAAO;AAAA,EACT;AACF;AA5MA,IAUI,YACA,aACA;AAZJ;AAAA;AAAA;AAUA,IAAI,aAAa;AACjB,IAAI,cAAc;AAClB,IAAI,gBAAgB;AAAA;AAAA;;;AC0Bb,IAAM,sBAAN,MAA0B;AAAA,EACvB;AAAA,EACA,WAA8B;AAAA,EAC9B,UAAU;AAAA,EAElB,YAAY,QAAkB;AAC5B,SAAK,SAAS;AAAA,MACZ,WAAW,OAAO;AAAA,MAClB,eAAe,OAAO;AAAA,MACtB,cAAc,OAAO;AAAA,MACrB,eAAe,OAAO,iBAAiB;AAAA,MACvC,sBAAsB,OAAO,wBAAwB;AAAA,IACvD;AAGA,QAAI,KAAK,OAAO,YAAY,KAAK,OAAO,kBAAkB,GAAG;AAC3D,YAAM,IAAI,MAAM,aAAa,KAAK,OAAO,SAAS,uCAAuC,KAAK,OAAO,aAAa,EAAE;AAAA,IACtH;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,MAAM,MAAM,SAAwC;AAClD,QAAI,QAAQ,WAAW,GAAG;AACxB,YAAM,IAAI,MAAM,uCAAuC;AAAA,IACzD;AAEA,UAAM,eAAe,KAAK,OAAO,YAAY,KAAK,OAAO;AACzD,UAAM,YAA4B,CAAC;AAEnC,YAAQ,IAAI,iBAAiB,KAAK,OAAO,aAAa,oBAAoB,KAAK,OAAO,YAAY,oBAAoB;AAGtH,aAAS,IAAI,GAAG,IAAI,KAAK,OAAO,eAAe,KAAK;AAClD,YAAM,WAAW,IAAI;AACrB,YAAM,SAAS,WAAW;AAG1B,YAAM,aAAa,QAAQ,IAAI,OAAK,EAAE,MAAM,UAAU,MAAM,CAAC;AAG7D,YAAM,eAAe,MAAM,KAAK,OAAO,YAAY,KAAK,OAAO,YAAY;AAC3E,gBAAU,KAAK,GAAG,YAAY;AAE9B,WAAK,IAAI,KAAK,MAAM,KAAK,MAAM,KAAK,OAAO,gBAAgB,GAAG;AAC5D,gBAAQ,IAAI,gBAAgB,IAAI,CAAC,IAAI,KAAK,OAAO,aAAa,aAAa;AAAA,MAC7E;AAAA,IACF;AAEA,SAAK,WAAW;AAAA,MACd;AAAA,MACA,eAAe,KAAK,OAAO;AAAA,MAC3B,cAAc,KAAK,OAAO;AAAA,MAC1B;AAAA,IACF;AAEA,SAAK,UAAU;AACf,YAAQ,IAAI,wBAAwB;AAAA,EACtC;AAAA;AAAA;AAAA;AAAA,EAKA,MAAc,OAAO,SAAyB,GAAoC;AAChF,UAAM,MAAM,QAAQ,CAAC,EAAE;AACvB,UAAM,IAAI,QAAQ;AAGlB,UAAM,YAAY,KAAK,eAAe,SAAS,CAAC;AAChD,UAAM,cAAc,IAAI,YAAY,CAAC;AACrC,QAAI,cAAc;AAElB,aAAS,OAAO,GAAG,OAAO,KAAK,OAAO,eAAe,QAAQ;AAE3D,UAAI,UAAU;AACd,eAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,YAAI,UAAU;AACd,YAAI,SAAS;AAEb,iBAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,gBAAM,OAAO,KAAK,gBAAgB,QAAQ,CAAC,GAAG,UAAU,CAAC,CAAC;AAC1D,cAAI,OAAO,SAAS;AAClB,sBAAU;AACV,qBAAS;AAAA,UACX;AAAA,QACF;AAEA,oBAAY,CAAC,IAAI;AACjB,mBAAW;AAAA,MACb;AAGA,UAAI,KAAK,IAAI,cAAc,OAAO,IAAI,KAAK,OAAO,sBAAsB;AACtE;AAAA,MACF;AACA,oBAAc;AAGd,YAAM,SAAS,IAAI,YAAY,CAAC;AAChC,YAAM,OAAO,MAAM,KAAK,EAAE,QAAQ,EAAE,GAAG,MAAM,IAAI,aAAa,GAAG,CAAC;AAElE,eAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,cAAM,UAAU,YAAY,CAAC;AAC7B,eAAO,OAAO;AACd,iBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,eAAK,OAAO,EAAE,CAAC,KAAK,QAAQ,CAAC,EAAE,CAAC;AAAA,QAClC;AAAA,MACF;AAEA,eAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,YAAI,OAAO,CAAC,IAAI,GAAG;AACjB,mBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,sBAAU,CAAC,EAAE,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC,IAAI,OAAO,CAAC;AAAA,UACzC;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKQ,eAAe,SAAyB,GAA2B;AACzE,UAAM,IAAI,QAAQ;AAClB,UAAM,MAAM,QAAQ,CAAC,EAAE;AACvB,UAAM,YAA4B,CAAC;AAGnC,UAAM,WAAW,KAAK,MAAM,KAAK,OAAO,IAAI,CAAC;AAC7C,cAAU,KAAK,IAAI,aAAa,QAAQ,QAAQ,CAAC,CAAC;AAGlD,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,YAAM,YAAY,IAAI,aAAa,CAAC;AACpC,UAAI,eAAe;AAGnB,eAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,YAAI,UAAU;AACd,mBAAW,YAAY,WAAW;AAChC,gBAAM,OAAO,KAAK,gBAAgB,QAAQ,CAAC,GAAG,QAAQ;AACtD,oBAAU,KAAK,IAAI,SAAS,IAAI;AAAA,QAClC;AACA,kBAAU,CAAC,IAAI;AACf,wBAAgB;AAAA,MAClB;AAGA,UAAI,IAAI,KAAK,OAAO,IAAI;AACxB,eAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,aAAK,UAAU,CAAC;AAChB,YAAI,KAAK,GAAG;AACV,oBAAU,KAAK,IAAI,aAAa,QAAQ,CAAC,CAAC,CAAC;AAC3C;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,SAAS,QAAwC;AAC/C,QAAI,CAAC,KAAK,WAAW,CAAC,KAAK,UAAU;AACnC,YAAM,IAAI,MAAM,6CAA6C;AAAA,IAC/D;AAEA,UAAM,QAAQ,IAAI,WAAW,KAAK,OAAO,aAAa;AACtD,UAAM,eAAe,KAAK,SAAS;AAGnC,QAAI,OAAO;AACX,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,cAAQ,OAAO,CAAC,IAAI,OAAO,CAAC;AAAA,IAC9B;AACA,WAAO,KAAK,KAAK,IAAI;AAGrB,aAAS,IAAI,GAAG,IAAI,KAAK,OAAO,eAAe,KAAK;AAClD,YAAM,WAAW,IAAI;AACrB,YAAM,YAAY,OAAO,MAAM,UAAU,WAAW,YAAY;AAGhE,UAAI,UAAU;AACd,UAAI,SAAS;AAEb,YAAM,iBAAiB,IAAI,KAAK,OAAO;AACvC,eAAS,IAAI,GAAG,IAAI,KAAK,OAAO,cAAc,KAAK;AACjD,cAAM,WAAW,KAAK,SAAS,UAAU,iBAAiB,CAAC;AAC3D,cAAM,OAAO,KAAK,gBAAgB,WAAW,QAAQ;AACrD,YAAI,OAAO,SAAS;AAClB,oBAAU;AACV,mBAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,CAAC,IAAI;AAAA,IACb;AAEA,WAAO,EAAE,OAAO,KAAK;AAAA,EACvB;AAAA;AAAA;AAAA;AAAA,EAKA,WAAW,YAA4C;AACrD,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,wBAAwB;AAAA,IAC1C;AAEA,UAAM,SAAS,IAAI,aAAa,KAAK,OAAO,SAAS;AACrD,UAAM,eAAe,KAAK,SAAS;AAEnC,aAAS,IAAI,GAAG,IAAI,KAAK,OAAO,eAAe,KAAK;AAClD,YAAM,OAAO,WAAW,MAAM,CAAC;AAC/B,YAAM,iBAAiB,IAAI,KAAK,OAAO;AACvC,YAAM,WAAW,KAAK,SAAS,UAAU,iBAAiB,IAAI;AAE9D,YAAM,WAAW,IAAI;AACrB,eAAS,IAAI,GAAG,IAAI,cAAc,KAAK;AACrC,eAAO,WAAW,CAAC,IAAI,SAAS,CAAC;AAAA,MACnC;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,mBAAmB,OAAqB,YAAsC;AAC5E,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,wBAAwB;AAAA,IAC1C;AAEA,QAAI,WAAW;AACf,UAAM,eAAe,KAAK,SAAS;AAEnC,aAAS,IAAI,GAAG,IAAI,KAAK,OAAO,eAAe,KAAK;AAClD,YAAM,OAAO,WAAW,MAAM,CAAC;AAC/B,YAAM,iBAAiB,IAAI,KAAK,OAAO;AACvC,YAAM,WAAW,KAAK,SAAS,UAAU,iBAAiB,IAAI;AAE9D,YAAM,WAAW,IAAI;AACrB,YAAM,iBAAiB,MAAM,MAAM,UAAU,WAAW,YAAY;AAEpE,kBAAY,KAAK,gBAAgB,gBAAgB,QAAQ;AAAA,IAC3D;AAEA,WAAO,KAAK,KAAK,QAAQ;AAAA,EAC3B;AAAA;AAAA;AAAA;AAAA,EAKA,cAAc,SAA6C;AACzD,WAAO,QAAQ,IAAI,OAAK,KAAK,SAAS,CAAC,CAAC;AAAA,EAC1C;AAAA;AAAA;AAAA;AAAA,EAKA,sBAA8B;AAG5B,UAAM,gBAAgB,KAAK,OAAO,YAAY;AAC9C,UAAM,kBAAkB,KAAK,OAAO,gBAAgB;AACpD,WAAO,gBAAgB;AAAA,EACzB;AAAA;AAAA;AAAA;AAAA,EAKA,iBAAyB;AACvB,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,uBAAuB;AAAA,IACzC;AAEA,WAAO,KAAK,UAAU;AAAA,MACpB,QAAQ,KAAK;AAAA,MACb,UAAU;AAAA,QACR,cAAc,KAAK,SAAS;AAAA,QAC5B,eAAe,KAAK,SAAS;AAAA,QAC7B,cAAc,KAAK,SAAS;AAAA,QAC5B,WAAW,KAAK,SAAS,UAAU,IAAI,OAAK,MAAM,KAAK,CAAC,CAAC;AAAA,MAC3D;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA,EAKA,eAAe,MAAoB;AACjC,UAAM,OAAO,KAAK,MAAM,IAAI;AAC5B,SAAK,SAAS,KAAK;AACnB,SAAK,WAAW;AAAA,MACd,cAAc,KAAK,SAAS;AAAA,MAC5B,eAAe,KAAK,SAAS;AAAA,MAC7B,cAAc,KAAK,SAAS;AAAA,MAC5B,WAAW,KAAK,SAAS,UAAU,IAAI,CAAC,MAAgB,IAAI,aAAa,CAAC,CAAC;AAAA,IAC7E;AACA,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA;AAAA;AAAA,EAKQ,gBAAgB,GAAiB,GAAyB;AAChE,QAAI,MAAM;AACV,aAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,YAAM,OAAO,EAAE,CAAC,IAAI,EAAE,CAAC;AACvB,aAAO,OAAO;AAAA,IAChB;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,WAKE;AACA,UAAM,mBAAmB,KAAK,oBAAoB;AAClD,UAAM,kBAAkB,KAAK,OAAO,gBAAgB;AACpD,UAAM,eAAe,KAAK,WACtB,KAAK,OAAO,gBAAgB,KAAK,OAAO,gBAAgB,KAAK,OAAO,YAAY,KAAK,OAAO,iBAAiB,IAC7G;AAEJ,WAAO;AAAA,MACL,SAAS,KAAK;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACF;AAKO,SAAS,UAAU,WAAwC;AAChE,SAAO,IAAI,oBAAoB;AAAA,IAC7B;AAAA,IACA,eAAe;AAAA,IACf,cAAc;AAAA,IACd,eAAe;AAAA,EACjB,CAAC;AACH;AAKO,SAAS,WAAW,WAAwC;AACjE,SAAO,IAAI,oBAAoB;AAAA,IAC7B;AAAA,IACA,eAAe;AAAA,IACf,cAAc;AAAA,IACd,eAAe;AAAA,EACjB,CAAC;AACH;AAKO,SAAS,WAAW,WAAwC;AACjE,SAAO,IAAI,oBAAoB;AAAA,IAC7B;AAAA,IACA,eAAe;AAAA,IACf,cAAc;AAAA,IACd,eAAe;AAAA,EACjB,CAAC;AACH;;;ACvXA,IAAM,UAAN,MAAiB;AAAA,EACP,QAA8C,CAAC;AAAA,EAEvD,KAAK,MAAS,UAAwB;AACpC,SAAK,MAAM,KAAK,EAAE,MAAM,SAAS,CAAC;AAClC,SAAK,SAAS,KAAK,MAAM,SAAS,CAAC;AAAA,EACrC;AAAA,EAEA,MAAqB;AACnB,QAAI,KAAK,MAAM,WAAW,EAAG,QAAO;AACpC,UAAM,SAAS,KAAK,MAAM,CAAC,EAAE;AAC7B,UAAM,OAAO,KAAK,MAAM,IAAI;AAC5B,QAAI,KAAK,MAAM,SAAS,GAAG;AACzB,WAAK,MAAM,CAAC,IAAI;AAChB,WAAK,WAAW,CAAC;AAAA,IACnB;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AA9DxB;AA+DI,YAAO,UAAK,MAAM,CAAC,MAAZ,mBAAe;AAAA,EACxB;AAAA,EAEA,OAAe;AACb,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EAEQ,SAAS,OAAqB;AACpC,WAAO,QAAQ,GAAG;AAChB,YAAM,cAAc,KAAK,OAAO,QAAQ,KAAK,CAAC;AAC9C,UAAI,KAAK,MAAM,KAAK,EAAE,YAAY,KAAK,MAAM,WAAW,EAAE,SAAU;AACpE,OAAC,KAAK,MAAM,KAAK,GAAG,KAAK,MAAM,WAAW,CAAC,IAAI,CAAC,KAAK,MAAM,WAAW,GAAG,KAAK,MAAM,KAAK,CAAC;AAC1F,cAAQ;AAAA,IACV;AAAA,EACF;AAAA,EAEQ,WAAW,OAAqB;AACtC,WAAO,MAAM;AACX,YAAM,YAAY,IAAI,QAAQ;AAC9B,YAAM,aAAa,IAAI,QAAQ;AAC/B,UAAI,WAAW;AAEf,UAAI,YAAY,KAAK,MAAM,UAAU,KAAK,MAAM,SAAS,EAAE,WAAW,KAAK,MAAM,QAAQ,EAAE,UAAU;AACnG,mBAAW;AAAA,MACb;AACA,UAAI,aAAa,KAAK,MAAM,UAAU,KAAK,MAAM,UAAU,EAAE,WAAW,KAAK,MAAM,QAAQ,EAAE,UAAU;AACrG,mBAAW;AAAA,MACb;AACA,UAAI,aAAa,MAAO;AAExB,OAAC,KAAK,MAAM,KAAK,GAAG,KAAK,MAAM,QAAQ,CAAC,IAAI,CAAC,KAAK,MAAM,QAAQ,GAAG,KAAK,MAAM,KAAK,CAAC;AACpF,cAAQ;AAAA,IACV;AAAA,EACF;AACF;AAEO,IAAM,YAAN,MAAgB;AAAA,EACb;AAAA,EACA,QAA+B,oBAAI,IAAI;AAAA,EACvC,aAA4B;AAAA,EAC5B,YAAY;AAAA,EACZ;AAAA,EAER,YAAY,SAA8B,CAAC,GAAG;AAC5C,SAAK,SAAS;AAAA,MACZ,WAAW,OAAO,aAAa;AAAA,MAC/B,GAAG,OAAO,KAAK;AAAA,MACf,gBAAgB,OAAO,kBAAkB;AAAA,MACzC,UAAU,OAAO,YAAY;AAAA,MAC7B,IAAI,OAAO,MAAM,IAAI,KAAK,IAAI,CAAC;AAAA,MAC/B,WAAW,OAAO,aAAa;AAAA,MAC/B,kBAAkB,OAAO,oBAAoB;AAAA,IAC/C;AAEA,SAAK,KAAK,KAAK,OAAO;AAAA,EACxB;AAAA;AAAA;AAAA;AAAA,EAKA,IAAI,QAAsB,IAAqB;AAC7C,UAAM,SAAS,OAAO,SAAY,KAAK,KAAK;AAC5C,UAAM,QAAQ,KAAK,YAAY;AAE/B,UAAM,OAAiB;AAAA,MACrB,IAAI;AAAA,MACJ;AAAA,MACA;AAAA,MACA,aAAa,oBAAI,IAAI;AAAA,IACvB;AAGA,aAAS,IAAI,GAAG,KAAK,OAAO,KAAK;AAC/B,WAAK,YAAY,IAAI,GAAG,CAAC,CAAC;AAAA,IAC5B;AAEA,QAAI,KAAK,eAAe,MAAM;AAE5B,WAAK,aAAa;AAClB,WAAK,MAAM,IAAI,QAAQ,IAAI;AAC3B,aAAO;AAAA,IACT;AAGA,UAAM,KAAK,KAAK;AAChB,QAAI,UAAU;AAGd,aAAS,KAAK,KAAK,MAAM,IAAI,EAAE,EAAG,OAAO,KAAK,OAAO,MAAM;AACzD,gBAAU,KAAK,YAAY,QAAQ,SAAS,GAAG,EAAE,EAAE,CAAC;AAAA,IACtD;AAGA,aAAS,KAAK,KAAK,IAAI,OAAO,KAAK,MAAM,IAAI,EAAE,EAAG,KAAK,GAAG,MAAM,GAAG,MAAM;AACvE,YAAM,aAAa,KAAK,YAAY,QAAQ,SAAS,KAAK,OAAO,gBAAgB,EAAE;AAGnF,YAAM,IAAI,OAAO,IAAI,KAAK,OAAO,IAAI,IAAI,KAAK,OAAO;AACrD,YAAM,YAAY,KAAK,gBAAgB,QAAQ,YAAY,CAAC;AAG5D,iBAAW,YAAY,WAAW;AAChC,aAAK,QAAQ,QAAQ,UAAU,EAAE;AACjC,aAAK,QAAQ,UAAU,QAAQ,EAAE;AAGjC,cAAM,eAAe,KAAK,MAAM,IAAI,QAAQ;AAC5C,cAAM,sBAAsB,aAAa,YAAY,IAAI,EAAE;AAC3D,YAAI,oBAAoB,SAAS,GAAG;AAClC,gBAAM,eAAe,KAAK;AAAA,YACxB,aAAa;AAAA,YACb;AAAA,YACA;AAAA,UACF;AACA,uBAAa,YAAY,IAAI,IAAI,YAAY;AAAA,QAC/C;AAAA,MACF;AAEA,gBAAU,WAAW,CAAC;AAAA,IACxB;AAGA,QAAI,QAAQ,KAAK,MAAM,IAAI,KAAK,UAAU,EAAG,OAAO;AAClD,WAAK,aAAa;AAAA,IACpB;AAEA,SAAK,MAAM,IAAI,QAAQ,IAAI;AAC3B,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,OAAO,OAAqB,GAAW,IAA6B;AAClE,QAAI,KAAK,eAAe,KAAM,QAAO,CAAC;AAEtC,SAAK,MAAM,KAAK,IAAI,KAAK,OAAO,UAAU,CAAC;AAE3C,QAAI,KAAK,KAAK;AACd,QAAI,UAAU;AAGd,aAAS,KAAK,KAAK,MAAM,IAAI,EAAE,EAAG,OAAO,KAAK,GAAG,MAAM;AACrD,gBAAU,KAAK,YAAY,OAAO,SAAS,GAAG,EAAE,EAAE,CAAC;AAAA,IACrD;AAGA,UAAM,aAAa,KAAK,YAAY,OAAO,SAAS,IAAI,CAAC;AAGzD,WAAO,WACJ,MAAM,GAAG,CAAC,EACV,IAAI,SAAO;AAAA,MACV;AAAA,MACA,UAAU,KAAK,SAAS,OAAO,KAAK,MAAM,IAAI,EAAE,EAAG,MAAM;AAAA,MACzD,QAAQ,KAAK,MAAM,IAAI,EAAE,EAAG;AAAA,IAC9B,EAAE;AAAA,EACN;AAAA;AAAA;AAAA;AAAA,EAKQ,YAAY,OAAqB,IAAY,IAAY,OAAyB;AACxF,UAAM,UAAU,oBAAI,IAAY;AAChC,UAAM,aAAa,IAAI,QAAgB;AACvC,UAAM,IAAI,IAAI,QAAgB;AAE9B,UAAM,OAAO,KAAK,SAAS,OAAO,KAAK,MAAM,IAAI,EAAE,EAAG,MAAM;AAC5D,eAAW,KAAK,IAAI,IAAI;AACxB,MAAE,KAAK,IAAI,CAAC,IAAI;AAChB,YAAQ,IAAI,EAAE;AAEd,WAAO,WAAW,KAAK,IAAI,GAAG;AAC5B,YAAM,IAAI,WAAW,IAAI;AACzB,YAAM,QAAQ,CAAC,EAAE,KAAK;AAEtB,YAAM,QAAQ,KAAK,SAAS,OAAO,KAAK,MAAM,IAAI,CAAC,EAAG,MAAM;AAC5D,UAAI,QAAQ,MAAO;AAEnB,YAAM,YAAY,KAAK,MAAM,IAAI,CAAC,EAAG,YAAY,IAAI,KAAK,KAAK,CAAC;AAChE,iBAAW,KAAK,WAAW;AACzB,YAAI,QAAQ,IAAI,CAAC,EAAG;AACpB,gBAAQ,IAAI,CAAC;AAEb,cAAM,QAAQ,KAAK,SAAS,OAAO,KAAK,MAAM,IAAI,CAAC,EAAG,MAAM;AAC5D,cAAMA,SAAQ,CAAC,EAAE,KAAK;AAEtB,YAAI,QAAQA,UAAS,EAAE,KAAK,IAAI,IAAI;AAClC,qBAAW,KAAK,GAAG,KAAK;AACxB,YAAE,KAAK,GAAG,CAAC,KAAK;AAEhB,cAAI,EAAE,KAAK,IAAI,IAAI;AACjB,cAAE,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAGA,UAAM,SAAmB,CAAC;AAC1B,WAAO,EAAE,KAAK,IAAI,GAAG;AACnB,aAAO,QAAQ,EAAE,IAAI,CAAE;AAAA,IACzB;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKQ,gBAAgB,MAAoB,YAAsB,GAAqB;AACrF,QAAI,WAAW,UAAU,EAAG,QAAO;AAGnC,UAAM,SAAS,WACZ,IAAI,SAAO;AAAA,MACV;AAAA,MACA,UAAU,KAAK,SAAS,MAAM,KAAK,MAAM,IAAI,EAAE,EAAG,MAAM;AAAA,IAC1D,EAAE,EACD,KAAK,CAAC,GAAG,MAAM,EAAE,WAAW,EAAE,QAAQ;AAEzC,WAAO,OAAO,MAAM,GAAG,CAAC,EAAE,IAAI,OAAK,EAAE,EAAE;AAAA,EACzC;AAAA;AAAA;AAAA;AAAA,EAKQ,QAAQ,MAAc,IAAY,OAAqB;AAC7D,UAAM,OAAO,KAAK,MAAM,IAAI,IAAI;AAChC,UAAM,cAAc,KAAK,YAAY,IAAI,KAAK;AAC9C,QAAI,CAAC,YAAY,SAAS,EAAE,GAAG;AAC7B,kBAAY,KAAK,EAAE;AAAA,IACrB;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKQ,cAAsB;AAC5B,QAAI,QAAQ;AACZ,WAAO,KAAK,OAAO,IAAI,KAAK,MAAM,QAAQ,KAAK,OAAO,YAAY,GAAG;AACnE;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKQ,SAAS,GAAiB,GAAyB;AACzD,YAAQ,KAAK,OAAO,kBAAkB;AAAA,MACpC,KAAK;AACH,eAAO,IAAI,KAAK,iBAAiB,GAAG,CAAC;AAAA,MACvC,KAAK;AACH,eAAO,KAAK,kBAAkB,GAAG,CAAC;AAAA,MACpC,KAAK;AACH,eAAO,KAAK,kBAAkB,GAAG,CAAC;AAAA,MACpC;AACE,eAAO,IAAI,KAAK,iBAAiB,GAAG,CAAC;AAAA,IACzC;AAAA,EACF;AAAA,EAEQ,iBAAiB,GAAiB,GAAyB;AACjE,QAAI,aAAa;AACjB,QAAI,QAAQ;AACZ,QAAI,QAAQ;AAEZ,aAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,oBAAc,EAAE,CAAC,IAAI,EAAE,CAAC;AACxB,eAAS,EAAE,CAAC,IAAI,EAAE,CAAC;AACnB,eAAS,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,IACrB;AAEA,WAAO,cAAc,KAAK,KAAK,KAAK,IAAI,KAAK,KAAK,KAAK;AAAA,EACzD;AAAA,EAEQ,kBAAkB,GAAiB,GAAyB;AAClE,QAAI,MAAM;AACV,aAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,YAAM,OAAO,EAAE,CAAC,IAAI,EAAE,CAAC;AACvB,aAAO,OAAO;AAAA,IAChB;AACA,WAAO,KAAK,KAAK,GAAG;AAAA,EACtB;AAAA,EAEQ,kBAAkB,GAAiB,GAAyB;AAClE,QAAI,MAAM;AACV,aAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,aAAO,KAAK,IAAI,EAAE,CAAC,IAAI,EAAE,CAAC,CAAC;AAAA,IAC7B;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,WAME;AACA,QAAI,KAAK,MAAM,SAAS,GAAG;AACzB,aAAO;AAAA,QACL,UAAU;AAAA,QACV,WAAW;AAAA,QACX,gBAAgB;AAAA,QAChB,iBAAiB;AAAA,QACjB,aAAa;AAAA,MACf;AAAA,IACF;AAEA,UAAM,WAAW,KAAK,IAAI,GAAG,MAAM,KAAK,KAAK,MAAM,OAAO,CAAC,EAAE,IAAI,OAAK,EAAE,KAAK,CAAC;AAC9E,QAAI,mBAAmB;AAEvB,eAAW,QAAQ,KAAK,MAAM,OAAO,GAAG;AACtC,iBAAW,eAAe,KAAK,YAAY,OAAO,GAAG;AACnD,4BAAoB,YAAY;AAAA,MAClC;AAAA,IACF;AAEA,UAAM,iBAAiB,mBAAmB,KAAK,MAAM;AAGrD,UAAM,cAAc,KAAK,OAAO,YAAY;AAC5C,UAAM,kBAAkB,iBAAiB;AACzC,UAAM,gBAAgB;AACtB,UAAM,cAAc,KAAK,MAAM,QAAQ,cAAc,kBAAkB;AAEvE,WAAO;AAAA,MACL,UAAU,KAAK,MAAM;AAAA,MACrB,WAAW,WAAW;AAAA,MACtB;AAAA,MACA,iBAAiB,KAAK,aAAa,KAAK,MAAM,IAAI,KAAK,UAAU,EAAG,QAAQ;AAAA,MAC5E;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,SAAiB;AACf,UAAM,OAAO;AAAA,MACX,QAAQ,KAAK;AAAA,MACb,YAAY,KAAK;AAAA,MACjB,WAAW,KAAK;AAAA,MAChB,OAAO,MAAM,KAAK,KAAK,MAAM,QAAQ,CAAC,EAAE,IAAI,CAAC,CAAC,IAAI,IAAI,OAAO;AAAA,QAC3D;AAAA,QACA,QAAQ,MAAM,KAAK,KAAK,MAAM;AAAA,QAC9B,OAAO,KAAK;AAAA,QACZ,aAAa,MAAM,KAAK,KAAK,YAAY,QAAQ,CAAC;AAAA,MACpD,EAAE;AAAA,IACJ;AAEA,WAAO,KAAK,UAAU,IAAI;AAAA,EAC5B;AAAA;AAAA;AAAA;AAAA,EAKA,OAAO,MAAoB;AACzB,UAAM,OAAO,KAAK,MAAM,IAAI;AAE5B,SAAK,SAAS,KAAK;AACnB,SAAK,aAAa,KAAK;AACvB,SAAK,YAAY,KAAK;AACtB,SAAK,MAAM,MAAM;AAEjB,eAAW,YAAY,KAAK,OAAO;AACjC,YAAM,OAAiB;AAAA,QACrB,IAAI,SAAS;AAAA,QACb,QAAQ,IAAI,aAAa,SAAS,MAAM;AAAA,QACxC,OAAO,SAAS;AAAA,QAChB,aAAa,IAAI,IAAI,SAAS,WAAW;AAAA,MAC3C;AACA,WAAK,MAAM,IAAI,SAAS,IAAI,IAAI;AAAA,IAClC;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,QAAc;AACZ,SAAK,MAAM,MAAM;AACjB,SAAK,aAAa;AAClB,SAAK,YAAY;AAAA,EACnB;AAAA;AAAA;AAAA;AAAA,EAKA,OAAe;AACb,WAAO,KAAK,MAAM;AAAA,EACpB;AACF;AAKO,SAAS,WAAW,WAA8B;AACvD,SAAO,IAAI,UAAU;AAAA,IACnB;AAAA,IACA,GAAG;AAAA,IACH,gBAAgB;AAAA,IAChB,UAAU;AAAA,EACZ,CAAC;AACH;AAKO,SAAS,eAAe,WAA8B;AAC3D,SAAO,IAAI,UAAU;AAAA,IACnB;AAAA,IACA,GAAG;AAAA,IACH,gBAAgB;AAAA,IAChB,UAAU;AAAA,EACZ,CAAC;AACH;AAKO,SAAS,mBAAmB,WAA8B;AAC/D,SAAO,IAAI,UAAU;AAAA,IACnB;AAAA,IACA,GAAG;AAAA,IACH,gBAAgB;AAAA,IAChB,UAAU;AAAA,EACZ,CAAC;AACH;;;ACxcO,IAAM,qBAAN,MAAyB;AAAA,EACtB;AAAA,EACA,QAA8B,oBAAI,IAAI;AAAA,EACtC,QAAmB,CAAC;AAAA,EACpB,mBAAwC,oBAAI,IAAI;AAAA,EAExD,YAAY,SAA6B,CAAC,GAAG;AAC3C,SAAK,SAAS;AAAA,MACZ,WAAW,OAAO,aAAa;AAAA,MAC/B,UAAU,OAAO,YAAY;AAAA,MAC7B,SAAS,OAAO,WAAW;AAAA,MAC3B,cAAc,OAAO,gBAAgB;AAAA,MACrC,eAAe,OAAO,iBAAiB;AAAA,IACzC;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,QAAQ,IAAY,UAA8B;AAChD,SAAK,MAAM,IAAI,IAAI;AAAA,MACjB;AAAA,MACA;AAAA,MACA,WAAW,CAAC;AAAA,IACd,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA,EAKA,QAAQ,MAAc,IAAY,SAAiB,GAAW;AAC5D,SAAK,MAAM,KAAK,EAAE,MAAM,IAAI,OAAO,CAAC;AAGpC,UAAM,WAAW,KAAK,MAAM,IAAI,IAAI;AACpC,UAAM,SAAS,KAAK,MAAM,IAAI,EAAE;AAEhC,QAAI,YAAY,CAAC,SAAS,UAAU,SAAS,EAAE,GAAG;AAChD,eAAS,UAAU,KAAK,EAAE;AAAA,IAC5B;AACA,QAAI,UAAU,CAAC,OAAO,UAAU,SAAS,IAAI,GAAG;AAC9C,aAAO,UAAU,KAAK,IAAI;AAAA,IAC5B;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,eAAe,QAA8B;AAC3C,UAAM,OAAO,KAAK,MAAM,IAAI,MAAM;AAClC,QAAI,CAAC,KAAM,OAAM,IAAI,MAAM,QAAQ,MAAM,YAAY;AAErD,UAAM,YAAY,KAAK;AACvB,QAAI,UAAU,WAAW,GAAG;AAC1B,aAAO,KAAK;AAAA,IACd;AAGA,UAAM,UAAU,KAAK,MAAM,KAAK,OAAO,YAAY,KAAK,OAAO,QAAQ;AACvE,UAAM,aAAa,IAAI,aAAa,KAAK,OAAO,SAAS;AAEzD,aAAS,IAAI,GAAG,IAAI,KAAK,OAAO,UAAU,KAAK;AAC7C,UAAI,eAAe;AACnB,YAAM,aAAa,IAAI,aAAa,OAAO;AAG3C,iBAAW,cAAc,WAAW;AAClC,cAAM,WAAW,KAAK,MAAM,IAAI,UAAU;AAG1C,cAAM,QAAQ,KAAK;AAAA,UACjB,KAAK;AAAA,UACL,SAAS;AAAA,UACT;AAAA,QACF;AAEA,wBAAgB;AAGhB,iBAAS,IAAI,GAAG,IAAI,WAAW,IAAI,SAAS,SAAS,QAAQ,KAAK;AAChE,qBAAW,CAAC,KAAK,QAAQ,SAAS,SAAS,CAAC;AAAA,QAC9C;AAAA,MACF;AAGA,UAAI,eAAe,GAAG;AACpB,iBAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,qBAAW,CAAC,KAAK;AAAA,QACnB;AAAA,MACF;AAGA,YAAM,SAAS,IAAI;AACnB,eAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,mBAAW,SAAS,CAAC,IAAI,WAAW,CAAC;AAAA,MACvC;AAAA,IACF;AAGA,aAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,KAAK;AAC1C,iBAAW,CAAC,IAAI,WAAW,CAAC,IAAI,IAAI,WAAW,CAAC,IAAI,OAAO,WAAW,CAAC;AAAA,IACzE;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKQ,sBACN,WACA,WACA,MACQ;AAER,QAAI,QAAQ;AACZ,UAAM,MAAM,KAAK,IAAI,UAAU,QAAQ,UAAU,MAAM;AAEvD,aAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,eAAS,UAAU,CAAC,IAAI,UAAU,CAAC;AAAA,IACrC;AAGA,WAAO,KAAK,IAAI,QAAQ,KAAK,KAAK,GAAG,CAAC;AAAA,EACxC;AAAA;AAAA;AAAA;AAAA,EAKA,cAAyC;AACvC,UAAM,cAAc,oBAAI,IAA0B;AAElD,eAAW,CAAC,MAAM,KAAK,KAAK,OAAO;AACjC,kBAAY,IAAI,QAAQ,KAAK,eAAe,MAAM,CAAC;AAAA,IACrD;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,OAAO,aAA8C;AACnD,eAAW,CAAC,QAAQ,QAAQ,KAAK,aAAa;AAC5C,YAAM,OAAO,KAAK,MAAM,IAAI,MAAM;AAClC,UAAI,MAAM;AACR,aAAK,WAAW;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,sBAAsB,QAAgB,OAAe,GAAiB;AACpE,UAAM,WAAW,oBAAI,IAA0B;AAC/C,aAAS,IAAI,QAAQ,KAAK,MAAM,IAAI,MAAM,EAAG,QAAQ;AAGrD,aAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,YAAM,cAAc,KAAK,YAAY;AACrC,WAAK,OAAO,WAAW;AAAA,IACzB;AAEA,WAAO,KAAK,MAAM,IAAI,MAAM,EAAG;AAAA,EACjC;AAAA;AAAA;AAAA;AAAA,EAKA,WAAW;AACT,WAAO;AAAA,MACL,UAAU,KAAK,MAAM;AAAA,MACrB,UAAU,KAAK,MAAM;AAAA,MACrB,WAAW,KAAK,MAAM,SAAS,KAAK,IAAI,KAAK,MAAM,MAAM,CAAC;AAAA,MAC1D,QAAQ,KAAK;AAAA,IACf;AAAA,EACF;AACF;AAcO,IAAM,2BAAN,MAA+B;AAAA,EAC5B;AAAA,EAER,YAAY,SAA6B,CAAC,GAAG;AAC3C,SAAK,SAAS;AAAA,MACZ,QAAQ,OAAO,UAAU;AAAA,MACzB,QAAQ,OAAO,UAAU;AAAA,IAC3B;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,OACE,OACA,YACA,GACU;AACV,QAAI,WAAW,WAAW,EAAG,QAAO,CAAC;AAErC,UAAM,WAAqB,CAAC;AAC5B,UAAM,YAAY,IAAI,IAAI,WAAW,IAAI,CAAC,GAAG,MAAM,CAAC,CAAC;AAGrD,QAAI,UAAU;AACd,QAAI,YAAY;AAEhB,aAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,KAAK;AAC1C,UAAI,WAAW,CAAC,EAAE,QAAQ,WAAW;AACnC,oBAAY,WAAW,CAAC,EAAE;AAC1B,kBAAU;AAAA,MACZ;AAAA,IACF;AAEA,aAAS,KAAK,WAAW,OAAO,EAAE,EAAE;AACpC,cAAU,OAAO,OAAO;AAGxB,WAAO,SAAS,SAAS,KAAK,UAAU,OAAO,GAAG;AAChD,UAAI,UAAU;AACd,UAAI,gBAAgB;AAEpB,iBAAW,OAAO,WAAW;AAC3B,cAAM,YAAY,WAAW,GAAG;AAGhC,cAAM,YAAY,KAAK,WAAW,OAAO,UAAU,MAAM;AAGzD,YAAI,gBAAgB;AACpB,mBAAW,cAAc,UAAU;AACjC,gBAAM,oBAAoB,WAAW,KAAK,OAAK,EAAE,OAAO,UAAU;AAClE,gBAAM,MAAM,KAAK,WAAW,UAAU,QAAQ,kBAAkB,MAAM;AACtE,0BAAgB,KAAK,IAAI,eAAe,GAAG;AAAA,QAC7C;AAGA,cAAM,MACJ,KAAK,OAAO,SAAS,aACpB,IAAI,KAAK,OAAO,UAAU;AAE7B,YAAI,MAAM,SAAS;AACjB,oBAAU;AACV,0BAAgB;AAAA,QAClB;AAAA,MACF;AAEA,UAAI,kBAAkB,IAAI;AACxB,iBAAS,KAAK,WAAW,aAAa,EAAE,EAAE;AAC1C,kBAAU,OAAO,aAAa;AAAA,MAChC,OAAO;AACL;AAAA,MACF;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKQ,WAAW,GAAiB,GAAyB;AAC3D,QAAI,KAAK,OAAO,WAAW,UAAU;AACnC,aAAO,KAAK,iBAAiB,GAAG,CAAC;AAAA,IACnC,OAAO;AAEL,YAAM,OAAO,KAAK,kBAAkB,GAAG,CAAC;AACxC,aAAO,KAAK,IAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEQ,iBAAiB,GAAiB,GAAyB;AACjE,QAAI,aAAa;AACjB,QAAI,QAAQ;AACZ,QAAI,QAAQ;AAEZ,aAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,oBAAc,EAAE,CAAC,IAAI,EAAE,CAAC;AACxB,eAAS,EAAE,CAAC,IAAI,EAAE,CAAC;AACnB,eAAS,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,IACrB;AAEA,WAAO,cAAc,KAAK,KAAK,KAAK,IAAI,KAAK,KAAK,KAAK;AAAA,EACzD;AAAA,EAEQ,kBAAkB,GAAiB,GAAyB;AAClE,QAAI,MAAM;AACV,aAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,YAAM,OAAO,EAAE,CAAC,IAAI,EAAE,CAAC;AACvB,aAAO,OAAO;AAAA,IAChB;AACA,WAAO,KAAK,KAAK,GAAG;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA,EAKA,UAAU,QAAsB;AAC9B,SAAK,OAAO,SAAS,KAAK,IAAI,GAAG,KAAK,IAAI,GAAG,MAAM,CAAC;AAAA,EACtD;AACF;AASO,IAAM,oBAAN,MAAwB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAO7B,OAAO,SACL,SACA,WACgB;AAChB,QAAI,QAAQ,WAAW,EAAG,QAAO,CAAC;AAElC,UAAM,cAAc,QAAQ,CAAC,EAAE;AAC/B,QAAI,aAAa,YAAa,QAAO;AAGrC,UAAM,SAAS,QAAQ,IAAI,OAAK,MAAM,KAAK,CAAC,CAAC;AAG7C,UAAM,OAAO,KAAK,YAAY,MAAM;AACpC,UAAM,WAAW,OAAO;AAAA,MAAI,SAC1B,IAAI,IAAI,CAAC,KAAK,MAAM,MAAM,KAAK,CAAC,CAAC;AAAA,IACnC;AAGA,UAAM,MAAM,KAAK,kBAAkB,QAAQ;AAG3C,UAAM,eAAe,KAAK,eAAe,KAAK,SAAS;AAGvD,UAAM,aAAa,SAAS,IAAI,SAAO;AACrC,YAAM,YAAY,IAAI,aAAa,SAAS;AAC5C,eAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,YAAI,MAAM;AACV,iBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,iBAAO,IAAI,CAAC,IAAI,aAAa,CAAC,EAAE,CAAC;AAAA,QACnC;AACA,kBAAU,CAAC,IAAI;AAAA,MACjB;AACA,aAAO;AAAA,IACT,CAAC;AAED,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,OAAe,YAAY,QAA8B;AACvD,UAAM,IAAI,OAAO;AACjB,UAAM,MAAM,OAAO,CAAC,EAAE;AACtB,UAAM,OAAO,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAElC,eAAW,OAAO,QAAQ;AACxB,eAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,aAAK,CAAC,KAAK,IAAI,CAAC;AAAA,MAClB;AAAA,IACF;AAEA,WAAO,KAAK,IAAI,OAAK,IAAI,CAAC;AAAA,EAC5B;AAAA;AAAA;AAAA;AAAA,EAKA,OAAe,kBAAkB,QAAgC;AAC/D,UAAM,IAAI,OAAO;AACjB,UAAM,MAAM,OAAO,CAAC,EAAE;AACtB,UAAM,MAAkB,MAAM;AAAA,MAAK,EAAE,QAAQ,IAAI;AAAA,MAAG,MAClD,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAAA,IACvB;AAEA,aAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,eAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAI,MAAM;AACV,mBAAW,OAAO,QAAQ;AACxB,iBAAO,IAAI,CAAC,IAAI,IAAI,CAAC;AAAA,QACvB;AACA,YAAI,CAAC,EAAE,CAAC,IAAI,IAAI,CAAC,EAAE,CAAC,IAAI,MAAM;AAAA,MAChC;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,OAAe,eACb,QACA,GACA,aAAqB,KACT;AACZ,UAAM,MAAM,OAAO;AACnB,UAAM,eAA2B,CAAC;AAElC,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAE1B,UAAI,IAAI,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC,EAAE,IAAI,MAAM,KAAK,OAAO,IAAI,GAAG;AAG5D,eAAS,OAAO,GAAG,OAAO,YAAY,QAAQ;AAE5C,cAAM,OAAO,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAClC,iBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,mBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,iBAAK,CAAC,KAAK,OAAO,CAAC,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,UAC/B;AAAA,QACF;AAGA,mBAAW,QAAQ,cAAc;AAC/B,cAAI,MAAM;AACV,mBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,mBAAO,KAAK,CAAC,IAAI,KAAK,CAAC;AAAA,UACzB;AACA,mBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,iBAAK,CAAC,KAAK,MAAM,KAAK,CAAC;AAAA,UACzB;AAAA,QACF;AAGA,YAAI,OAAO;AACX,mBAAW,OAAO,MAAM;AACtB,kBAAQ,MAAM;AAAA,QAChB;AACA,eAAO,KAAK,KAAK,IAAI;AAErB,YAAI,OAAO,MAAO;AAElB,YAAI,KAAK,IAAI,SAAO,MAAM,IAAI;AAAA,MAChC;AAEA,mBAAa,KAAK,CAAC;AAAA,IACrB;AAEA,WAAO;AAAA,EACT;AACF;AASO,IAAM,iBAAN,MAAqB;AAAA;AAAA;AAAA;AAAA,EAI1B,OAAO,sBACL,OACA,SACc;AACd,UAAM,eAAe,IAAI,aAAa,QAAQ,MAAM;AAGpD,QAAI,YAAY;AAChB,aAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,mBAAa,MAAM,CAAC,IAAI,MAAM,CAAC;AAAA,IACjC;AACA,gBAAY,KAAK,KAAK,SAAS;AAG/B,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,YAAM,SAAS,QAAQ,CAAC;AACxB,UAAI,aAAa;AACjB,UAAI,aAAa;AAEjB,eAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,sBAAc,MAAM,CAAC,IAAI,OAAO,CAAC;AACjC,sBAAc,OAAO,CAAC,IAAI,OAAO,CAAC;AAAA,MACpC;AAEA,mBAAa,KAAK,KAAK,UAAU;AACjC,mBAAa,CAAC,IAAI,cAAc,YAAY;AAAA,IAC9C;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,OAAO,eAAe,SAAyC;AAC7D,WAAO,QAAQ,IAAI,OAAK;AACtB,UAAI,OAAO;AACX,eAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,gBAAQ,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,MACpB;AACA,aAAO,KAAK,KAAK,IAAI;AAErB,YAAM,aAAa,IAAI,aAAa,EAAE,MAAM;AAC5C,eAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,mBAAW,CAAC,IAAI,EAAE,CAAC,IAAI;AAAA,MACzB;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;;;ACrhBO,IAAM,mBAAN,MAAuB;AAAA,EACpB,aAAkB;AAAA,EAClB,eAA6B;AAAA,EAC7B,YAA0B;AAAA,EAC1B;AAAA,EAER,YAAY,SAA0B,CAAC,GAAG;AACxC,SAAK,SAAS;AAAA,MACZ,WAAW;AAAA,MACX,UAAU;AAAA,MACV,WAAW;AAAA,MACX,WAAW;AAAA,MACX,SAAS;AAAA,MACT,GAAG;AAAA,IACL;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,kBAAgC;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA,EAKA,WAAyB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA,EAKA,MAAM,aAA4B;AAChC,QAAI,KAAK,iBAAiB,SAAU;AACpC,QAAI,KAAK,iBAAiB,WAAW;AAEnC,aAAO,KAAK,iBAAiB,WAAW;AACtC,cAAM,IAAI,QAAQ,aAAW,WAAW,SAAS,EAAE,CAAC;AAAA,MACtD;AACA;AAAA,IACF;AAEA,SAAK,eAAe;AAEpB,QAAI;AACF,UAAI,CAAC,KAAK,OAAO,SAAS;AAExB,aAAK,eAAe;AACpB;AAAA,MACF;AAGA,YAAM,aAAa,MAAM;AACzB,WAAK,aAAa,MAAM,WAAW,SAAS;AAC5C,WAAK,eAAe;AAAA,IACtB,SAAS,OAAO;AACd,WAAK,YAAY,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC;AACzE,WAAK,eAAe;AACpB,cAAQ,KAAK,+CAA+C,KAAK,UAAU,OAAO;AAAA,IAEpF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,MAAM,eACJ,OACA,MACA,QACuB;AA7G3B;AA8GI,UAAM,KAAK,WAAW;AAEtB,SAAI,UAAK,eAAL,mBAAiB,gBAAgB;AACnC,UAAI;AACF,eAAO,KAAK,WAAW,eAAe,OAAO,MAAM,QAAQ,KAAK,MAAM;AAAA,MACxE,SAAS,OAAO;AACd,gBAAQ,KAAK,gDAAgD,KAAK;AAAA,MACpE;AAAA,IACF;AAGA,WAAO,KAAK,uBAAuB,OAAO,MAAM,MAAM;AAAA,EACxD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,oBACJ,OACA,MACuB;AAvI3B;AAwII,UAAM,KAAK,WAAW;AAEtB,SAAI,UAAK,eAAL,mBAAiB,qBAAqB;AACxC,UAAI;AACF,eAAO,KAAK,WAAW,oBAAoB,OAAO,MAAM,KAAK,MAAM;AAAA,MACrE,SAAS,OAAO;AACd,gBAAQ,KAAK,qDAAqD,KAAK;AAAA,MACzE;AAAA,IACF;AAGA,WAAO,KAAK,4BAA4B,OAAO,IAAI;AAAA,EACrD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAM,oBACJ,UACA,SAA8B,CAAC,GAK7B;AApKN;AAqKI,UAAM,KAAK,WAAW;AAEtB,UAAM,aAAa;AAAA,MACjB,WAAW;AAAA,MACX,aAAa;AAAA,MACb,gBAAgB;AAAA,MAChB,GAAG;AAAA,IACL;AAEA,SAAI,UAAK,eAAL,mBAAiB,qBAAqB;AACxC,UAAI;AACF,eAAO,KAAK,WAAW,oBAAoB,UAAU,UAAU;AAAA,MACjE,SAAS,OAAO;AACd,gBAAQ,KAAK,qDAAqD,KAAK;AAAA,MACzE;AAAA,IACF;AAGA,WAAO,KAAK,4BAA4B,UAAU,UAAU;AAAA,EAC9D;AAAA;AAAA;AAAA;AAAA,EAKA,UAAgB;AACd,SAAK,aAAa;AAClB,SAAK,eAAe;AACpB,SAAK,YAAY;AAAA,EACnB;AAAA;AAAA;AAAA;AAAA,EAMQ,uBACN,OACA,MACA,QACc;AACd,UAAM,EAAE,YAAY,IAAI,IAAI,KAAK;AACjC,UAAM,SAAS,KAAK,SAAS;AAC7B,UAAM,SAAS,IAAI,aAAa,MAAM,MAAM;AAE5C,aAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK,WAAW;AAChD,YAAM,IAAI,MAAM,MAAM,GAAG,IAAI,SAAS;AACtC,UAAI,aAAa;AACjB,YAAM,UAAU,IAAI,aAAa,MAAM;AAGvC,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,cAAM,IAAI,KAAK,MAAM,IAAI,YAAY,IAAI,KAAK,SAAS;AACvD,YAAI,MAAM;AACV,iBAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,iBAAO,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,QACnB;AACA,gBAAQ,CAAC,IAAI,KAAK,IAAI,MAAM,KAAK,KAAK,SAAS,CAAC;AAChD,sBAAc,QAAQ,CAAC;AAAA,MACzB;AAGA,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,gBAAQ,CAAC,KAAM,cAAc;AAC7B,cAAM,IAAI,OAAO,MAAM,IAAI,YAAY,IAAI,KAAK,SAAS;AACzD,iBAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,iBAAO,IAAI,CAAC,KAAK,QAAQ,CAAC,IAAI,EAAE,CAAC;AAAA,QACnC;AAAA,MACF;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAAA,EAEQ,4BACN,OACA,MACc;AACd,UAAM,EAAE,YAAY,GAAK,IAAI,KAAK;AAClC,UAAM,IAAI,KAAK,IAAI,SAAS;AAC5B,UAAM,eAAe,IAAI,aAAa,KAAK,SAAS,MAAM,MAAM;AAGhE,aAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,KAAK;AAC5C,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,aAAa;AACjB,UAAI,QAAQ;AACZ,UAAI,QAAQ;AAEZ,eAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,sBAAc,MAAM,CAAC,IAAI,KAAK,SAAS,CAAC;AACxC,iBAAS,MAAM,CAAC,IAAI,MAAM,CAAC;AAC3B,iBAAS,KAAK,SAAS,CAAC,IAAI,KAAK,SAAS,CAAC;AAAA,MAC7C;AAGA,YAAM,YAAY,KAAK,KAAK,QAAQ,QAAQ,IAAI,UAAU;AAG1D,YAAM,WAAW,KAAK,MAAM,IAAI,IAAI,IAAI,YAAY,SAAS;AAG7D,mBAAa,CAAC,IAAI,KAAK,IAAI;AAAA,IAC7B;AAEA,WAAO;AAAA,EACT;AAAA,EAEQ,4BACN,UACA,QAKC;AACD,UAAM,EAAE,YAAY,KAAK,cAAc,IAAI,iBAAiB,EAAE,IAAI;AAClE,UAAM,eAID,CAAC;AACN,UAAM,OAAO,oBAAI,IAAY;AAG7B,aAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,UAAI,KAAK,IAAI,CAAC,EAAG;AAEjB,YAAM,UAA0B,CAAC,SAAS,CAAC,CAAC;AAC5C,WAAK,IAAI,CAAC;AAEV,eAAS,IAAI,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AAC5C,YAAI,KAAK,IAAI,CAAC,EAAG;AAGjB,cAAM,aAAa,KAAK,iBAAiB,SAAS,CAAC,GAAG,SAAS,CAAC,CAAC;AAEjE,YAAI,aAAa,WAAW;AAC1B,kBAAQ,KAAK,SAAS,CAAC,CAAC;AACxB,eAAK,IAAI,CAAC;AAAA,QACZ;AAAA,MACF;AAGA,UAAI,QAAQ,UAAU,gBAAgB;AAEpC,cAAM,WAAW,IAAI,aAAa,SAAS,CAAC,EAAE,MAAM;AACpD,mBAAW,OAAO,SAAS;AACzB,mBAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,qBAAS,CAAC,KAAK,IAAI,CAAC,IAAI,QAAQ;AAAA,UAClC;AAAA,QACF;AAGA,YAAI,OAAO;AACX,iBAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,kBAAQ,SAAS,CAAC,IAAI,SAAS,CAAC;AAAA,QAClC;AACA,eAAO,KAAK,KAAK,IAAI;AACrB,YAAI,OAAO,GAAG;AACZ,mBAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,qBAAS,CAAC,KAAK;AAAA,UACjB;AAAA,QACF;AAEA,qBAAa,KAAK;AAAA,UAChB,QAAQ;AAAA,UACR,OAAO,QAAQ;AAAA,UACf,SAAS;AAAA,QACX,CAAC;AAAA,MACH;AAEA,UAAI,aAAa,UAAU,YAAa;AAAA,IAC1C;AAEA,WAAO;AAAA,EACT;AAAA,EAEQ,iBAAiB,GAAiB,GAAyB;AACjE,QAAI,MAAM;AACV,QAAI,QAAQ;AACZ,QAAI,QAAQ;AAEZ,aAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,aAAO,EAAE,CAAC,IAAI,EAAE,CAAC;AACjB,eAAS,EAAE,CAAC,IAAI,EAAE,CAAC;AACnB,eAAS,EAAE,CAAC,IAAI,EAAE,CAAC;AAAA,IACrB;AAEA,UAAM,cAAc,KAAK,KAAK,QAAQ,KAAK;AAC3C,WAAO,cAAc,IAAI,MAAM,cAAc;AAAA,EAC/C;AACF;AAKO,SAAS,gBAAgB,QAA4C;AAC1E,SAAO,IAAI,iBAAiB,MAAM;AACpC;AAKO,SAAS,sBAAwC;AACtD,SAAO,IAAI,iBAAiB;AAAA,IAC1B,WAAW;AAAA,IACX,UAAU;AAAA,IACV,WAAW;AAAA,IACX,SAAS;AAAA,EACX,CAAC;AACH;AAKO,SAAS,0BAA4C;AAC1D,SAAO,IAAI,iBAAiB;AAAA,IAC1B,WAAW;AAAA,IACX,UAAU;AAAA,IACV,WAAW;AAAA,IACX,SAAS;AAAA,EACX,CAAC;AACH;;;AChTO,SAAS,iBAAiB;AAC/B,SAAO;AAAA,IACL,WAAW,eAAe;AAAA,IAC1B,kBAAkB,sBAAsB;AAAA,IACxC,YAAY,OAAQ,WAAmB,WAAW;AAAA,IAClD,UAAUC,gBAAe;AAAA,IACzB,mBAAmB,OAAO,sBAAsB;AAAA,EAClD;AACF;AAKA,eAAeA,kBAAmC;AAChD,MAAI;AAEF,QAAI,OAAQ,WAAmB,gBAAgB,aAAa;AAC1D,aAAO;AAAA,IACT;AAGA,UAAM,WAAW,IAAI,WAAW;AAAA,MAC9B;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAC1C;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAC1C;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAC1C;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,MAAM;AAAA,IACtC,CAAC;AAED,UAAM,KAAM,WAAmB;AAC/B,UAAM,SAAS,MAAM,GAAG,YAAY,QAAQ;AAC5C,WAAO,kBAAkB,GAAG;AAAA,EAC9B,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AASO,IAAM,uBAAuB;AAAA,EAClC,IAAI,EAAE,SAAS,MAAM;AAAA,EACrB,MAAM,EAAE,SAAS,MAAM;AAAA,EACvB,KAAK,EAAE,SAAS,MAAM,UAAU,EAAE;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM,QAAQ,IAAI;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM;AACxB;AAKO,IAAM,wBAAwB;AAAA,EACnC,IAAI,EAAE,SAAS,MAAM,YAAY,EAAE;AAAA,EACnC,MAAM,EAAE,SAAS,MAAM,GAAG,GAAG;AAAA,EAC7B,KAAK,EAAE,SAAS,MAAM,UAAU,EAAE;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM,QAAQ,IAAI;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM;AACxB;AAKO,IAAM,uBAAuB;AAAA,EAClC,IAAI,EAAE,SAAS,MAAM,YAAY,GAAG;AAAA,EACpC,MAAM,EAAE,SAAS,MAAM,GAAG,GAAG;AAAA,EAC7B,KAAK,EAAE,SAAS,MAAM,UAAU,EAAE;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM,QAAQ,IAAI;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM,WAAW,IAAI;AACvC;AAKO,IAAM,0BAA0B;AAAA,EACrC,IAAI,EAAE,SAAS,MAAM,YAAY,GAAG;AAAA;AAAA,EACpC,MAAM,EAAE,SAAS,MAAM,GAAG,EAAE;AAAA;AAAA,EAC5B,KAAK,EAAE,SAAS,MAAM;AAAA,EACtB,KAAK,EAAE,SAAS,MAAM;AAAA,EACtB,KAAK,EAAE,SAAS,MAAM,WAAW,GAAG;AAAA;AACtC;AAKO,IAAM,yBAAyB;AAAA,EACpC,IAAI,EAAE,SAAS,MAAM;AAAA;AAAA,EACrB,MAAM,EAAE,SAAS,MAAM,GAAG,IAAI,UAAU,IAAI;AAAA;AAAA,EAC5C,KAAK,EAAE,SAAS,MAAM;AAAA,EACtB,KAAK,EAAE,SAAS,MAAM;AAAA,EACtB,KAAK,EAAE,SAAS,MAAM;AACxB;AAKO,IAAM,2BAA2B;AAAA,EACtC,IAAI,EAAE,SAAS,MAAM;AAAA;AAAA,EACrB,MAAM,EAAE,SAAS,MAAM,GAAG,IAAI,gBAAgB,IAAI;AAAA;AAAA,EAClD,KAAK,EAAE,SAAS,MAAM,UAAU,EAAE;AAAA;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM,QAAQ,IAAI;AAAA;AAAA,EAClC,KAAK,EAAE,SAAS,MAAM;AAAA;AACxB;AAMO,IAAM,UAAU;AAAA,EACrB,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,UAAU;AAAA,EACV,MAAM;AACR;AASO,SAAS,oBACd,YACA,WACA,QAMA;AAzNF;AA0NE,MAAI,cAAc,aAAa,YAAY;AAG3C,OAAI,YAAO,OAAP,mBAAW,SAAS;AACtB,UAAM,aAAa,OAAO,GAAG,cAAc;AAC3C,kBAAc,cAAc,aAAa;AAAA,EAC3C;AAGA,OAAI,YAAO,QAAP,mBAAY,SAAS;AACvB,UAAM,YAAY,OAAO,IAAI,aAAa,YAAY;AACtD,kBAAc,aAAa,YAAY;AAAA,EACzC;AAGA,MAAI,aAAa;AACjB,OAAI,YAAO,SAAP,mBAAa,SAAS;AACxB,UAAM,IAAI,OAAO,KAAK,KAAK;AAC3B,UAAM,iBAAiB,IAAI;AAC3B,iBAAa,aAAa,iBAAiB;AAAA,EAC7C;AAEA,QAAM,QAAQ,cAAc;AAE5B,SAAO;AAAA,IACL,SAAS;AAAA,IACT,OAAO;AAAA,IACP;AAAA,IACA,SAAS,SAAS,OAAO;AAAA,EAC3B;AACF;AAKO,SAAS,gBAAgB,YAAoB,WAAmB;AACrE,MAAI,aAAa,KAAM;AACrB,WAAO;AAAA,MACL,MAAM;AAAA,MACN,QAAQ;AAAA,MACR,QAAQ;AAAA,IACV;AAAA,EACF,WAAW,aAAa,KAAO;AAC7B,WAAO;AAAA,MACL,MAAM;AAAA,MACN,QAAQ;AAAA,MACR,QAAQ;AAAA,IACV;AAAA,EACF,OAAO;AACL,WAAO;AAAA,MACL,MAAM;AAAA,MACN,QAAQ;AAAA,MACR,QAAQ;AAAA,IACV;AAAA,EACF;AACF;AAKA,eAAsB,gBACpB,UACA,aAAqB,KACrB,IAAY,IACZ,YAAoB,KAQnB;AACD,QAAM,QAAkB,CAAC;AAEzB,WAAS,IAAI,GAAG,IAAI,YAAY,KAAK;AACnC,UAAM,QAAQ,IAAI,aAAa,SAAS;AACxC,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,YAAM,CAAC,IAAI,KAAK,OAAO,IAAI;AAAA,IAC7B;AAEA,UAAM,QAAQ,YAAY,IAAI;AAC9B,aAAS,OAAO,CAAC;AACjB,UAAM,MAAM,YAAY,IAAI;AAE5B,UAAM,KAAK,MAAM,KAAK;AAAA,EACxB;AAEA,QAAM,KAAK,CAAC,GAAG,MAAM,IAAI,CAAC;AAE1B,SAAO;AAAA,IACL,WAAW,MAAM,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI,MAAM;AAAA,IACpD,WAAW,MAAM,CAAC;AAAA,IAClB,WAAW,MAAM,MAAM,SAAS,CAAC;AAAA,IACjC,OAAO,MAAM,KAAK,MAAM,MAAM,SAAS,GAAG,CAAC;AAAA,IAC3C,OAAO,MAAM,KAAK,MAAM,MAAM,SAAS,IAAI,CAAC;AAAA,IAC5C,OAAO,MAAM,KAAK,MAAM,MAAM,SAAS,IAAI,CAAC;AAAA,EAC9C;AACF;",
   "names": ["fDist", "detectWasmSIMD"]
 }

package/dist/agentdb.browser.min.js

@@ -1,3 +1,3 @@
-/*! AgentDB Browser Bundle v2.0.0-alpha.2.10 | MIT | https://agentdb.ruv.io */
+/*! AgentDB Browser Bundle v2.0.0-alpha.2.13 | MIT | https://agentdb.ruv.io */
 var L=Object.defineProperty;var z=(h,t)=>()=>(h&&(t=h(h=0)),t);var H=(h,t)=>{for(var e in t)L(h,e,{get:t[e],enumerable:!0})};var I={};H(I,{initWASM:()=>j,wasmLoadError:()=>P,wasmModule:()=>p});async function j(){return p||M||(M=(async()=>{try{if(typeof WebAssembly>"u")throw new Error("WebAssembly not supported in this browser");let h=await _();return console.log(`WASM SIMD support: ${h}`),p={flashAttention:N(),hyperbolicAttention:D(),memoryConsolidation:k(),simdSupported:h},console.log("\u2705 WASM attention module loaded"),p}catch(h){return P=h,console.warn("\u26A0\uFE0F  WASM loading failed, using fallback:",h.message),p={flashAttention:N(),hyperbolicAttention:D(),memoryConsolidation:k(),simdSupported:!1},p}finally{M=null}})(),M)}async function _(){try{let h=new Uint8Array([0,97,115,109,1,0,0,0,1,5,1,96,0,1,123,3,2,1,0,10,10,1,8,0,253,12,253,12,253,84,11]);return await WebAssembly.instantiate(h)instanceof WebAssembly.Instance}catch{return!1}}function N(){return(h,t,e,o={})=>{let{dim:n=384,numHeads:r=4,blockSize:s=64}=o,i=t.length/n,a=new Float32Array(h.length);for(let c=0;c<h.length;c+=n){let u=h.slice(c,c+n),m=0,d=new Float32Array(i);for(let l=0;l<i;l++){let f=t.slice(l*n,(l+1)*n),b=0;for(let g=0;g<n;g++)b+=u[g]*f[g];d[l]=Math.exp(b/Math.sqrt(n)),m+=d[l]}for(let l=0;l<i;l++){d[l]/=m;let f=e.slice(l*n,(l+1)*n);for(let b=0;b<n;b++)a[c+b]+=d[l]*f[b]}}return a}}function D(){return(h,t,e={})=>{let{curvature:o=-1}=e,n=Math.abs(o),r=new Float32Array(t.length/h.length);for(let s=0;s<r.length;s++){let i=s*h.length,a=0,c=0,u=0;for(let l=0;l<h.length;l++)a+=h[l]*t[i+l],c+=h[l]*h[l],u+=t[i+l]*t[i+l];let m=Math.sqrt(c+u-2*a),d=Math.acosh(1+2*n*m*m);r[s]=1/(1+d)}return r}}function k(){return(h,t={})=>{let{threshold:e=.8,maxClusters:o=10}=t,n=[],r=new Set;for(let s=0;s<h.length;s++){if(r.has(s))continue;let i=[h[s]];r.add(s);for(let c=s+1;c<h.length;c++){if(r.has(c))continue;let u=0,m=0,d=0;for(let f=0;f<h[s].length;f++)u+=h[s][f]*h[c][f],m+=h[s][f]*h[s][f],d+=h[c][f]*h[c][f];u/(Math.sqrt(m*d)||1)>e&&(i.push(h[c]),r.add(c))}let a=new Float32Array(h[s].length);for(let c of i)for(let u=0;u<a.length;u++)a[u]+=c[u]/i.length;if(n.push({memory:a,count:i.size,members:i}),n.length>=o)break}return n}}var p,M,P,W=z(()=>{"use strict";p=null,M=null,P=null});var y=class{config;codebook=null;trained=!1;constructor(t){if(this.config={dimension:t.dimension,numSubvectors:t.numSubvectors,numCentroids:t.numCentroids,maxIterations:t.maxIterations||50,convergenceThreshold:t.convergenceThreshold||1e-4},this.config.dimension%this.config.numSubvectors!==0)throw new Error(`Dimension ${this.config.dimension} must be divisible by numSubvectors ${this.config.numSubvectors}`)}async train(t){if(t.length===0)throw new Error("Training requires at least one vector");let e=this.config.dimension/this.config.numSubvectors,o=[];console.log(`[PQ] Training ${this.config.numSubvectors} subvectors with ${this.config.numCentroids} centroids each...`);for(let n=0;n<this.config.numSubvectors;n++){let r=n*e,s=r+e,i=t.map(c=>c.slice(r,s)),a=await this.kMeans(i,this.config.numCentroids);o.push(...a),((n+1)%4===0||n===this.config.numSubvectors-1)&&console.log(`[PQ] Trained ${n+1}/${this.config.numSubvectors} subvectors`)}this.codebook={subvectorDim:e,numSubvectors:this.config.numSubvectors,numCentroids:this.config.numCentroids,centroids:o},this.trained=!0,console.log("[PQ] Training complete")}async kMeans(t,e){let o=t[0].length,n=t.length,r=this.kMeansPlusPlus(t,e),s=new Uint32Array(n),i=1/0;for(let a=0;a<this.config.maxIterations;a++){let c=0;for(let d=0;d<n;d++){let l=1/0,f=0;for(let b=0;b<e;b++){let g=this.squaredDistance(t[d],r[b]);g<l&&(l=g,f=b)}s[d]=f,c+=l}if(Math.abs(i-c)<this.config.convergenceThreshold)break;i=c;let u=new Uint32Array(e),m=Array.from({length:e},()=>new Float32Array(o));for(let d=0;d<n;d++){let l=s[d];u[l]++;for(let f=0;f<o;f++)m[l][f]+=t[d][f]}for(let d=0;d<e;d++)if(u[d]>0)for(let l=0;l<o;l++)r[d][l]=m[d][l]/u[d]}return r}kMeansPlusPlus(t,e){let o=t.length,n=t[0].length,r=[],s=Math.floor(Math.random()*o);r.push(new Float32Array(t[s]));for(let i=1;i<e;i++){let a=new Float32Array(o),c=0;for(let m=0;m<o;m++){let d=1/0;for(let l of r){let f=this.squaredDistance(t[m],l);d=Math.min(d,f)}a[m]=d,c+=d}let u=Math.random()*c;for(let m=0;m<o;m++)if(u-=a[m],u<=0){r.push(new Float32Array(t[m]));break}}return r}compress(t){if(!this.trained||!this.codebook)throw new Error("Codebook must be trained before compression");let e=new Uint8Array(this.config.numSubvectors),o=this.codebook.subvectorDim,n=0;for(let r=0;r<t.length;r++)n+=t[r]*t[r];n=Math.sqrt(n);for(let r=0;r<this.config.numSubvectors;r++){let s=r*o,i=t.slice(s,s+o),a=1/0,c=0,u=r*this.config.numCentroids;for(let m=0;m<this.config.numCentroids;m++){let d=this.codebook.centroids[u+m],l=this.squaredDistance(i,d);l<a&&(a=l,c=m)}e[r]=c}return{codes:e,norm:n}}decompress(t){if(!this.codebook)throw new Error("Codebook not available");let e=new Float32Array(this.config.dimension),o=this.codebook.subvectorDim;for(let n=0;n<this.config.numSubvectors;n++){let r=t.codes[n],s=n*this.config.numCentroids,i=this.codebook.centroids[s+r],a=n*o;for(let c=0;c<o;c++)e[a+c]=i[c]}return e}asymmetricDistance(t,e){if(!this.codebook)throw new Error("Codebook not available");let o=0,n=this.codebook.subvectorDim;for(let r=0;r<this.config.numSubvectors;r++){let s=e.codes[r],i=r*this.config.numCentroids,a=this.codebook.centroids[i+s],c=r*n,u=t.slice(c,c+n);o+=this.squaredDistance(u,a)}return Math.sqrt(o)}batchCompress(t){return t.map(e=>this.compress(e))}getCompressionRatio(){let t=this.config.dimension*4,e=this.config.numSubvectors+4;return t/e}exportCodebook(){if(!this.codebook)throw new Error("No codebook to export");return JSON.stringify({config:this.config,codebook:{subvectorDim:this.codebook.subvectorDim,numSubvectors:this.codebook.numSubvectors,numCentroids:this.codebook.numCentroids,centroids:this.codebook.centroids.map(t=>Array.from(t))}})}importCodebook(t){let e=JSON.parse(t);this.config=e.config,this.codebook={subvectorDim:e.codebook.subvectorDim,numSubvectors:e.codebook.numSubvectors,numCentroids:e.codebook.numCentroids,centroids:e.codebook.centroids.map(o=>new Float32Array(o))},this.trained=!0}squaredDistance(t,e){let o=0;for(let n=0;n<t.length;n++){let r=t[n]-e[n];o+=r*r}return o}getStats(){let t=this.getCompressionRatio(),e=this.config.numSubvectors+4,o=this.codebook?this.config.numSubvectors*this.config.numCentroids*(this.config.dimension/this.config.numSubvectors)*4:0;return{trained:this.trained,compressionRatio:t,memoryPerVector:e,codebookSize:o}}};function B(h){return new y({dimension:h,numSubvectors:8,numCentroids:256,maxIterations:50})}function Q(h){return new y({dimension:h,numSubvectors:16,numCentroids:256,maxIterations:50})}function R(h){return new y({dimension:h,numSubvectors:32,numCentroids:256,maxIterations:50})}var S=class{items=[];push(t,e){this.items.push({item:t,priority:e}),this.bubbleUp(this.items.length-1)}pop(){if(this.items.length===0)return;let t=this.items[0].item,e=this.items.pop();return this.items.length>0&&(this.items[0]=e,this.bubbleDown(0)),t}peek(){var t;return(t=this.items[0])==null?void 0:t.item}size(){return this.items.length}bubbleUp(t){for(;t>0;){let e=Math.floor((t-1)/2);if(this.items[t].priority>=this.items[e].priority)break;[this.items[t],this.items[e]]=[this.items[e],this.items[t]],t=e}}bubbleDown(t){for(;;){let e=2*t+1,o=2*t+2,n=t;if(e<this.items.length&&this.items[e].priority<this.items[n].priority&&(n=e),o<this.items.length&&this.items[o].priority<this.items[n].priority&&(n=o),n===t)break;[this.items[t],this.items[n]]=[this.items[n],this.items[t]],t=n}}},A=class{config;nodes=new Map;entryPoint=null;currentId=0;ml;constructor(t={}){this.config={dimension:t.dimension||384,M:t.M||16,efConstruction:t.efConstruction||200,efSearch:t.efSearch||50,ml:t.ml||1/Math.log(2),maxLayers:t.maxLayers||16,distanceFunction:t.distanceFunction||"cosine"},this.ml=this.config.ml}add(t,e){let o=e!==void 0?e:this.currentId++,n=this.randomLevel(),r={id:o,vector:t,level:n,connections:new Map};for(let a=0;a<=n;a++)r.connections.set(a,[]);if(this.entryPoint===null)return this.entryPoint=o,this.nodes.set(o,r),o;let s=this.entryPoint,i=s;for(let a=this.nodes.get(s).level;a>n;a--)i=this.searchLayer(t,i,1,a)[0];for(let a=Math.min(n,this.nodes.get(s).level);a>=0;a--){let c=this.searchLayer(t,i,this.config.efConstruction,a),u=a===0?this.config.M*2:this.config.M,m=this.selectNeighbors(t,c,u);for(let d of m){this.connect(o,d,a),this.connect(d,o,a);let l=this.nodes.get(d),f=l.connections.get(a);if(f.length>u){let b=this.selectNeighbors(l.vector,f,u);l.connections.set(a,b)}}i=c[0]}return n>this.nodes.get(this.entryPoint).level&&(this.entryPoint=o),this.nodes.set(o,r),o}search(t,e,o){if(this.entryPoint===null)return[];o=o||Math.max(this.config.efSearch,e);let n=this.entryPoint,r=n;for(let i=this.nodes.get(n).level;i>0;i--)r=this.searchLayer(t,r,1,i)[0];return this.searchLayer(t,r,o,0).slice(0,e).map(i=>({id:i,distance:this.distance(t,this.nodes.get(i).vector),vector:this.nodes.get(i).vector}))}searchLayer(t,e,o,n){let r=new Set,s=new S,i=new S,a=this.distance(t,this.nodes.get(e).vector);for(s.push(e,a),i.push(e,-a),r.add(e);s.size()>0;){let u=s.pop(),m=-i.peek();if(this.distance(t,this.nodes.get(u).vector)>m)break;let l=this.nodes.get(u).connections.get(n)||[];for(let f of l){if(r.has(f))continue;r.add(f);let b=this.distance(t,this.nodes.get(f).vector),g=-i.peek();(b<g||i.size()<o)&&(s.push(f,b),i.push(f,-b),i.size()>o&&i.pop())}}let c=[];for(;i.size()>0;)c.unshift(i.pop());return c}selectNeighbors(t,e,o){return e.length<=o?e:e.map(r=>({id:r,distance:this.distance(t,this.nodes.get(r).vector)})).sort((r,s)=>r.distance-s.distance).slice(0,o).map(r=>r.id)}connect(t,e,o){let r=this.nodes.get(t).connections.get(o);r.includes(e)||r.push(e)}randomLevel(){let t=0;for(;Math.random()<this.ml&&t<this.config.maxLayers-1;)t++;return t}distance(t,e){switch(this.config.distanceFunction){case"cosine":return 1-this.cosineSimilarity(t,e);case"euclidean":return this.euclideanDistance(t,e);case"manhattan":return this.manhattanDistance(t,e);default:return 1-this.cosineSimilarity(t,e)}}cosineSimilarity(t,e){let o=0,n=0,r=0;for(let s=0;s<t.length;s++)o+=t[s]*e[s],n+=t[s]*t[s],r+=e[s]*e[s];return o/(Math.sqrt(n)*Math.sqrt(r))}euclideanDistance(t,e){let o=0;for(let n=0;n<t.length;n++){let r=t[n]-e[n];o+=r*r}return Math.sqrt(o)}manhattanDistance(t,e){let o=0;for(let n=0;n<t.length;n++)o+=Math.abs(t[n]-e[n]);return o}getStats(){if(this.nodes.size===0)return{numNodes:0,numLayers:0,avgConnections:0,entryPointLevel:0,memoryBytes:0};let t=Math.max(...Array.from(this.nodes.values()).map(a=>a.level)),e=0;for(let a of this.nodes.values())for(let c of a.connections.values())e+=c.length;let o=e/this.nodes.size,n=this.config.dimension*4,r=o*4,i=this.nodes.size*(n+r+100);return{numNodes:this.nodes.size,numLayers:t+1,avgConnections:o,entryPointLevel:this.entryPoint?this.nodes.get(this.entryPoint).level:0,memoryBytes:i}}export(){let t={config:this.config,entryPoint:this.entryPoint,currentId:this.currentId,nodes:Array.from(this.nodes.entries()).map(([e,o])=>({id:e,vector:Array.from(o.vector),level:o.level,connections:Array.from(o.connections.entries())}))};return JSON.stringify(t)}import(t){let e=JSON.parse(t);this.config=e.config,this.entryPoint=e.entryPoint,this.currentId=e.currentId,this.nodes.clear();for(let o of e.nodes){let n={id:o.id,vector:new Float32Array(o.vector),level:o.level,connections:new Map(o.connections)};this.nodes.set(o.id,n)}}clear(){this.nodes.clear(),this.entryPoint=null,this.currentId=0}size(){return this.nodes.size}};function q(h){return new A({dimension:h,M:16,efConstruction:200,efSearch:50})}function G(h){return new A({dimension:h,M:8,efConstruction:100,efSearch:30})}function O(h){return new A({dimension:h,M:32,efConstruction:400,efSearch:100})}var w=class{config;nodes=new Map;edges=[];attentionWeights=new Map;constructor(t={}){this.config={hiddenDim:t.hiddenDim||64,numHeads:t.numHeads||4,dropout:t.dropout||.1,learningRate:t.learningRate||.01,attentionType:t.attentionType||"gat"}}addNode(t,e){this.nodes.set(t,{id:t,features:e,neighbors:[]})}addEdge(t,e,o=1){this.edges.push({from:t,to:e,weight:o});let n=this.nodes.get(t),r=this.nodes.get(e);n&&!n.neighbors.includes(e)&&n.neighbors.push(e),r&&!r.neighbors.includes(t)&&r.neighbors.push(t)}graphAttention(t){let e=this.nodes.get(t);if(!e)throw new Error(`Node ${t} not found`);let o=e.neighbors;if(o.length===0)return e.features;let n=Math.floor(this.config.hiddenDim/this.config.numHeads),r=new Float32Array(this.config.hiddenDim);for(let s=0;s<this.config.numHeads;s++){let i=0,a=new Float32Array(n);for(let u of o){let m=this.nodes.get(u),d=this.computeAttentionScore(e.features,m.features,s);i+=d;for(let l=0;l<n&&l<m.features.length;l++)a[l]+=d*m.features[l]}if(i>0)for(let u=0;u<n;u++)a[u]/=i;let c=s*n;for(let u=0;u<n;u++)r[c+u]=a[u]}for(let s=0;s<r.length;s++)r[s]=r[s]>0?r[s]:.01*r[s];return r}computeAttentionScore(t,e,o){let n=0,r=Math.min(t.length,e.length);for(let s=0;s<r;s++)n+=t[s]*e[s];return Math.exp(n/Math.sqrt(r))}messagePass(){let t=new Map;for(let[e]of this.nodes)t.set(e,this.graphAttention(e));return t}update(t){for(let[e,o]of t){let n=this.nodes.get(e);n&&(n.features=o)}}computeGraphEmbedding(t,e=2){new Map().set(t,this.nodes.get(t).features);for(let n=0;n<e;n++){let r=this.messagePass();this.update(r)}return this.nodes.get(t).features}getStats(){return{numNodes:this.nodes.size,numEdges:this.edges.length,avgDegree:this.edges.length/Math.max(this.nodes.size,1),config:this.config}}},x=class{config;constructor(t={}){this.config={lambda:t.lambda||.7,metric:t.metric||"cosine"}}rerank(t,e,o){if(e.length===0)return[];let n=[],r=new Set(e.map((a,c)=>c)),s=0,i=-1/0;for(let a=0;a<e.length;a++)e[a].score>i&&(i=e[a].score,s=a);for(n.push(e[s].id),r.delete(s);n.length<o&&r.size>0;){let a=-1/0,c=-1;for(let u of r){let m=e[u],d=this.similarity(t,m.vector),l=-1/0;for(let b of n){let g=e.find(E=>E.id===b),T=this.similarity(m.vector,g.vector);l=Math.max(l,T)}let f=this.config.lambda*d-(1-this.config.lambda)*l;f>a&&(a=f,c=u)}if(c!==-1)n.push(e[c].id),r.delete(c);else break}return n}similarity(t,e){return this.config.metric==="cosine"?this.cosineSimilarity(t,e):1/(1+this.euclideanDistance(t,e))}cosineSimilarity(t,e){let o=0,n=0,r=0;for(let s=0;s<t.length;s++)o+=t[s]*e[s],n+=t[s]*t[s],r+=e[s]*e[s];return o/(Math.sqrt(n)*Math.sqrt(r))}euclideanDistance(t,e){let o=0;for(let n=0;n<t.length;n++){let r=t[n]-e[n];o+=r*r}return Math.sqrt(o)}setLambda(t){this.config.lambda=Math.max(0,Math.min(1,t))}},F=class{static compress(t,e){if(t.length===0)return[];let o=t[0].length;if(e>=o)return t;let n=t.map(u=>Array.from(u)),r=this.computeMean(n),s=n.map(u=>u.map((m,d)=>m-r[d])),i=this.computeCovariance(s),a=this.powerIteration(i,e);return s.map(u=>{let m=new Float32Array(e);for(let d=0;d<e;d++){let l=0;for(let f=0;f<o;f++)l+=u[f]*a[d][f];m[d]=l}return m})}static computeMean(t){let e=t.length,o=t[0].length,n=new Array(o).fill(0);for(let r of t)for(let s=0;s<o;s++)n[s]+=r[s];return n.map(r=>r/e)}static computeCovariance(t){let e=t.length,o=t[0].length,n=Array.from({length:o},()=>new Array(o).fill(0));for(let r=0;r<o;r++)for(let s=0;s<=r;s++){let i=0;for(let a of t)i+=a[r]*a[s];n[r][s]=n[s][r]=i/e}return n}static powerIteration(t,e,o=100){let n=t.length,r=[];for(let s=0;s<e;s++){let i=new Array(n).fill(0).map(()=>Math.random()-.5);for(let a=0;a<o;a++){let c=new Array(n).fill(0);for(let m=0;m<n;m++)for(let d=0;d<n;d++)c[m]+=t[m][d]*i[d];for(let m of r){let d=0;for(let l=0;l<n;l++)d+=c[l]*m[l];for(let l=0;l<n;l++)c[l]-=d*m[l]}let u=0;for(let m of c)u+=m*m;if(u=Math.sqrt(u),u<1e-10)break;i=c.map(m=>m/u)}r.push(i)}return r}},C=class{static batchCosineSimilarity(t,e){let o=new Float32Array(e.length),n=0;for(let r=0;r<t.length;r++)n+=t[r]*t[r];n=Math.sqrt(n);for(let r=0;r<e.length;r++){let s=e[r],i=0,a=0;for(let c=0;c<t.length;c++)i+=t[c]*s[c],a+=s[c]*s[c];a=Math.sqrt(a),o[r]=i/(n*a)}return o}static batchNormalize(t){return t.map(e=>{let o=0;for(let r=0;r<e.length;r++)o+=e[r]*e[r];o=Math.sqrt(o);let n=new Float32Array(e.length);for(let r=0;r<e.length;r++)n[r]=e[r]/o;return n})}};var v=class{wasmModule=null;loadingState="idle";loadError=null;config;constructor(t={}){this.config={dimension:384,numHeads:4,blockSize:64,curvature:-1,useWASM:!0,...t}}getLoadingState(){return this.loadingState}getError(){return this.loadError}async initialize(){if(this.loadingState!=="loaded"){if(this.loadingState==="loading"){for(;this.loadingState==="loading";)await new Promise(t=>setTimeout(t,50));return}this.loadingState="loading";try{if(!this.config.useWASM){this.loadingState="loaded";return}let t=await Promise.resolve().then(()=>(W(),I));this.wasmModule=await t.initWASM(),this.loadingState="loaded"}catch(t){this.loadError=t instanceof Error?t:new Error(String(t)),this.loadingState="error",console.warn("WASM initialization failed, using fallback:",this.loadError.message)}}}async flashAttention(t,e,o){var n;if(await this.initialize(),(n=this.wasmModule)!=null&&n.flashAttention)try{return this.wasmModule.flashAttention(t,e,o,this.config)}catch(r){console.warn("WASM flash attention failed, using fallback:",r)}return this.flashAttentionFallback(t,e,o)}async hyperbolicAttention(t,e){var o;if(await this.initialize(),(o=this.wasmModule)!=null&&o.hyperbolicAttention)try{return this.wasmModule.hyperbolicAttention(t,e,this.config)}catch(n){console.warn("WASM hyperbolic attention failed, using fallback:",n)}return this.hyperbolicAttentionFallback(t,e)}async consolidateMemories(t,e={}){var n;await this.initialize();let o={threshold:.8,maxClusters:10,minClusterSize:1,...e};if((n=this.wasmModule)!=null&&n.memoryConsolidation)try{return this.wasmModule.memoryConsolidation(t,o)}catch(r){console.warn("WASM memory consolidation failed, using fallback:",r)}return this.consolidateMemoriesFallback(t,o)}dispose(){this.wasmModule=null,this.loadingState="idle",this.loadError=null}flashAttentionFallback(t,e,o){let{dimension:n=384}=this.config,r=e.length/n,s=new Float32Array(t.length);for(let i=0;i<t.length;i+=n){let a=t.slice(i,i+n),c=0,u=new Float32Array(r);for(let m=0;m<r;m++){let d=e.slice(m*n,(m+1)*n),l=0;for(let f=0;f<n;f++)l+=a[f]*d[f];u[m]=Math.exp(l/Math.sqrt(n)),c+=u[m]}for(let m=0;m<r;m++){u[m]/=c||1;let d=o.slice(m*n,(m+1)*n);for(let l=0;l<n;l++)s[i+l]+=u[m]*d[l]}}return s}hyperbolicAttentionFallback(t,e){let{curvature:o=-1}=this.config,n=Math.abs(o),r=new Float32Array(e.length/t.length);for(let s=0;s<r.length;s++){let i=s*t.length,a=0,c=0,u=0;for(let l=0;l<t.length;l++)a+=t[l]*e[i+l],c+=t[l]*t[l],u+=e[i+l]*e[i+l];let m=Math.sqrt(c+u-2*a),d=Math.acosh(1+2*n*m*m);r[s]=1/(1+d)}return r}consolidateMemoriesFallback(t,e){let{threshold:o=.8,maxClusters:n=10,minClusterSize:r=1}=e,s=[],i=new Set;for(let a=0;a<t.length;a++){if(i.has(a))continue;let c=[t[a]];i.add(a);for(let u=a+1;u<t.length;u++){if(i.has(u))continue;this.cosineSimilarity(t[a],t[u])>o&&(c.push(t[u]),i.add(u))}if(c.length>=r){let u=new Float32Array(t[a].length);for(let d of c)for(let l=0;l<u.length;l++)u[l]+=d[l]/c.length;let m=0;for(let d=0;d<u.length;d++)m+=u[d]*u[d];if(m=Math.sqrt(m),m>0)for(let d=0;d<u.length;d++)u[d]/=m;s.push({memory:u,count:c.length,members:c})}if(s.length>=n)break}return s}cosineSimilarity(t,e){let o=0,n=0,r=0;for(let i=0;i<t.length;i++)o+=t[i]*e[i],n+=t[i]*t[i],r+=e[i]*e[i];let s=Math.sqrt(n*r);return s>0?o/s:0}};function U(h){return new v(h)}function $(){return new v({dimension:256,numHeads:2,blockSize:32,useWASM:!0})}function V(){return new v({dimension:768,numHeads:8,blockSize:128,useWASM:!0})}function ot(){return{indexedDB:"indexedDB"in globalThis,broadcastChannel:"BroadcastChannel"in globalThis,webWorkers:typeof globalThis.Worker<"u",wasmSIMD:J(),sharedArrayBuffer:typeof SharedArrayBuffer<"u"}}async function J(){try{if(typeof globalThis.WebAssembly>"u")return!1;let h=new Uint8Array([0,97,115,109,1,0,0,0,1,5,1,96,0,1,123,3,2,1,0,10,10,1,8,0,253,12,253,12,253,84,11]),t=globalThis.WebAssembly;return await t.instantiate(h)instanceof t.Instance}catch{return!1}}var Z={pq:{enabled:!1},hnsw:{enabled:!1},gnn:{enabled:!0,numHeads:2},mmr:{enabled:!0,lambda:.7},svd:{enabled:!1}},K={pq:{enabled:!0,subvectors:8},hnsw:{enabled:!0,M:16},gnn:{enabled:!0,numHeads:4},mmr:{enabled:!0,lambda:.7},svd:{enabled:!1}},Y={pq:{enabled:!0,subvectors:16},hnsw:{enabled:!0,M:32},gnn:{enabled:!0,numHeads:4},mmr:{enabled:!0,lambda:.7},svd:{enabled:!0,targetDim:128}},st={pq:{enabled:!0,subvectors:32},hnsw:{enabled:!0,M:8},gnn:{enabled:!1},mmr:{enabled:!1},svd:{enabled:!0,targetDim:64}},it={pq:{enabled:!1},hnsw:{enabled:!0,M:32,efSearch:100},gnn:{enabled:!1},mmr:{enabled:!1},svd:{enabled:!1}},at={pq:{enabled:!1},hnsw:{enabled:!0,M:48,efConstruction:400},gnn:{enabled:!0,numHeads:8},mmr:{enabled:!0,lambda:.8},svd:{enabled:!1}},ct={major:2,minor:0,patch:0,prerelease:"alpha.2",features:"advanced",full:"2.0.0-alpha.2+advanced"};function lt(h,t,e){var s,i,a;let o=h*t*4;if((s=e.pq)!=null&&s.enabled){let c=e.pq.subvectors||8;o=h*(c+4)}if((i=e.svd)!=null&&i.enabled){let c=e.svd.targetDim||t/2;o=h*c*4}let n=0;if((a=e.hnsw)!=null&&a.enabled){let u=(e.hnsw.M||16)*1.5;n=h*u*4}let r=o+n;return{vectors:o,index:n,total:r,totalMB:r/(1024*1024)}}function ut(h,t){return h<1e3?{name:"SMALL_DATASET",config:Z,reason:"Small dataset, linear search is fast enough"}:h<1e4?{name:"MEDIUM_DATASET",config:K,reason:"Medium dataset, HNSW + PQ8 recommended"}:{name:"LARGE_DATASET",config:Y,reason:"Large dataset, aggressive compression + HNSW recommended"}}async function mt(h,t=100,e=10,o=384){let n=[];for(let r=0;r<t;r++){let s=new Float32Array(o);for(let c=0;c<o;c++)s[c]=Math.random()-.5;let i=performance.now();h(s,e);let a=performance.now();n.push(a-i)}return n.sort((r,s)=>r-s),{avgTimeMs:n.reduce((r,s)=>r+s,0)/n.length,minTimeMs:n[0],maxTimeMs:n[n.length-1],p50Ms:n[Math.floor(n.length*.5)],p95Ms:n[Math.floor(n.length*.95)],p99Ms:n[Math.floor(n.length*.99)]}}export{v as AttentionBrowser,C as BatchProcessor,w as GraphNeuralNetwork,A as HNSWIndex,Y as LARGE_DATASET_CONFIG,K as MEDIUM_DATASET_CONFIG,st as MEMORY_OPTIMIZED_CONFIG,x as MaximalMarginalRelevance,y as ProductQuantization,at as QUALITY_OPTIMIZED_CONFIG,Z as SMALL_DATASET_CONFIG,it as SPEED_OPTIMIZED_CONFIG,F as TensorCompression,ct as VERSION,mt as benchmarkSearch,V as createAccurateAttention,O as createAccurateHNSW,U as createAttention,$ as createFastAttention,G as createFastHNSW,q as createHNSW,Q as createPQ16,R as createPQ32,B as createPQ8,ot as detectFeatures,lt as estimateMemoryUsage,ut as recommendConfig};
 //# sourceMappingURL=agentdb.browser.min.js.map