@anvia/core 0.26.0 → 1.0.0-rc.2

package/dist/chunk-MELYDKWQ.js

@@ -1,337 +0,0 @@
-import {
-  cosineSimilarity,
-  embedText
-} from "./chunk-MNWK2USR.js";
-import {
-  toProviderJsonSchema
-} from "./chunk-WQKHFADH.js";
-
-// src/vector-store/index.ts
-import { z } from "zod";
-
-// src/tool/create-tool.ts
-function createTool(options) {
-  const parameters = toProviderJsonSchema(options.input);
-  const definition = () => ({
-    name: options.name,
-    description: options.description,
-    parameters
-  });
-  const call = async (args, context = {}) => {
-    const parsedArgs = options.input.parse(args);
-    const result = await options.execute(parsedArgs, context);
-    return options.output === void 0 ? result : options.output.parse(result);
-  };
-  const parseApprovalArgs = (args) => options.input.parse(args);
-  if (options.approval === void 0) {
-    return { name: options.name, definition, call, parseApprovalArgs };
-  }
-  return {
-    name: options.name,
-    approval: options.approval,
-    definition,
-    call,
-    parseApprovalArgs
-  };
-}
-
-// src/vector-store/filter.ts
-var vectorFilter = {
-  eq(key, value) {
-    return { type: "eq", key, value };
-  },
-  gt(key, value) {
-    return { type: "gt", key, value };
-  },
-  lt(key, value) {
-    return { type: "lt", key, value };
-  },
-  and(left, right) {
-    return { type: "and", filters: [left, right] };
-  },
-  or(left, right) {
-    return { type: "or", filters: [left, right] };
-  }
-};
-function matchesVectorFilter(metadata, filter) {
-  if (filter === void 0) {
-    return true;
-  }
-  if (metadata === void 0) {
-    return false;
-  }
-  switch (filter.type) {
-    case "eq":
-      return metadata[filter.key] === filter.value;
-    case "gt":
-      return compare(metadata[filter.key], filter.value) > 0;
-    case "lt":
-      return compare(metadata[filter.key], filter.value) < 0;
-    case "and":
-      return matchesVectorFilter(metadata, filter.filters[0]) && matchesVectorFilter(metadata, filter.filters[1]);
-    case "or":
-      return matchesVectorFilter(metadata, filter.filters[0]) || matchesVectorFilter(metadata, filter.filters[1]);
-  }
-}
-function compare(left, right) {
-  if (typeof left === "number" && typeof right === "number") {
-    return left - right;
-  }
-  if (typeof left === "string" && typeof right === "string") {
-    return left.localeCompare(right);
-  }
-  if (typeof left === "boolean" && typeof right === "boolean") {
-    return Number(left) - Number(right);
-  }
-  return 0;
-}
-
-// src/vector-store/lsh.ts
-var LshIndex = class {
-  constructor(dimensions, options) {
-    this.options = options;
-    const rng = seededRandom(options.seed ?? 42);
-    this.hyperplanes = [];
-    for (let index = 0; index < options.numTables * options.numHyperplanes; index += 1) {
-      const plane = Array.from({ length: dimensions }, () => rng() * 2 - 1);
-      const norm = Math.sqrt(plane.reduce((sum, value) => sum + value ** 2, 0));
-      this.hyperplanes.push(norm === 0 ? plane : plane.map((value) => value / norm));
-    }
-    this.tables = Array.from({ length: options.numTables }, () => /* @__PURE__ */ new Map());
-  }
-  options;
-  hyperplanes;
-  tables;
-  insert(id, vector) {
-    for (let table = 0; table < this.options.numTables; table += 1) {
-      const hash = this.hash(vector, table);
-      const bucket = this.tables[table]?.get(hash) ?? /* @__PURE__ */ new Set();
-      bucket.add(id);
-      this.tables[table]?.set(hash, bucket);
-    }
-  }
-  query(vector) {
-    const candidates = /* @__PURE__ */ new Set();
-    for (let table = 0; table < this.options.numTables; table += 1) {
-      const hash = this.hash(vector, table);
-      for (const id of this.tables[table]?.get(hash) ?? []) {
-        candidates.add(id);
-      }
-    }
-    return candidates;
-  }
-  hash(vector, table) {
-    let hash = "";
-    const start = table * this.options.numHyperplanes;
-    for (let offset = 0; offset < this.options.numHyperplanes; offset += 1) {
-      const plane = this.hyperplanes[start + offset];
-      const dot = vector.reduce((sum, value, index) => sum + value * (plane[index] ?? 0), 0);
-      hash += dot >= 0 ? "1" : "0";
-    }
-    return hash;
-  }
-};
-function seededRandom(seed) {
-  let state = seed >>> 0;
-  return () => {
-    state = state * 1664525 + 1013904223 >>> 0;
-    return state / 4294967296;
-  };
-}
-
-// src/vector-store/index.ts
-var InMemoryVectorStore = class _InMemoryVectorStore {
-  documents = /* @__PURE__ */ new Map();
-  indexStrategy;
-  lshIndex;
-  embeddingDimension;
-  constructor(options = {}) {
-    this.indexStrategy = options.index ?? { type: "bruteForce" };
-  }
-  static fromDocuments(documents, options = {}) {
-    const store = new _InMemoryVectorStore(options);
-    store.addDocuments(documents);
-    return store;
-  }
-  addDocuments(documents) {
-    this.validateDocumentDimensions(documents);
-    for (const document of documents) {
-      this.documents.set(document.id, document);
-    }
-    this.rebuildLshIndex();
-    return this;
-  }
-  get(id) {
-    return this.documents.get(id);
-  }
-  values() {
-    return [...this.documents.values()];
-  }
-  len() {
-    return this.documents.size;
-  }
-  isEmpty() {
-    return this.documents.size === 0;
-  }
-  index(model) {
-    return new InMemoryVectorIndex(model, this);
-  }
-  candidates(queryEmbedding) {
-    this.validateQueryDimension(queryEmbedding);
-    if (this.indexStrategy.type !== "lsh" || this.lshIndex === void 0) {
-      return this.values();
-    }
-    const candidateIds = this.lshIndex.query(queryEmbedding.vector);
-    if (candidateIds.size === 0) {
-      return this.values();
-    }
-    return [...candidateIds].flatMap((id) => {
-      const document = this.documents.get(id);
-      return document === void 0 ? [] : [document];
-    });
-  }
-  rebuildLshIndex() {
-    if (this.indexStrategy.type !== "lsh") {
-      this.lshIndex = void 0;
-      return;
-    }
-    const firstEmbedding = this.values().flatMap((document) => document.embeddings)[0];
-    if (firstEmbedding === void 0) {
-      this.lshIndex = void 0;
-      return;
-    }
-    const index = new LshIndex(firstEmbedding.vector.length, this.indexStrategy);
-    for (const document of this.documents.values()) {
-      for (const embedding of document.embeddings) {
-        index.insert(document.id, embedding.vector);
-      }
-    }
-    this.lshIndex = index;
-  }
-  validateDocumentDimensions(documents) {
-    let dimension = this.embeddingDimension;
-    for (const document of documents) {
-      for (const embedding of document.embeddings) {
-        dimension = validateEmbeddingDimension(dimension, embedding, document.id);
-      }
-    }
-    this.embeddingDimension = dimension;
-  }
-  validateQueryDimension(queryEmbedding) {
-    if (this.embeddingDimension === void 0) {
-      return;
-    }
-    validateEmbeddingDimension(this.embeddingDimension, queryEmbedding, "query");
-  }
-};
-var InMemoryVectorIndex = class {
-  constructor(model, store) {
-    this.model = model;
-    this.store = store;
-  }
-  model;
-  store;
-  async search(request) {
-    const queryEmbedding = await embedText(this.model, request.query);
-    return this.store.candidates(queryEmbedding).filter((document) => matchesVectorFilter(document.metadata, request.filter)).flatMap((document) => {
-      const score = bestScore(queryEmbedding, document.embeddings);
-      if (score === void 0) {
-        return [];
-      }
-      if (request.threshold !== void 0 && score < request.threshold) {
-        return [];
-      }
-      const result = {
-        score,
-        id: document.id,
-        document: document.document
-      };
-      if (document.metadata !== void 0) {
-        result.metadata = document.metadata;
-      }
-      return [result];
-    }).sort((left, right) => right.score - left.score).slice(0, Math.max(0, Math.trunc(request.topK)));
-  }
-  async searchIds(request) {
-    return (await this.search(request)).map(({ score, id }) => ({ score, id }));
-  }
-  async inspect(request) {
-    const limit = Math.max(0, Math.trunc(request.limit));
-    const start = Math.max(0, Math.trunc(Number(request.cursor ?? "0")));
-    const documents = this.store.values().filter((document) => matchesVectorFilter(document.metadata, request.filter));
-    const page = documents.slice(start, start + limit);
-    const nextOffset = start + page.length;
-    const result = {
-      items: page.map((document) => {
-        const item = {
-          id: document.id,
-          document: document.document
-        };
-        if (document.metadata !== void 0) {
-          item.metadata = document.metadata;
-        }
-        return item;
-      })
-    };
-    if (nextOffset < documents.length) {
-      result.nextCursor = String(nextOffset);
-    }
-    result.totalCount = documents.length;
-    return result;
-  }
-  asTool(options) {
-    return createVectorSearchTool(this, options);
-  }
-};
-function createVectorSearchTool(index, options) {
-  return createTool({
-    name: options.name,
-    description: options.description ?? "Search a vector store for documents relevant to the provided query.",
-    input: z.object({
-      query: z.string().describe("The query string to search for relevant documents."),
-      topK: z.number().int().positive().optional().describe("The maximum number of results.")
-    }),
-    output: z.array(
-      z.object({
-        score: z.number(),
-        id: z.string(),
-        document: z.any(),
-        metadata: z.record(z.string(), z.union([z.string(), z.number(), z.boolean(), z.null()])).optional()
-      })
-    ),
-    execute: ({ query, topK }) => index.search({
-      query,
-      topK: topK ?? options.topK ?? 5,
-      threshold: options.threshold,
-      filter: options.filter
-    })
-  });
-}
-function bestScore(queryEmbedding, embeddings) {
-  let best;
-  for (const embedding of embeddings) {
-    const score = cosineSimilarity(queryEmbedding.vector, embedding.vector);
-    best = best === void 0 ? score : Math.max(best, score);
-  }
-  return best;
-}
-function validateEmbeddingDimension(expectedDimension, embedding, id) {
-  if (expectedDimension === void 0) {
-    return embedding.vector.length;
-  }
-  if (embedding.vector.length !== expectedDimension) {
-    throw new Error(
-      `Vector dimension mismatch: expected ${expectedDimension} dimensions but received ${embedding.vector.length} for ${id}`
-    );
-  }
-  return expectedDimension;
-}
-
-export {
-  createTool,
-  vectorFilter,
-  InMemoryVectorStore,
-  InMemoryVectorIndex,
-  createVectorSearchTool
-};
-//# sourceMappingURL=chunk-MELYDKWQ.js.map

package/dist/chunk-MELYDKWQ.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/vector-store/index.ts","../src/tool/create-tool.ts","../src/vector-store/filter.ts","../src/vector-store/lsh.ts"],"sourcesContent":["import { z } from \"zod\";\nimport {\n  cosineSimilarity,\n  type EmbeddedDocument,\n  type Embedding,\n  type EmbeddingModel,\n  embedText,\n  type VectorMetadata,\n} from \"../embeddings\";\nimport { createTool } from \"../tool/create-tool\";\nimport type { Tool } from \"../tool/tool\";\nimport { matchesVectorFilter } from \"./filter\";\nimport { LshIndex } from \"./lsh\";\nimport type {\n  IndexStrategy,\n  VectorInspectItem,\n  VectorInspectPage,\n  VectorInspectRequest,\n  VectorSearchIndex,\n  VectorSearchRequest,\n  VectorSearchResult,\n  VectorSearchToolOptions,\n} from \"./types\";\n\nexport { vectorFilter } from \"./filter\";\nexport type * from \"./types\";\n\ntype StoredDocument<T, Metadata extends VectorMetadata> = EmbeddedDocument<T, Metadata>;\n\nexport class InMemoryVectorStore<T, Metadata extends VectorMetadata = VectorMetadata> {\n  private readonly documents = new Map<string, StoredDocument<T, Metadata>>();\n  private indexStrategy: IndexStrategy;\n  private lshIndex: LshIndex | undefined;\n  private embeddingDimension: number | undefined;\n\n  constructor(options: { index?: IndexStrategy } = {}) {\n    this.indexStrategy = options.index ?? { type: \"bruteForce\" };\n  }\n\n  static fromDocuments<T, Metadata extends VectorMetadata = VectorMetadata>(\n    documents: Array<EmbeddedDocument<T, Metadata>>,\n    options: { index?: IndexStrategy } = {},\n  ): InMemoryVectorStore<T, Metadata> {\n    const store = new InMemoryVectorStore<T, Metadata>(options);\n    store.addDocuments(documents);\n    return store;\n  }\n\n  addDocuments(documents: Array<EmbeddedDocument<T, Metadata>>): this {\n    this.validateDocumentDimensions(documents);\n    for (const document of documents) {\n      this.documents.set(document.id, document);\n    }\n    this.rebuildLshIndex();\n    return this;\n  }\n\n  get(id: string): StoredDocument<T, Metadata> | undefined {\n    return this.documents.get(id);\n  }\n\n  values(): Array<StoredDocument<T, Metadata>> {\n    return [...this.documents.values()];\n  }\n\n  len(): number {\n    return this.documents.size;\n  }\n\n  isEmpty(): boolean {\n    return this.documents.size === 0;\n  }\n\n  index(model: EmbeddingModel): InMemoryVectorIndex<T, Metadata> {\n    return new InMemoryVectorIndex(model, this);\n  }\n\n  candidates(queryEmbedding: Embedding): Array<StoredDocument<T, Metadata>> {\n    this.validateQueryDimension(queryEmbedding);\n    if (this.indexStrategy.type !== \"lsh\" || this.lshIndex === undefined) {\n      return this.values();\n    }\n\n    const candidateIds = this.lshIndex.query(queryEmbedding.vector);\n    if (candidateIds.size === 0) {\n      return this.values();\n    }\n\n    return [...candidateIds].flatMap((id) => {\n      const document = this.documents.get(id);\n      return document === undefined ? [] : [document];\n    });\n  }\n\n  private rebuildLshIndex(): void {\n    if (this.indexStrategy.type !== \"lsh\") {\n      this.lshIndex = undefined;\n      return;\n    }\n\n    const firstEmbedding = this.values().flatMap((document) => document.embeddings)[0];\n    if (firstEmbedding === undefined) {\n      this.lshIndex = undefined;\n      return;\n    }\n\n    const index = new LshIndex(firstEmbedding.vector.length, this.indexStrategy);\n    for (const document of this.documents.values()) {\n      for (const embedding of document.embeddings) {\n        index.insert(document.id, embedding.vector);\n      }\n    }\n    this.lshIndex = index;\n  }\n\n  private validateDocumentDimensions(documents: Array<EmbeddedDocument<T, Metadata>>): void {\n    let dimension = this.embeddingDimension;\n    for (const document of documents) {\n      for (const embedding of document.embeddings) {\n        dimension = validateEmbeddingDimension(dimension, embedding, document.id);\n      }\n    }\n    this.embeddingDimension = dimension;\n  }\n\n  private validateQueryDimension(queryEmbedding: Embedding): void {\n    if (this.embeddingDimension === undefined) {\n      return;\n    }\n    validateEmbeddingDimension(this.embeddingDimension, queryEmbedding, \"query\");\n  }\n}\n\nexport class InMemoryVectorIndex<T, Metadata extends VectorMetadata = VectorMetadata>\n  implements VectorSearchIndex<T, Metadata>\n{\n  constructor(\n    private readonly model: EmbeddingModel,\n    private readonly store: InMemoryVectorStore<T, Metadata>,\n  ) {}\n\n  async search(request: VectorSearchRequest): Promise<Array<VectorSearchResult<T, Metadata>>> {\n    const queryEmbedding = await embedText(this.model, request.query);\n    return this.store\n      .candidates(queryEmbedding)\n      .filter((document) => matchesVectorFilter(document.metadata, request.filter))\n      .flatMap((document) => {\n        const score = bestScore(queryEmbedding, document.embeddings);\n        if (score === undefined) {\n          return [];\n        }\n        if (request.threshold !== undefined && score < request.threshold) {\n          return [];\n        }\n        const result: VectorSearchResult<T, Metadata> = {\n          score,\n          id: document.id,\n          document: document.document,\n        };\n        if (document.metadata !== undefined) {\n          result.metadata = document.metadata;\n        }\n        return [result];\n      })\n      .sort((left, right) => right.score - left.score)\n      .slice(0, Math.max(0, Math.trunc(request.topK)));\n  }\n\n  async searchIds(request: VectorSearchRequest): Promise<Array<{ score: number; id: string }>> {\n    return (await this.search(request)).map(({ score, id }) => ({ score, id }));\n  }\n\n  async inspect(request: VectorInspectRequest): Promise<VectorInspectPage<T, Metadata>> {\n    const limit = Math.max(0, Math.trunc(request.limit));\n    const start = Math.max(0, Math.trunc(Number(request.cursor ?? \"0\")));\n    const documents = this.store\n      .values()\n      .filter((document) => matchesVectorFilter(document.metadata, request.filter));\n    const page = documents.slice(start, start + limit);\n    const nextOffset = start + page.length;\n    const result: VectorInspectPage<T, Metadata> = {\n      items: page.map((document) => {\n        const item: VectorInspectItem<T, Metadata> = {\n          id: document.id,\n          document: document.document,\n        };\n        if (document.metadata !== undefined) {\n          item.metadata = document.metadata;\n        }\n        return item;\n      }),\n    };\n    if (nextOffset < documents.length) {\n      result.nextCursor = String(nextOffset);\n    }\n    result.totalCount = documents.length;\n    return result;\n  }\n\n  asTool(options: VectorSearchToolOptions): Tool<{ query: string; topK?: number }, unknown> {\n    return createVectorSearchTool(this, options);\n  }\n}\n\nexport function createVectorSearchTool<T, Metadata extends VectorMetadata>(\n  index: VectorSearchIndex<T, Metadata>,\n  options: VectorSearchToolOptions,\n): Tool<{ query: string; topK?: number }, Array<VectorSearchResult<T, Metadata>>> {\n  return createTool({\n    name: options.name,\n    description:\n      options.description ?? \"Search a vector store for documents relevant to the provided query.\",\n    input: z.object({\n      query: z.string().describe(\"The query string to search for relevant documents.\"),\n      topK: z.number().int().positive().optional().describe(\"The maximum number of results.\"),\n    }),\n    output: z.array(\n      z.object({\n        score: z.number(),\n        id: z.string(),\n        document: z.any(),\n        metadata: z\n          .record(z.string(), z.union([z.string(), z.number(), z.boolean(), z.null()]))\n          .optional(),\n      }),\n    ),\n    execute: ({ query, topK }) =>\n      index.search({\n        query,\n        topK: topK ?? options.topK ?? 5,\n        threshold: options.threshold,\n        filter: options.filter,\n      }),\n  }) as Tool<{ query: string; topK?: number }, Array<VectorSearchResult<T, Metadata>>>;\n}\n\nfunction bestScore(queryEmbedding: Embedding, embeddings: Embedding[]): number | undefined {\n  let best: number | undefined;\n  for (const embedding of embeddings) {\n    const score = cosineSimilarity(queryEmbedding.vector, embedding.vector);\n    best = best === undefined ? score : Math.max(best, score);\n  }\n  return best;\n}\n\nfunction validateEmbeddingDimension(\n  expectedDimension: number | undefined,\n  embedding: Embedding,\n  id: string,\n): number {\n  if (expectedDimension === undefined) {\n    return embedding.vector.length;\n  }\n  if (embedding.vector.length !== expectedDimension) {\n    throw new Error(\n      `Vector dimension mismatch: expected ${expectedDimension} dimensions but received ${embedding.vector.length} for ${id}`,\n    );\n  }\n  return expectedDimension;\n}\n","import type { z } from \"zod\";\nimport { toProviderJsonSchema, type ZodSchema } from \"../schema/zod-schema\";\nimport type { Tool, ToolApprovalPolicy, ToolCallContext } from \"./tool\";\n\nexport type CreateToolOptions<\n  InputSchema extends ZodSchema,\n  OutputSchema extends ZodSchema | undefined = undefined,\n  Output = unknown,\n> = {\n  name: string;\n  description: string;\n  input: InputSchema;\n  output?: OutputSchema;\n  approval?: ToolApprovalPolicy<z.output<InputSchema>>;\n  execute(\n    args: z.output<InputSchema>,\n    context: ToolCallContext,\n  ): OutputSchema extends ZodSchema\n    ? z.input<OutputSchema> | Promise<z.input<OutputSchema>>\n    : Output | Promise<Output>;\n};\n\ntype CreateToolOutput<\n  OutputSchema extends ZodSchema | undefined,\n  Output,\n> = OutputSchema extends ZodSchema ? z.output<OutputSchema> : Output;\n\nexport function createTool<InputSchema extends ZodSchema, Output = unknown>(\n  options: CreateToolOptions<InputSchema, undefined, Output> & { output?: undefined },\n): Tool<z.output<InputSchema>, Output>;\n\nexport function createTool<InputSchema extends ZodSchema, OutputSchema extends ZodSchema>(\n  options: CreateToolOptions<InputSchema, OutputSchema>,\n): Tool<z.output<InputSchema>, z.output<OutputSchema>>;\n\nexport function createTool<\n  InputSchema extends ZodSchema,\n  OutputSchema extends ZodSchema | undefined = undefined,\n  Output = unknown,\n>(\n  options: CreateToolOptions<InputSchema, OutputSchema, Output>,\n): Tool<z.output<InputSchema>, CreateToolOutput<OutputSchema, Output>> {\n  const parameters = toProviderJsonSchema(options.input);\n  const definition = () => ({\n    name: options.name,\n    description: options.description,\n    parameters,\n  });\n  const call = async (\n    args: z.output<InputSchema>,\n    context: ToolCallContext = {},\n  ): Promise<CreateToolOutput<OutputSchema, Output>> => {\n    const parsedArgs = options.input.parse(args);\n    const result = await options.execute(parsedArgs, context);\n    return (\n      options.output === undefined ? result : options.output.parse(result)\n    ) as CreateToolOutput<OutputSchema, Output>;\n  };\n  const parseApprovalArgs = (args: unknown): z.output<InputSchema> => options.input.parse(args);\n\n  if (options.approval === undefined) {\n    return { name: options.name, definition, call, parseApprovalArgs };\n  }\n  return {\n    name: options.name,\n    approval: options.approval,\n    definition,\n    call,\n    parseApprovalArgs,\n  };\n}\n","import type { VectorMetadata, VectorMetadataValue } from \"../embeddings\";\nimport type { VectorFilter } from \"./types\";\n\nexport type { VectorFilter } from \"./types\";\n\nexport const vectorFilter = {\n  eq(key: string, value: VectorMetadataValue): VectorFilter {\n    return { type: \"eq\", key, value };\n  },\n  gt(key: string, value: VectorMetadataValue): VectorFilter {\n    return { type: \"gt\", key, value };\n  },\n  lt(key: string, value: VectorMetadataValue): VectorFilter {\n    return { type: \"lt\", key, value };\n  },\n  and(left: VectorFilter, right: VectorFilter): VectorFilter {\n    return { type: \"and\", filters: [left, right] };\n  },\n  or(left: VectorFilter, right: VectorFilter): VectorFilter {\n    return { type: \"or\", filters: [left, right] };\n  },\n};\n\nexport function matchesVectorFilter(\n  metadata: VectorMetadata | undefined,\n  filter: VectorFilter | undefined,\n): boolean {\n  if (filter === undefined) {\n    return true;\n  }\n  if (metadata === undefined) {\n    return false;\n  }\n\n  switch (filter.type) {\n    case \"eq\":\n      return metadata[filter.key] === filter.value;\n    case \"gt\":\n      return compare(metadata[filter.key], filter.value) > 0;\n    case \"lt\":\n      return compare(metadata[filter.key], filter.value) < 0;\n    case \"and\":\n      return (\n        matchesVectorFilter(metadata, filter.filters[0]) &&\n        matchesVectorFilter(metadata, filter.filters[1])\n      );\n    case \"or\":\n      return (\n        matchesVectorFilter(metadata, filter.filters[0]) ||\n        matchesVectorFilter(metadata, filter.filters[1])\n      );\n  }\n}\n\nfunction compare(left: VectorMetadataValue | undefined, right: VectorMetadataValue): number {\n  if (typeof left === \"number\" && typeof right === \"number\") {\n    return left - right;\n  }\n  if (typeof left === \"string\" && typeof right === \"string\") {\n    return left.localeCompare(right);\n  }\n  if (typeof left === \"boolean\" && typeof right === \"boolean\") {\n    return Number(left) - Number(right);\n  }\n  return 0;\n}\n","import type { LshOptions } from \"./types\";\n\nexport type { LshOptions } from \"./types\";\n\nexport class LshIndex {\n  private readonly hyperplanes: number[][];\n  private readonly tables: Array<Map<string, Set<string>>>;\n\n  constructor(\n    dimensions: number,\n    private readonly options: LshOptions,\n  ) {\n    const rng = seededRandom(options.seed ?? 42);\n    this.hyperplanes = [];\n    for (let index = 0; index < options.numTables * options.numHyperplanes; index += 1) {\n      const plane = Array.from({ length: dimensions }, () => rng() * 2 - 1);\n      const norm = Math.sqrt(plane.reduce((sum, value) => sum + value ** 2, 0));\n      this.hyperplanes.push(norm === 0 ? plane : plane.map((value) => value / norm));\n    }\n    this.tables = Array.from({ length: options.numTables }, () => new Map());\n  }\n\n  insert(id: string, vector: number[]): void {\n    for (let table = 0; table < this.options.numTables; table += 1) {\n      const hash = this.hash(vector, table);\n      const bucket = this.tables[table]?.get(hash) ?? new Set<string>();\n      bucket.add(id);\n      this.tables[table]?.set(hash, bucket);\n    }\n  }\n\n  query(vector: number[]): Set<string> {\n    const candidates = new Set<string>();\n    for (let table = 0; table < this.options.numTables; table += 1) {\n      const hash = this.hash(vector, table);\n      for (const id of this.tables[table]?.get(hash) ?? []) {\n        candidates.add(id);\n      }\n    }\n    return candidates;\n  }\n\n  private hash(vector: number[], table: number): string {\n    let hash = \"\";\n    const start = table * this.options.numHyperplanes;\n    for (let offset = 0; offset < this.options.numHyperplanes; offset += 1) {\n      const plane = this.hyperplanes[start + offset] as number[];\n      const dot = vector.reduce((sum, value, index) => sum + value * (plane[index] ?? 0), 0);\n      hash += dot >= 0 ? \"1\" : \"0\";\n    }\n    return hash;\n  }\n}\n\nfunction seededRandom(seed: number): () => number {\n  let state = seed >>> 0;\n  return () => {\n    state = (state * 1664525 + 1013904223) >>> 0;\n    return state / 0x100000000;\n  };\n}\n"],"mappings":";;;;;;;;;AAAA,SAAS,SAAS;;;ACmCX,SAAS,WAKd,SACqE;AACrE,QAAM,aAAa,qBAAqB,QAAQ,KAAK;AACrD,QAAM,aAAa,OAAO;AAAA,IACxB,MAAM,QAAQ;AAAA,IACd,aAAa,QAAQ;AAAA,IACrB;AAAA,EACF;AACA,QAAM,OAAO,OACX,MACA,UAA2B,CAAC,MACwB;AACpD,UAAM,aAAa,QAAQ,MAAM,MAAM,IAAI;AAC3C,UAAM,SAAS,MAAM,QAAQ,QAAQ,YAAY,OAAO;AACxD,WACE,QAAQ,WAAW,SAAY,SAAS,QAAQ,OAAO,MAAM,MAAM;AAAA,EAEvE;AACA,QAAM,oBAAoB,CAAC,SAAyC,QAAQ,MAAM,MAAM,IAAI;AAE5F,MAAI,QAAQ,aAAa,QAAW;AAClC,WAAO,EAAE,MAAM,QAAQ,MAAM,YAAY,MAAM,kBAAkB;AAAA,EACnE;AACA,SAAO;AAAA,IACL,MAAM,QAAQ;AAAA,IACd,UAAU,QAAQ;AAAA,IAClB;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;;;ACjEO,IAAM,eAAe;AAAA,EAC1B,GAAG,KAAa,OAA0C;AACxD,WAAO,EAAE,MAAM,MAAM,KAAK,MAAM;AAAA,EAClC;AAAA,EACA,GAAG,KAAa,OAA0C;AACxD,WAAO,EAAE,MAAM,MAAM,KAAK,MAAM;AAAA,EAClC;AAAA,EACA,GAAG,KAAa,OAA0C;AACxD,WAAO,EAAE,MAAM,MAAM,KAAK,MAAM;AAAA,EAClC;AAAA,EACA,IAAI,MAAoB,OAAmC;AACzD,WAAO,EAAE,MAAM,OAAO,SAAS,CAAC,MAAM,KAAK,EAAE;AAAA,EAC/C;AAAA,EACA,GAAG,MAAoB,OAAmC;AACxD,WAAO,EAAE,MAAM,MAAM,SAAS,CAAC,MAAM,KAAK,EAAE;AAAA,EAC9C;AACF;AAEO,SAAS,oBACd,UACA,QACS;AACT,MAAI,WAAW,QAAW;AACxB,WAAO;AAAA,EACT;AACA,MAAI,aAAa,QAAW;AAC1B,WAAO;AAAA,EACT;AAEA,UAAQ,OAAO,MAAM;AAAA,IACnB,KAAK;AACH,aAAO,SAAS,OAAO,GAAG,MAAM,OAAO;AAAA,IACzC,KAAK;AACH,aAAO,QAAQ,SAAS,OAAO,GAAG,GAAG,OAAO,KAAK,IAAI;AAAA,IACvD,KAAK;AACH,aAAO,QAAQ,SAAS,OAAO,GAAG,GAAG,OAAO,KAAK,IAAI;AAAA,IACvD,KAAK;AACH,aACE,oBAAoB,UAAU,OAAO,QAAQ,CAAC,CAAC,KAC/C,oBAAoB,UAAU,OAAO,QAAQ,CAAC,CAAC;AAAA,IAEnD,KAAK;AACH,aACE,oBAAoB,UAAU,OAAO,QAAQ,CAAC,CAAC,KAC/C,oBAAoB,UAAU,OAAO,QAAQ,CAAC,CAAC;AAAA,EAErD;AACF;AAEA,SAAS,QAAQ,MAAuC,OAAoC;AAC1F,MAAI,OAAO,SAAS,YAAY,OAAO,UAAU,UAAU;AACzD,WAAO,OAAO;AAAA,EAChB;AACA,MAAI,OAAO,SAAS,YAAY,OAAO,UAAU,UAAU;AACzD,WAAO,KAAK,cAAc,KAAK;AAAA,EACjC;AACA,MAAI,OAAO,SAAS,aAAa,OAAO,UAAU,WAAW;AAC3D,WAAO,OAAO,IAAI,IAAI,OAAO,KAAK;AAAA,EACpC;AACA,SAAO;AACT;;;AC7DO,IAAM,WAAN,MAAe;AAAA,EAIpB,YACE,YACiB,SACjB;AADiB;AAEjB,UAAM,MAAM,aAAa,QAAQ,QAAQ,EAAE;AAC3C,SAAK,cAAc,CAAC;AACpB,aAAS,QAAQ,GAAG,QAAQ,QAAQ,YAAY,QAAQ,gBAAgB,SAAS,GAAG;AAClF,YAAM,QAAQ,MAAM,KAAK,EAAE,QAAQ,WAAW,GAAG,MAAM,IAAI,IAAI,IAAI,CAAC;AACpE,YAAM,OAAO,KAAK,KAAK,MAAM,OAAO,CAAC,KAAK,UAAU,MAAM,SAAS,GAAG,CAAC,CAAC;AACxE,WAAK,YAAY,KAAK,SAAS,IAAI,QAAQ,MAAM,IAAI,CAAC,UAAU,QAAQ,IAAI,CAAC;AAAA,IAC/E;AACA,SAAK,SAAS,MAAM,KAAK,EAAE,QAAQ,QAAQ,UAAU,GAAG,MAAM,oBAAI,IAAI,CAAC;AAAA,EACzE;AAAA,EAVmB;AAAA,EALF;AAAA,EACA;AAAA,EAgBjB,OAAO,IAAY,QAAwB;AACzC,aAAS,QAAQ,GAAG,QAAQ,KAAK,QAAQ,WAAW,SAAS,GAAG;AAC9D,YAAM,OAAO,KAAK,KAAK,QAAQ,KAAK;AACpC,YAAM,SAAS,KAAK,OAAO,KAAK,GAAG,IAAI,IAAI,KAAK,oBAAI,IAAY;AAChE,aAAO,IAAI,EAAE;AACb,WAAK,OAAO,KAAK,GAAG,IAAI,MAAM,MAAM;AAAA,IACtC;AAAA,EACF;AAAA,EAEA,MAAM,QAA+B;AACnC,UAAM,aAAa,oBAAI,IAAY;AACnC,aAAS,QAAQ,GAAG,QAAQ,KAAK,QAAQ,WAAW,SAAS,GAAG;AAC9D,YAAM,OAAO,KAAK,KAAK,QAAQ,KAAK;AACpC,iBAAW,MAAM,KAAK,OAAO,KAAK,GAAG,IAAI,IAAI,KAAK,CAAC,GAAG;AACpD,mBAAW,IAAI,EAAE;AAAA,MACnB;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,KAAK,QAAkB,OAAuB;AACpD,QAAI,OAAO;AACX,UAAM,QAAQ,QAAQ,KAAK,QAAQ;AACnC,aAAS,SAAS,GAAG,SAAS,KAAK,QAAQ,gBAAgB,UAAU,GAAG;AACtE,YAAM,QAAQ,KAAK,YAAY,QAAQ,MAAM;AAC7C,YAAM,MAAM,OAAO,OAAO,CAAC,KAAK,OAAO,UAAU,MAAM,SAAS,MAAM,KAAK,KAAK,IAAI,CAAC;AACrF,cAAQ,OAAO,IAAI,MAAM;AAAA,IAC3B;AACA,WAAO;AAAA,EACT;AACF;AAEA,SAAS,aAAa,MAA4B;AAChD,MAAI,QAAQ,SAAS;AACrB,SAAO,MAAM;AACX,YAAS,QAAQ,UAAU,eAAgB;AAC3C,WAAO,QAAQ;AAAA,EACjB;AACF;;;AH/BO,IAAM,sBAAN,MAAM,qBAAyE;AAAA,EACnE,YAAY,oBAAI,IAAyC;AAAA,EAClE;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,UAAqC,CAAC,GAAG;AACnD,SAAK,gBAAgB,QAAQ,SAAS,EAAE,MAAM,aAAa;AAAA,EAC7D;AAAA,EAEA,OAAO,cACL,WACA,UAAqC,CAAC,GACJ;AAClC,UAAM,QAAQ,IAAI,qBAAiC,OAAO;AAC1D,UAAM,aAAa,SAAS;AAC5B,WAAO;AAAA,EACT;AAAA,EAEA,aAAa,WAAuD;AAClE,SAAK,2BAA2B,SAAS;AACzC,eAAW,YAAY,WAAW;AAChC,WAAK,UAAU,IAAI,SAAS,IAAI,QAAQ;AAAA,IAC1C;AACA,SAAK,gBAAgB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,IAAI,IAAqD;AACvD,WAAO,KAAK,UAAU,IAAI,EAAE;AAAA,EAC9B;AAAA,EAEA,SAA6C;AAC3C,WAAO,CAAC,GAAG,KAAK,UAAU,OAAO,CAAC;AAAA,EACpC;AAAA,EAEA,MAAc;AACZ,WAAO,KAAK,UAAU;AAAA,EACxB;AAAA,EAEA,UAAmB;AACjB,WAAO,KAAK,UAAU,SAAS;AAAA,EACjC;AAAA,EAEA,MAAM,OAAyD;AAC7D,WAAO,IAAI,oBAAoB,OAAO,IAAI;AAAA,EAC5C;AAAA,EAEA,WAAW,gBAA+D;AACxE,SAAK,uBAAuB,cAAc;AAC1C,QAAI,KAAK,cAAc,SAAS,SAAS,KAAK,aAAa,QAAW;AACpE,aAAO,KAAK,OAAO;AAAA,IACrB;AAEA,UAAM,eAAe,KAAK,SAAS,MAAM,eAAe,MAAM;AAC9D,QAAI,aAAa,SAAS,GAAG;AAC3B,aAAO,KAAK,OAAO;AAAA,IACrB;AAEA,WAAO,CAAC,GAAG,YAAY,EAAE,QAAQ,CAAC,OAAO;AACvC,YAAM,WAAW,KAAK,UAAU,IAAI,EAAE;AACtC,aAAO,aAAa,SAAY,CAAC,IAAI,CAAC,QAAQ;AAAA,IAChD,CAAC;AAAA,EACH;AAAA,EAEQ,kBAAwB;AAC9B,QAAI,KAAK,cAAc,SAAS,OAAO;AACrC,WAAK,WAAW;AAChB;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK,OAAO,EAAE,QAAQ,CAAC,aAAa,SAAS,UAAU,EAAE,CAAC;AACjF,QAAI,mBAAmB,QAAW;AAChC,WAAK,WAAW;AAChB;AAAA,IACF;AAEA,UAAM,QAAQ,IAAI,SAAS,eAAe,OAAO,QAAQ,KAAK,aAAa;AAC3E,eAAW,YAAY,KAAK,UAAU,OAAO,GAAG;AAC9C,iBAAW,aAAa,SAAS,YAAY;AAC3C,cAAM,OAAO,SAAS,IAAI,UAAU,MAAM;AAAA,MAC5C;AAAA,IACF;AACA,SAAK,WAAW;AAAA,EAClB;AAAA,EAEQ,2BAA2B,WAAuD;AACxF,QAAI,YAAY,KAAK;AACrB,eAAW,YAAY,WAAW;AAChC,iBAAW,aAAa,SAAS,YAAY;AAC3C,oBAAY,2BAA2B,WAAW,WAAW,SAAS,EAAE;AAAA,MAC1E;AAAA,IACF;AACA,SAAK,qBAAqB;AAAA,EAC5B;AAAA,EAEQ,uBAAuB,gBAAiC;AAC9D,QAAI,KAAK,uBAAuB,QAAW;AACzC;AAAA,IACF;AACA,+BAA2B,KAAK,oBAAoB,gBAAgB,OAAO;AAAA,EAC7E;AACF;AAEO,IAAM,sBAAN,MAEP;AAAA,EACE,YACmB,OACA,OACjB;AAFiB;AACA;AAAA,EAChB;AAAA,EAFgB;AAAA,EACA;AAAA,EAGnB,MAAM,OAAO,SAA+E;AAC1F,UAAM,iBAAiB,MAAM,UAAU,KAAK,OAAO,QAAQ,KAAK;AAChE,WAAO,KAAK,MACT,WAAW,cAAc,EACzB,OAAO,CAAC,aAAa,oBAAoB,SAAS,UAAU,QAAQ,MAAM,CAAC,EAC3E,QAAQ,CAAC,aAAa;AACrB,YAAM,QAAQ,UAAU,gBAAgB,SAAS,UAAU;AAC3D,UAAI,UAAU,QAAW;AACvB,eAAO,CAAC;AAAA,MACV;AACA,UAAI,QAAQ,cAAc,UAAa,QAAQ,QAAQ,WAAW;AAChE,eAAO,CAAC;AAAA,MACV;AACA,YAAM,SAA0C;AAAA,QAC9C;AAAA,QACA,IAAI,SAAS;AAAA,QACb,UAAU,SAAS;AAAA,MACrB;AACA,UAAI,SAAS,aAAa,QAAW;AACnC,eAAO,WAAW,SAAS;AAAA,MAC7B;AACA,aAAO,CAAC,MAAM;AAAA,IAChB,CAAC,EACA,KAAK,CAAC,MAAM,UAAU,MAAM,QAAQ,KAAK,KAAK,EAC9C,MAAM,GAAG,KAAK,IAAI,GAAG,KAAK,MAAM,QAAQ,IAAI,CAAC,CAAC;AAAA,EACnD;AAAA,EAEA,MAAM,UAAU,SAA6E;AAC3F,YAAQ,MAAM,KAAK,OAAO,OAAO,GAAG,IAAI,CAAC,EAAE,OAAO,GAAG,OAAO,EAAE,OAAO,GAAG,EAAE;AAAA,EAC5E;AAAA,EAEA,MAAM,QAAQ,SAAwE;AACpF,UAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,QAAQ,KAAK,CAAC;AACnD,UAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,OAAO,QAAQ,UAAU,GAAG,CAAC,CAAC;AACnE,UAAM,YAAY,KAAK,MACpB,OAAO,EACP,OAAO,CAAC,aAAa,oBAAoB,SAAS,UAAU,QAAQ,MAAM,CAAC;AAC9E,UAAM,OAAO,UAAU,MAAM,OAAO,QAAQ,KAAK;AACjD,UAAM,aAAa,QAAQ,KAAK;AAChC,UAAM,SAAyC;AAAA,MAC7C,OAAO,KAAK,IAAI,CAAC,aAAa;AAC5B,cAAM,OAAuC;AAAA,UAC3C,IAAI,SAAS;AAAA,UACb,UAAU,SAAS;AAAA,QACrB;AACA,YAAI,SAAS,aAAa,QAAW;AACnC,eAAK,WAAW,SAAS;AAAA,QAC3B;AACA,eAAO;AAAA,MACT,CAAC;AAAA,IACH;AACA,QAAI,aAAa,UAAU,QAAQ;AACjC,aAAO,aAAa,OAAO,UAAU;AAAA,IACvC;AACA,WAAO,aAAa,UAAU;AAC9B,WAAO;AAAA,EACT;AAAA,EAEA,OAAO,SAAmF;AACxF,WAAO,uBAAuB,MAAM,OAAO;AAAA,EAC7C;AACF;AAEO,SAAS,uBACd,OACA,SACgF;AAChF,SAAO,WAAW;AAAA,IAChB,MAAM,QAAQ;AAAA,IACd,aACE,QAAQ,eAAe;AAAA,IACzB,OAAO,EAAE,OAAO;AAAA,MACd,OAAO,EAAE,OAAO,EAAE,SAAS,oDAAoD;AAAA,MAC/E,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,SAAS,gCAAgC;AAAA,IACxF,CAAC;AAAA,IACD,QAAQ,EAAE;AAAA,MACR,EAAE,OAAO;AAAA,QACP,OAAO,EAAE,OAAO;AAAA,QAChB,IAAI,EAAE,OAAO;AAAA,QACb,UAAU,EAAE,IAAI;AAAA,QAChB,UAAU,EACP,OAAO,EAAE,OAAO,GAAG,EAAE,MAAM,CAAC,EAAE,OAAO,GAAG,EAAE,OAAO,GAAG,EAAE,QAAQ,GAAG,EAAE,KAAK,CAAC,CAAC,CAAC,EAC3E,SAAS;AAAA,MACd,CAAC;AAAA,IACH;AAAA,IACA,SAAS,CAAC,EAAE,OAAO,KAAK,MACtB,MAAM,OAAO;AAAA,MACX;AAAA,MACA,MAAM,QAAQ,QAAQ,QAAQ;AAAA,MAC9B,WAAW,QAAQ;AAAA,MACnB,QAAQ,QAAQ;AAAA,IAClB,CAAC;AAAA,EACL,CAAC;AACH;AAEA,SAAS,UAAU,gBAA2B,YAA6C;AACzF,MAAI;AACJ,aAAW,aAAa,YAAY;AAClC,UAAM,QAAQ,iBAAiB,eAAe,QAAQ,UAAU,MAAM;AACtE,WAAO,SAAS,SAAY,QAAQ,KAAK,IAAI,MAAM,KAAK;AAAA,EAC1D;AACA,SAAO;AACT;AAEA,SAAS,2BACP,mBACA,WACA,IACQ;AACR,MAAI,sBAAsB,QAAW;AACnC,WAAO,UAAU,OAAO;AAAA,EAC1B;AACA,MAAI,UAAU,OAAO,WAAW,mBAAmB;AACjD,UAAM,IAAI;AAAA,MACR,uCAAuC,iBAAiB,4BAA4B,UAAU,OAAO,MAAM,QAAQ,EAAE;AAAA,IACvH;AAAA,EACF;AACA,SAAO;AACT;","names":[]}

package/dist/chunk-MNWK2USR.js

@@ -1,243 +0,0 @@
-import {
-  mapWithConcurrency
-} from "./chunk-OIMLU4SF.js";
-
-// src/embeddings/distance.ts
-function dotProduct(left, right) {
-  assertSameDimensions(left, right);
-  return left.reduce((sum, value, index) => sum + value * right[index], 0);
-}
-function cosineSimilarity(left, right) {
-  assertSameDimensions(left, right);
-  const leftMagnitude = magnitude(left);
-  const rightMagnitude = magnitude(right);
-  if (leftMagnitude === 0 || rightMagnitude === 0) {
-    return 0;
-  }
-  return dotProduct(left, right) / (leftMagnitude * rightMagnitude);
-}
-function angularDistance(left, right) {
-  const similarity = Math.max(-1, Math.min(1, cosineSimilarity(left, right)));
-  return Math.acos(similarity) / Math.PI;
-}
-function euclideanDistance(left, right) {
-  assertSameDimensions(left, right);
-  return Math.sqrt(
-    left.reduce((sum, value, index) => sum + (value - right[index]) ** 2, 0)
-  );
-}
-function manhattanDistance(left, right) {
-  assertSameDimensions(left, right);
-  return left.reduce((sum, value, index) => sum + Math.abs(value - right[index]), 0);
-}
-function chebyshevDistance(left, right) {
-  assertSameDimensions(left, right);
-  return left.reduce(
-    (max, value, index) => Math.max(max, Math.abs(value - right[index])),
-    0
-  );
-}
-function magnitude(vector) {
-  return Math.sqrt(vector.reduce((sum, value) => sum + value ** 2, 0));
-}
-function assertSameDimensions(left, right) {
-  if (left.length !== right.length) {
-    throw new Error(`Vector dimension mismatch: ${left.length} !== ${right.length}`);
-  }
-}
-
-// src/embeddings/embed.ts
-async function embedText(model, text) {
-  const embeddings = await embedTexts(model, [text]);
-  const embedding = embeddings[0];
-  if (embedding === void 0) {
-    throw new Error("Embedding model returned no embeddings");
-  }
-  return embedding;
-}
-async function embedTexts(model, texts) {
-  if (texts.length === 0) {
-    return [];
-  }
-  const maxBatchSize = Math.max(1, Math.trunc(model.maxBatchSize ?? texts.length));
-  const batches = [];
-  for (let index = 0; index < texts.length; index += maxBatchSize) {
-    batches.push(texts.slice(index, index + maxBatchSize));
-  }
-  const results = await mapWithConcurrency(batches, 1, (batch) => model.embedTexts(batch));
-  const embeddings = results.flat();
-  if (embeddings.length !== texts.length) {
-    throw new Error(
-      `Embedding model returned ${embeddings.length} embeddings for ${texts.length} texts`
-    );
-  }
-  return embeddings;
-}
-async function embedSparseTexts(model, texts) {
-  if (texts.length === 0) {
-    return [];
-  }
-  const maxBatchSize = Math.max(1, Math.trunc(model.maxBatchSize ?? texts.length));
-  const batches = [];
-  for (let index = 0; index < texts.length; index += maxBatchSize) {
-    batches.push(texts.slice(index, index + maxBatchSize));
-  }
-  const results = await mapWithConcurrency(batches, 1, (batch) => model.embedTexts(batch));
-  const embeddings = results.flat();
-  if (embeddings.length !== texts.length) {
-    throw new Error(
-      `Sparse embedding model returned ${embeddings.length} embeddings for ${texts.length} texts`
-    );
-  }
-  return embeddings;
-}
-async function embedSparseQuery(model, query) {
-  return model.embedQuery(query);
-}
-async function embedDocuments(model, documents, options) {
-  const prepared = prepareDocuments(documents, options);
-  const flatTexts = prepared.flatMap(
-    (item, documentIndex) => item.texts.map((text) => ({ documentIndex, text }))
-  );
-  const embeddings = await mapWithConcurrency(
-    chunk(flatTexts, Math.max(1, Math.trunc(model.maxBatchSize ?? (flatTexts.length || 1)))),
-    Math.max(1, Math.trunc(options.concurrency ?? 1)),
-    async (batch) => {
-      const batchEmbeddings = await model.embedTexts(batch.map((item) => item.text));
-      if (batchEmbeddings.length !== batch.length) {
-        throw new Error(
-          `Embedding model returned ${batchEmbeddings.length} embeddings for ${batch.length} texts`
-        );
-      }
-      return batch.map((item, index) => ({
-        documentIndex: item.documentIndex,
-        embedding: batchEmbeddings[index]
-      }));
-    }
-  );
-  const byDocument = /* @__PURE__ */ new Map();
-  for (const item of embeddings.flat()) {
-    const list = byDocument.get(item.documentIndex) ?? [];
-    list.push(item.embedding);
-    byDocument.set(item.documentIndex, list);
-  }
-  return prepared.map((item, index) => {
-    const documentEmbeddings = byDocument.get(index) ?? [];
-    if (item.metadata === void 0) {
-      return { id: item.id, document: item.document, embeddings: documentEmbeddings };
-    }
-    return {
-      id: item.id,
-      document: item.document,
-      metadata: item.metadata,
-      embeddings: documentEmbeddings
-    };
-  });
-}
-async function embedHybridDocuments(models, documents, options) {
-  const prepared = prepareDocuments(documents, options);
-  const flatTexts = prepared.flatMap(
-    (item, documentIndex) => item.texts.map((text) => ({ documentIndex, text }))
-  );
-  const concurrency = Math.max(1, Math.trunc(options.concurrency ?? 1));
-  const denseBatches = chunk(
-    flatTexts,
-    Math.max(1, Math.trunc(models.dense.maxBatchSize ?? (flatTexts.length || 1)))
-  );
-  const sparseBatches = chunk(
-    flatTexts,
-    Math.max(1, Math.trunc(models.sparse.maxBatchSize ?? (flatTexts.length || 1)))
-  );
-  const [denseResults, sparseResults] = await Promise.all([
-    mapWithConcurrency(denseBatches, concurrency, async (batch) => {
-      const batchEmbeddings = await models.dense.embedTexts(batch.map((item) => item.text));
-      if (batchEmbeddings.length !== batch.length) {
-        throw new Error(
-          `Embedding model returned ${batchEmbeddings.length} embeddings for ${batch.length} texts`
-        );
-      }
-      return batch.map((item, index) => ({
-        documentIndex: item.documentIndex,
-        embedding: batchEmbeddings[index]
-      }));
-    }),
-    mapWithConcurrency(sparseBatches, concurrency, async (batch) => {
-      const batchEmbeddings = await models.sparse.embedTexts(batch.map((item) => item.text));
-      if (batchEmbeddings.length !== batch.length) {
-        throw new Error(
-          `Sparse embedding model returned ${batchEmbeddings.length} embeddings for ${batch.length} texts`
-        );
-      }
-      return batch.map((item, index) => ({
-        documentIndex: item.documentIndex,
-        embedding: batchEmbeddings[index]
-      }));
-    })
-  ]);
-  const denseByDocument = /* @__PURE__ */ new Map();
-  for (const item of denseResults.flat()) {
-    const list = denseByDocument.get(item.documentIndex) ?? [];
-    list.push(item.embedding);
-    denseByDocument.set(item.documentIndex, list);
-  }
-  const sparseByDocument = /* @__PURE__ */ new Map();
-  for (const item of sparseResults.flat()) {
-    const list = sparseByDocument.get(item.documentIndex) ?? [];
-    list.push(item.embedding);
-    sparseByDocument.set(item.documentIndex, list);
-  }
-  return prepared.map((item, index) => {
-    const embeddings = denseByDocument.get(index) ?? [];
-    const sparseEmbeddings = sparseByDocument.get(index) ?? [];
-    if (embeddings.length !== sparseEmbeddings.length) {
-      throw new Error(
-        `Hybrid embedding produced ${embeddings.length} dense and ${sparseEmbeddings.length} sparse vectors for document ${item.id}`
-      );
-    }
-    if (item.metadata === void 0) {
-      return { id: item.id, document: item.document, embeddings, sparseEmbeddings };
-    }
-    return {
-      id: item.id,
-      document: item.document,
-      metadata: item.metadata,
-      embeddings,
-      sparseEmbeddings
-    };
-  });
-}
-function prepareDocuments(documents, options) {
-  return documents.map((document, index) => {
-    const content = options.content(document, index);
-    const texts = Array.isArray(content) ? content : [content];
-    return {
-      id: options.id?.(document, index) ?? `doc${index}`,
-      document,
-      metadata: options.metadata?.(document, index),
-      texts
-    };
-  });
-}
-function chunk(items, size) {
-  const chunks = [];
-  for (let index = 0; index < items.length; index += size) {
-    chunks.push(items.slice(index, index + size));
-  }
-  return chunks;
-}
-
-export {
-  dotProduct,
-  cosineSimilarity,
-  angularDistance,
-  euclideanDistance,
-  manhattanDistance,
-  chebyshevDistance,
-  embedText,
-  embedTexts,
-  embedSparseTexts,
-  embedSparseQuery,
-  embedDocuments,
-  embedHybridDocuments
-};
-//# sourceMappingURL=chunk-MNWK2USR.js.map

package/dist/chunk-MNWK2USR.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/embeddings/distance.ts","../src/embeddings/embed.ts"],"sourcesContent":["export function dotProduct(left: number[], right: number[]): number {\n  assertSameDimensions(left, right);\n  return left.reduce((sum, value, index) => sum + value * (right[index] as number), 0);\n}\n\nexport function cosineSimilarity(left: number[], right: number[]): number {\n  assertSameDimensions(left, right);\n  const leftMagnitude = magnitude(left);\n  const rightMagnitude = magnitude(right);\n  if (leftMagnitude === 0 || rightMagnitude === 0) {\n    return 0;\n  }\n  return dotProduct(left, right) / (leftMagnitude * rightMagnitude);\n}\n\nexport function angularDistance(left: number[], right: number[]): number {\n  const similarity = Math.max(-1, Math.min(1, cosineSimilarity(left, right)));\n  return Math.acos(similarity) / Math.PI;\n}\n\nexport function euclideanDistance(left: number[], right: number[]): number {\n  assertSameDimensions(left, right);\n  return Math.sqrt(\n    left.reduce((sum, value, index) => sum + (value - (right[index] as number)) ** 2, 0),\n  );\n}\n\nexport function manhattanDistance(left: number[], right: number[]): number {\n  assertSameDimensions(left, right);\n  return left.reduce((sum, value, index) => sum + Math.abs(value - (right[index] as number)), 0);\n}\n\nexport function chebyshevDistance(left: number[], right: number[]): number {\n  assertSameDimensions(left, right);\n  return left.reduce(\n    (max, value, index) => Math.max(max, Math.abs(value - (right[index] as number))),\n    0,\n  );\n}\n\nfunction magnitude(vector: number[]): number {\n  return Math.sqrt(vector.reduce((sum, value) => sum + value ** 2, 0));\n}\n\nfunction assertSameDimensions(left: number[], right: number[]): void {\n  if (left.length !== right.length) {\n    throw new Error(`Vector dimension mismatch: ${left.length} !== ${right.length}`);\n  }\n}\n","import { mapWithConcurrency } from \"../internal/concurrency\";\nimport type {\n  EmbedDocumentsOptions,\n  EmbeddedDocument,\n  Embedding,\n  EmbeddingModel,\n  EmbedHybridDocumentsOptions,\n  SparseEmbedding,\n  SparseEmbeddingModel,\n  VectorMetadata,\n} from \"./types\";\n\nexport async function embedText(model: EmbeddingModel, text: string): Promise<Embedding> {\n  const embeddings = await embedTexts(model, [text]);\n  const embedding = embeddings[0];\n  if (embedding === undefined) {\n    throw new Error(\"Embedding model returned no embeddings\");\n  }\n  return embedding;\n}\n\nexport async function embedTexts(model: EmbeddingModel, texts: string[]): Promise<Embedding[]> {\n  if (texts.length === 0) {\n    return [];\n  }\n\n  const maxBatchSize = Math.max(1, Math.trunc(model.maxBatchSize ?? texts.length));\n  const batches: string[][] = [];\n  for (let index = 0; index < texts.length; index += maxBatchSize) {\n    batches.push(texts.slice(index, index + maxBatchSize));\n  }\n\n  const results = await mapWithConcurrency(batches, 1, (batch) => model.embedTexts(batch));\n  const embeddings = results.flat();\n  if (embeddings.length !== texts.length) {\n    throw new Error(\n      `Embedding model returned ${embeddings.length} embeddings for ${texts.length} texts`,\n    );\n  }\n  return embeddings;\n}\n\nexport async function embedSparseTexts(\n  model: SparseEmbeddingModel,\n  texts: string[],\n): Promise<SparseEmbedding[]> {\n  if (texts.length === 0) {\n    return [];\n  }\n\n  const maxBatchSize = Math.max(1, Math.trunc(model.maxBatchSize ?? texts.length));\n  const batches: string[][] = [];\n  for (let index = 0; index < texts.length; index += maxBatchSize) {\n    batches.push(texts.slice(index, index + maxBatchSize));\n  }\n\n  const results = await mapWithConcurrency(batches, 1, (batch) => model.embedTexts(batch));\n  const embeddings = results.flat();\n  if (embeddings.length !== texts.length) {\n    throw new Error(\n      `Sparse embedding model returned ${embeddings.length} embeddings for ${texts.length} texts`,\n    );\n  }\n  return embeddings;\n}\n\nexport async function embedSparseQuery(\n  model: SparseEmbeddingModel,\n  query: string,\n): Promise<SparseEmbedding> {\n  return model.embedQuery(query);\n}\n\nexport async function embedDocuments<T, Metadata extends VectorMetadata = VectorMetadata>(\n  model: EmbeddingModel,\n  documents: T[],\n  options: EmbedDocumentsOptions<T, Metadata>,\n): Promise<Array<EmbeddedDocument<T, Metadata>>> {\n  const prepared = prepareDocuments(documents, options);\n\n  const flatTexts = prepared.flatMap((item, documentIndex) =>\n    item.texts.map((text) => ({ documentIndex, text })),\n  );\n  const embeddings = await mapWithConcurrency(\n    chunk(flatTexts, Math.max(1, Math.trunc(model.maxBatchSize ?? (flatTexts.length || 1)))),\n    Math.max(1, Math.trunc(options.concurrency ?? 1)),\n    async (batch) => {\n      const batchEmbeddings = await model.embedTexts(batch.map((item) => item.text));\n      if (batchEmbeddings.length !== batch.length) {\n        throw new Error(\n          `Embedding model returned ${batchEmbeddings.length} embeddings for ${batch.length} texts`,\n        );\n      }\n      return batch.map((item, index) => ({\n        documentIndex: item.documentIndex,\n        embedding: batchEmbeddings[index] as Embedding,\n      }));\n    },\n  );\n\n  const byDocument = new Map<number, Embedding[]>();\n  for (const item of embeddings.flat()) {\n    const list = byDocument.get(item.documentIndex) ?? [];\n    list.push(item.embedding);\n    byDocument.set(item.documentIndex, list);\n  }\n\n  return prepared.map((item, index) => {\n    const documentEmbeddings = byDocument.get(index) ?? [];\n    if (item.metadata === undefined) {\n      return { id: item.id, document: item.document, embeddings: documentEmbeddings };\n    }\n    return {\n      id: item.id,\n      document: item.document,\n      metadata: item.metadata,\n      embeddings: documentEmbeddings,\n    };\n  });\n}\n\nexport async function embedHybridDocuments<T, Metadata extends VectorMetadata = VectorMetadata>(\n  models: EmbedHybridDocumentsOptions,\n  documents: T[],\n  options: EmbedDocumentsOptions<T, Metadata>,\n): Promise<Array<EmbeddedDocument<T, Metadata>>> {\n  const prepared = prepareDocuments(documents, options);\n  const flatTexts = prepared.flatMap((item, documentIndex) =>\n    item.texts.map((text) => ({ documentIndex, text })),\n  );\n  const concurrency = Math.max(1, Math.trunc(options.concurrency ?? 1));\n\n  const denseBatches = chunk(\n    flatTexts,\n    Math.max(1, Math.trunc(models.dense.maxBatchSize ?? (flatTexts.length || 1))),\n  );\n  const sparseBatches = chunk(\n    flatTexts,\n    Math.max(1, Math.trunc(models.sparse.maxBatchSize ?? (flatTexts.length || 1))),\n  );\n\n  const [denseResults, sparseResults] = await Promise.all([\n    mapWithConcurrency(denseBatches, concurrency, async (batch) => {\n      const batchEmbeddings = await models.dense.embedTexts(batch.map((item) => item.text));\n      if (batchEmbeddings.length !== batch.length) {\n        throw new Error(\n          `Embedding model returned ${batchEmbeddings.length} embeddings for ${batch.length} texts`,\n        );\n      }\n      return batch.map((item, index) => ({\n        documentIndex: item.documentIndex,\n        embedding: batchEmbeddings[index] as Embedding,\n      }));\n    }),\n    mapWithConcurrency(sparseBatches, concurrency, async (batch) => {\n      const batchEmbeddings = await models.sparse.embedTexts(batch.map((item) => item.text));\n      if (batchEmbeddings.length !== batch.length) {\n        throw new Error(\n          `Sparse embedding model returned ${batchEmbeddings.length} embeddings for ${batch.length} texts`,\n        );\n      }\n      return batch.map((item, index) => ({\n        documentIndex: item.documentIndex,\n        embedding: batchEmbeddings[index] as SparseEmbedding,\n      }));\n    }),\n  ]);\n\n  const denseByDocument = new Map<number, Embedding[]>();\n  for (const item of denseResults.flat()) {\n    const list = denseByDocument.get(item.documentIndex) ?? [];\n    list.push(item.embedding);\n    denseByDocument.set(item.documentIndex, list);\n  }\n\n  const sparseByDocument = new Map<number, SparseEmbedding[]>();\n  for (const item of sparseResults.flat()) {\n    const list = sparseByDocument.get(item.documentIndex) ?? [];\n    list.push(item.embedding);\n    sparseByDocument.set(item.documentIndex, list);\n  }\n\n  return prepared.map((item, index) => {\n    const embeddings = denseByDocument.get(index) ?? [];\n    const sparseEmbeddings = sparseByDocument.get(index) ?? [];\n    if (embeddings.length !== sparseEmbeddings.length) {\n      throw new Error(\n        `Hybrid embedding produced ${embeddings.length} dense and ${sparseEmbeddings.length} sparse vectors for document ${item.id}`,\n      );\n    }\n    if (item.metadata === undefined) {\n      return { id: item.id, document: item.document, embeddings, sparseEmbeddings };\n    }\n    return {\n      id: item.id,\n      document: item.document,\n      metadata: item.metadata,\n      embeddings,\n      sparseEmbeddings,\n    };\n  });\n}\n\nfunction prepareDocuments<T, Metadata extends VectorMetadata>(\n  documents: T[],\n  options: EmbedDocumentsOptions<T, Metadata>,\n) {\n  return documents.map((document, index) => {\n    const content = options.content(document, index);\n    const texts = Array.isArray(content) ? content : [content];\n    return {\n      id: options.id?.(document, index) ?? `doc${index}`,\n      document,\n      metadata: options.metadata?.(document, index),\n      texts,\n    };\n  });\n}\n\nfunction chunk<T>(items: T[], size: number): T[][] {\n  const chunks: T[][] = [];\n  for (let index = 0; index < items.length; index += size) {\n    chunks.push(items.slice(index, index + size));\n  }\n  return chunks;\n}\n"],"mappings":";;;;;AAAO,SAAS,WAAW,MAAgB,OAAyB;AAClE,uBAAqB,MAAM,KAAK;AAChC,SAAO,KAAK,OAAO,CAAC,KAAK,OAAO,UAAU,MAAM,QAAS,MAAM,KAAK,GAAc,CAAC;AACrF;AAEO,SAAS,iBAAiB,MAAgB,OAAyB;AACxE,uBAAqB,MAAM,KAAK;AAChC,QAAM,gBAAgB,UAAU,IAAI;AACpC,QAAM,iBAAiB,UAAU,KAAK;AACtC,MAAI,kBAAkB,KAAK,mBAAmB,GAAG;AAC/C,WAAO;AAAA,EACT;AACA,SAAO,WAAW,MAAM,KAAK,KAAK,gBAAgB;AACpD;AAEO,SAAS,gBAAgB,MAAgB,OAAyB;AACvE,QAAM,aAAa,KAAK,IAAI,IAAI,KAAK,IAAI,GAAG,iBAAiB,MAAM,KAAK,CAAC,CAAC;AAC1E,SAAO,KAAK,KAAK,UAAU,IAAI,KAAK;AACtC;AAEO,SAAS,kBAAkB,MAAgB,OAAyB;AACzE,uBAAqB,MAAM,KAAK;AAChC,SAAO,KAAK;AAAA,IACV,KAAK,OAAO,CAAC,KAAK,OAAO,UAAU,OAAO,QAAS,MAAM,KAAK,MAAiB,GAAG,CAAC;AAAA,EACrF;AACF;AAEO,SAAS,kBAAkB,MAAgB,OAAyB;AACzE,uBAAqB,MAAM,KAAK;AAChC,SAAO,KAAK,OAAO,CAAC,KAAK,OAAO,UAAU,MAAM,KAAK,IAAI,QAAS,MAAM,KAAK,CAAY,GAAG,CAAC;AAC/F;AAEO,SAAS,kBAAkB,MAAgB,OAAyB;AACzE,uBAAqB,MAAM,KAAK;AAChC,SAAO,KAAK;AAAA,IACV,CAAC,KAAK,OAAO,UAAU,KAAK,IAAI,KAAK,KAAK,IAAI,QAAS,MAAM,KAAK,CAAY,CAAC;AAAA,IAC/E;AAAA,EACF;AACF;AAEA,SAAS,UAAU,QAA0B;AAC3C,SAAO,KAAK,KAAK,OAAO,OAAO,CAAC,KAAK,UAAU,MAAM,SAAS,GAAG,CAAC,CAAC;AACrE;AAEA,SAAS,qBAAqB,MAAgB,OAAuB;AACnE,MAAI,KAAK,WAAW,MAAM,QAAQ;AAChC,UAAM,IAAI,MAAM,8BAA8B,KAAK,MAAM,QAAQ,MAAM,MAAM,EAAE;AAAA,EACjF;AACF;;;ACpCA,eAAsB,UAAU,OAAuB,MAAkC;AACvF,QAAM,aAAa,MAAM,WAAW,OAAO,CAAC,IAAI,CAAC;AACjD,QAAM,YAAY,WAAW,CAAC;AAC9B,MAAI,cAAc,QAAW;AAC3B,UAAM,IAAI,MAAM,wCAAwC;AAAA,EAC1D;AACA,SAAO;AACT;AAEA,eAAsB,WAAW,OAAuB,OAAuC;AAC7F,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO,CAAC;AAAA,EACV;AAEA,QAAM,eAAe,KAAK,IAAI,GAAG,KAAK,MAAM,MAAM,gBAAgB,MAAM,MAAM,CAAC;AAC/E,QAAM,UAAsB,CAAC;AAC7B,WAAS,QAAQ,GAAG,QAAQ,MAAM,QAAQ,SAAS,cAAc;AAC/D,YAAQ,KAAK,MAAM,MAAM,OAAO,QAAQ,YAAY,CAAC;AAAA,EACvD;AAEA,QAAM,UAAU,MAAM,mBAAmB,SAAS,GAAG,CAAC,UAAU,MAAM,WAAW,KAAK,CAAC;AACvF,QAAM,aAAa,QAAQ,KAAK;AAChC,MAAI,WAAW,WAAW,MAAM,QAAQ;AACtC,UAAM,IAAI;AAAA,MACR,4BAA4B,WAAW,MAAM,mBAAmB,MAAM,MAAM;AAAA,IAC9E;AAAA,EACF;AACA,SAAO;AACT;AAEA,eAAsB,iBACpB,OACA,OAC4B;AAC5B,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO,CAAC;AAAA,EACV;AAEA,QAAM,eAAe,KAAK,IAAI,GAAG,KAAK,MAAM,MAAM,gBAAgB,MAAM,MAAM,CAAC;AAC/E,QAAM,UAAsB,CAAC;AAC7B,WAAS,QAAQ,GAAG,QAAQ,MAAM,QAAQ,SAAS,cAAc;AAC/D,YAAQ,KAAK,MAAM,MAAM,OAAO,QAAQ,YAAY,CAAC;AAAA,EACvD;AAEA,QAAM,UAAU,MAAM,mBAAmB,SAAS,GAAG,CAAC,UAAU,MAAM,WAAW,KAAK,CAAC;AACvF,QAAM,aAAa,QAAQ,KAAK;AAChC,MAAI,WAAW,WAAW,MAAM,QAAQ;AACtC,UAAM,IAAI;AAAA,MACR,mCAAmC,WAAW,MAAM,mBAAmB,MAAM,MAAM;AAAA,IACrF;AAAA,EACF;AACA,SAAO;AACT;AAEA,eAAsB,iBACpB,OACA,OAC0B;AAC1B,SAAO,MAAM,WAAW,KAAK;AAC/B;AAEA,eAAsB,eACpB,OACA,WACA,SAC+C;AAC/C,QAAM,WAAW,iBAAiB,WAAW,OAAO;AAEpD,QAAM,YAAY,SAAS;AAAA,IAAQ,CAAC,MAAM,kBACxC,KAAK,MAAM,IAAI,CAAC,UAAU,EAAE,eAAe,KAAK,EAAE;AAAA,EACpD;AACA,QAAM,aAAa,MAAM;AAAA,IACvB,MAAM,WAAW,KAAK,IAAI,GAAG,KAAK,MAAM,MAAM,iBAAiB,UAAU,UAAU,EAAE,CAAC,CAAC;AAAA,IACvF,KAAK,IAAI,GAAG,KAAK,MAAM,QAAQ,eAAe,CAAC,CAAC;AAAA,IAChD,OAAO,UAAU;AACf,YAAM,kBAAkB,MAAM,MAAM,WAAW,MAAM,IAAI,CAAC,SAAS,KAAK,IAAI,CAAC;AAC7E,UAAI,gBAAgB,WAAW,MAAM,QAAQ;AAC3C,cAAM,IAAI;AAAA,UACR,4BAA4B,gBAAgB,MAAM,mBAAmB,MAAM,MAAM;AAAA,QACnF;AAAA,MACF;AACA,aAAO,MAAM,IAAI,CAAC,MAAM,WAAW;AAAA,QACjC,eAAe,KAAK;AAAA,QACpB,WAAW,gBAAgB,KAAK;AAAA,MAClC,EAAE;AAAA,IACJ;AAAA,EACF;AAEA,QAAM,aAAa,oBAAI,IAAyB;AAChD,aAAW,QAAQ,WAAW,KAAK,GAAG;AACpC,UAAM,OAAO,WAAW,IAAI,KAAK,aAAa,KAAK,CAAC;AACpD,SAAK,KAAK,KAAK,SAAS;AACxB,eAAW,IAAI,KAAK,eAAe,IAAI;AAAA,EACzC;AAEA,SAAO,SAAS,IAAI,CAAC,MAAM,UAAU;AACnC,UAAM,qBAAqB,WAAW,IAAI,KAAK,KAAK,CAAC;AACrD,QAAI,KAAK,aAAa,QAAW;AAC/B,aAAO,EAAE,IAAI,KAAK,IAAI,UAAU,KAAK,UAAU,YAAY,mBAAmB;AAAA,IAChF;AACA,WAAO;AAAA,MACL,IAAI,KAAK;AAAA,MACT,UAAU,KAAK;AAAA,MACf,UAAU,KAAK;AAAA,MACf,YAAY;AAAA,IACd;AAAA,EACF,CAAC;AACH;AAEA,eAAsB,qBACpB,QACA,WACA,SAC+C;AAC/C,QAAM,WAAW,iBAAiB,WAAW,OAAO;AACpD,QAAM,YAAY,SAAS;AAAA,IAAQ,CAAC,MAAM,kBACxC,KAAK,MAAM,IAAI,CAAC,UAAU,EAAE,eAAe,KAAK,EAAE;AAAA,EACpD;AACA,QAAM,cAAc,KAAK,IAAI,GAAG,KAAK,MAAM,QAAQ,eAAe,CAAC,CAAC;AAEpE,QAAM,eAAe;AAAA,IACnB;AAAA,IACA,KAAK,IAAI,GAAG,KAAK,MAAM,OAAO,MAAM,iBAAiB,UAAU,UAAU,EAAE,CAAC;AAAA,EAC9E;AACA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA,KAAK,IAAI,GAAG,KAAK,MAAM,OAAO,OAAO,iBAAiB,UAAU,UAAU,EAAE,CAAC;AAAA,EAC/E;AAEA,QAAM,CAAC,cAAc,aAAa,IAAI,MAAM,QAAQ,IAAI;AAAA,IACtD,mBAAmB,cAAc,aAAa,OAAO,UAAU;AAC7D,YAAM,kBAAkB,MAAM,OAAO,MAAM,WAAW,MAAM,IAAI,CAAC,SAAS,KAAK,IAAI,CAAC;AACpF,UAAI,gBAAgB,WAAW,MAAM,QAAQ;AAC3C,cAAM,IAAI;AAAA,UACR,4BAA4B,gBAAgB,MAAM,mBAAmB,MAAM,MAAM;AAAA,QACnF;AAAA,MACF;AACA,aAAO,MAAM,IAAI,CAAC,MAAM,WAAW;AAAA,QACjC,eAAe,KAAK;AAAA,QACpB,WAAW,gBAAgB,KAAK;AAAA,MAClC,EAAE;AAAA,IACJ,CAAC;AAAA,IACD,mBAAmB,eAAe,aAAa,OAAO,UAAU;AAC9D,YAAM,kBAAkB,MAAM,OAAO,OAAO,WAAW,MAAM,IAAI,CAAC,SAAS,KAAK,IAAI,CAAC;AACrF,UAAI,gBAAgB,WAAW,MAAM,QAAQ;AAC3C,cAAM,IAAI;AAAA,UACR,mCAAmC,gBAAgB,MAAM,mBAAmB,MAAM,MAAM;AAAA,QAC1F;AAAA,MACF;AACA,aAAO,MAAM,IAAI,CAAC,MAAM,WAAW;AAAA,QACjC,eAAe,KAAK;AAAA,QACpB,WAAW,gBAAgB,KAAK;AAAA,MAClC,EAAE;AAAA,IACJ,CAAC;AAAA,EACH,CAAC;AAED,QAAM,kBAAkB,oBAAI,IAAyB;AACrD,aAAW,QAAQ,aAAa,KAAK,GAAG;AACtC,UAAM,OAAO,gBAAgB,IAAI,KAAK,aAAa,KAAK,CAAC;AACzD,SAAK,KAAK,KAAK,SAAS;AACxB,oBAAgB,IAAI,KAAK,eAAe,IAAI;AAAA,EAC9C;AAEA,QAAM,mBAAmB,oBAAI,IAA+B;AAC5D,aAAW,QAAQ,cAAc,KAAK,GAAG;AACvC,UAAM,OAAO,iBAAiB,IAAI,KAAK,aAAa,KAAK,CAAC;AAC1D,SAAK,KAAK,KAAK,SAAS;AACxB,qBAAiB,IAAI,KAAK,eAAe,IAAI;AAAA,EAC/C;AAEA,SAAO,SAAS,IAAI,CAAC,MAAM,UAAU;AACnC,UAAM,aAAa,gBAAgB,IAAI,KAAK,KAAK,CAAC;AAClD,UAAM,mBAAmB,iBAAiB,IAAI,KAAK,KAAK,CAAC;AACzD,QAAI,WAAW,WAAW,iBAAiB,QAAQ;AACjD,YAAM,IAAI;AAAA,QACR,6BAA6B,WAAW,MAAM,cAAc,iBAAiB,MAAM,gCAAgC,KAAK,EAAE;AAAA,MAC5H;AAAA,IACF;AACA,QAAI,KAAK,aAAa,QAAW;AAC/B,aAAO,EAAE,IAAI,KAAK,IAAI,UAAU,KAAK,UAAU,YAAY,iBAAiB;AAAA,IAC9E;AACA,WAAO;AAAA,MACL,IAAI,KAAK;AAAA,MACT,UAAU,KAAK;AAAA,MACf,UAAU,KAAK;AAAA,MACf;AAAA,MACA;AAAA,IACF;AAAA,EACF,CAAC;AACH;AAEA,SAAS,iBACP,WACA,SACA;AACA,SAAO,UAAU,IAAI,CAAC,UAAU,UAAU;AACxC,UAAM,UAAU,QAAQ,QAAQ,UAAU,KAAK;AAC/C,UAAM,QAAQ,MAAM,QAAQ,OAAO,IAAI,UAAU,CAAC,OAAO;AACzD,WAAO;AAAA,MACL,IAAI,QAAQ,KAAK,UAAU,KAAK,KAAK,MAAM,KAAK;AAAA,MAChD;AAAA,MACA,UAAU,QAAQ,WAAW,UAAU,KAAK;AAAA,MAC5C;AAAA,IACF;AAAA,EACF,CAAC;AACH;AAEA,SAAS,MAAS,OAAY,MAAqB;AACjD,QAAM,SAAgB,CAAC;AACvB,WAAS,QAAQ,GAAG,QAAQ,MAAM,QAAQ,SAAS,MAAM;AACvD,WAAO,KAAK,MAAM,MAAM,OAAO,QAAQ,IAAI,CAAC;AAAA,EAC9C;AACA,SAAO;AACT;","names":[]}