@xberg-io/langchain-xberg 1.0.14 → 1.1.0

package/README.md

@@ -11,7 +11,7 @@
 
 A [LangChain.js](https://js.langchain.com) document loader for [Xberg](https://github.com/xberg-io/xberg).
 Point it at a file, a directory, or raw bytes and it returns LangChain `Document`s with the extracted
-text, tables, and rich metadata from 101 document formats — with optional OCR for scans and images.
+text, tables, and rich metadata from 107 document formats — with optional OCR for scans and images.
 
 Extraction runs locally in-process through the `@xberg-io/xberg` native binding. No API key, no cloud
 call, no data leaves your machine.
@@ -69,13 +69,13 @@ const pages = await loader.load(); // metadata.page is 0-indexed
 
 ## Supported formats
 
-Xberg extracts from 101 formats including PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, and more. See the
+Xberg extracts from 107 formats including PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, and more. See the
 [Xberg documentation](https://docs.xberg.io) for the full list and the extraction configuration
 reference.
 
 ## Part of Xberg.io
 
-- [Xberg](https://github.com/xberg-io/xberg) — document intelligence: text, tables, metadata from 101 formats with optional OCR.
+- [Xberg](https://github.com/xberg-io/xberg) — document intelligence: text, tables, metadata from 107 formats with optional OCR.
 - [Xberg Enterprise](https://github.com/xberg-io/xberg-enterprise) — managed extraction API with SDKs, dashboards, and observability.
 
 ## License

package/dist/index.cjs

@@ -162,8 +162,7 @@ function pagesToDocuments(document, source) {
   const baseMetadata = buildMetadata(document, source);
   const documents = [];
   for (const page of document.pages ?? []) {
-    const metadata = { ...baseMetadata };
-    metadata.page = page.pageNumber - 1;
+    const metadata = { ...baseMetadata, page: page.pageNumber - 1 };
     if (page.isBlank != null) {
       metadata.is_blank = page.isBlank;
     }
@@ -233,7 +232,7 @@ var XbergLoader = class extends import_base.BaseDocumentLoader {
       result = batch ? await (0, import_xberg.extractBatch)(inputs, this.config ?? null) : await (0, import_xberg.extract)(inputs[0], this.config ?? null);
     } catch (error) {
       const source = sources[0] ?? "input";
-      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`);
+      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`, { cause: error });
     }
     return resultToDocuments(result, sources, {
       chunking: isChunkingEnabled(this.config),

package/dist/index.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/index.ts","../src/loader.ts","../src/mapping.ts"],"sourcesContent":["export { XbergLoader, type XbergLoaderOptions } from \"./loader\";\n","import { stat } from \"node:fs/promises\";\nimport { BaseDocumentLoader } from \"@langchain/core/document_loaders/base\";\nimport type { Document } from \"@langchain/core/documents\";\nimport { extract, extractBatch, ExtractInputKind } from \"@xberg-io/xberg\";\nimport type { ExtractInput, ExtractionConfig } from \"@xberg-io/xberg\";\nimport fastGlob from \"fast-glob\";\nimport { isChunkingEnabled, isPerPageEnabled, resultToDocuments, type ResultEnvelope } from \"./mapping\";\n\nconst DEFAULT_GLOB = \"**/*\";\n\nexport interface XbergLoaderOptions {\n  /** File path, list of file paths, or directory path to load. */\n  filePath?: string | string[];\n  /** Raw bytes to extract from. Mutually exclusive with `filePath`. */\n  data?: Uint8Array;\n  /** MIME type hint. Required when using `data`, optional for `filePath`. */\n  mimeType?: string;\n  /** Glob pattern for directory mode. Defaults to matching all files. */\n  glob?: string;\n  /** Xberg extraction configuration controlling output format, OCR, pages, chunking, etc. */\n  config?: ExtractionConfig;\n}\n\n/**\n * Load documents using Xberg, supporting 101 file formats with true async extraction.\n *\n * By default each source becomes one Document. Enable `chunking` on the\n * `ExtractionConfig` to emit one Document per chunk, or `pages` for one Document\n * per page. Multiple paths (a list or a directory glob) are extracted with a single\n * `extractBatch` call so concurrency happens Rust-side.\n */\nexport class XbergLoader extends BaseDocumentLoader {\n  private readonly filePath?: string | string[];\n  private readonly data?: Uint8Array;\n  private readonly mimeType?: string;\n  private readonly glob?: string;\n  private readonly config?: ExtractionConfig;\n\n  constructor(options: XbergLoaderOptions) {\n    super();\n    const { filePath, data, mimeType, glob, config } = options;\n    if (filePath === undefined && data === undefined) {\n      throw new Error(\"Either 'filePath' or 'data' must be provided.\");\n    }\n    if (filePath !== undefined && data !== undefined) {\n      throw new Error(\"Cannot specify both 'filePath' and 'data'. Use one or the other.\");\n    }\n    if (data !== undefined && mimeType === undefined) {\n      throw new Error(\"'mimeType' is required when using 'data'.\");\n    }\n\n    this.filePath = filePath;\n    this.data = data;\n    this.mimeType = mimeType;\n    this.glob = glob;\n    this.config = config;\n  }\n\n  async load(): Promise<Document[]> {\n    const { inputs, sources, batch } = await this.buildInputs();\n    if (inputs.length === 0) {\n      return [];\n    }\n\n    let result;\n    try {\n      result = batch ? await extractBatch(inputs, this.config ?? null) : await extract(inputs[0], this.config ?? null);\n    } catch (error) {\n      const source = sources[0] ?? \"input\";\n      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`);\n    }\n\n    return resultToDocuments(result as unknown as ResultEnvelope, sources, {\n      chunking: isChunkingEnabled(this.config),\n      perPage: isPerPageEnabled(this.config),\n    });\n  }\n\n  private async buildInputs(): Promise<{ inputs: ExtractInput[]; sources: string[]; batch: boolean }> {\n    if (this.data !== undefined) {\n      const source = `bytes://${this.mimeType}`;\n      const input: ExtractInput = { kind: ExtractInputKind.Bytes, bytes: this.data, mimeType: this.mimeType };\n      return { inputs: [input], sources: [source], batch: false };\n    }\n\n    const { paths, batch } = await this.resolvePaths();\n    const inputs: ExtractInput[] = paths.map((path) => ({\n      kind: ExtractInputKind.Uri,\n      uri: path,\n      mimeType: this.mimeType,\n    }));\n    return { inputs, sources: paths, batch };\n  }\n\n  private async resolvePaths(): Promise<{ paths: string[]; batch: boolean }> {\n    const filePath = this.filePath;\n    if (Array.isArray(filePath)) {\n      return { paths: filePath, batch: true };\n    }\n    if (typeof filePath === \"string\") {\n      if (await isDirectory(filePath)) {\n        const pattern = this.glob ?? DEFAULT_GLOB;\n        const matches = await fastGlob(pattern, { cwd: filePath, onlyFiles: true, absolute: true });\n        return { paths: matches.sort(), batch: true };\n      }\n      return { paths: [filePath], batch: false };\n    }\n    return { paths: [], batch: false };\n  }\n}\n\nasync function isDirectory(path: string): Promise<boolean> {\n  try {\n    const stats = await stat(path);\n    return stats.isDirectory();\n  } catch {\n    return false;\n  }\n}\n\nfunction errorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n","import { Document } from \"@langchain/core/documents\";\nimport type { DocumentCounts, ExtractionConfig, Keyword, Metadata, ProcessingWarning, Table } from \"@xberg-io/xberg\";\n\n// Metadata fields (from Xberg Metadata) that carry JSON-friendly scalar/list values,\n// mapped from the binding's camelCase keys to the snake_case output keys used by the\n// Python adapter. Opaque nested fields (pages, format, imagePreprocessing, jsonSchema,\n// error) are skipped because they are native objects, not plain data. ~keep\nconst METADATA_FIELDS: ReadonlyArray<readonly [string, keyof Metadata]> = [\n  [\"title\", \"title\"],\n  [\"subject\", \"subject\"],\n  [\"authors\", \"authors\"],\n  [\"keywords\", \"keywords\"],\n  [\"language\", \"language\"],\n  [\"created_at\", \"createdAt\"],\n  [\"modified_at\", \"modifiedAt\"],\n  [\"created_by\", \"createdBy\"],\n  [\"modified_by\", \"modifiedBy\"],\n  [\"category\", \"category\"],\n  [\"tags\", \"tags\"],\n  [\"document_version\", \"documentVersion\"],\n  [\"abstract_text\", \"abstractText\"],\n  [\"output_format\", \"outputFormat\"],\n  [\"ocr_used\", \"ocrUsed\"],\n  [\"extraction_duration_ms\", \"extractionDurationMs\"],\n];\n\nconst CONTENT_SEPARATOR = \"\\n\\n\";\n\nexport interface ChunkMetadataShape {\n  chunkIndex: number;\n  totalChunks: number;\n  headingPath?: string[];\n  tokenCount?: number;\n  firstPage?: number;\n  lastPage?: number;\n}\n\nexport interface ChunkShape {\n  content: string;\n  chunkType: unknown;\n  metadata: ChunkMetadataShape;\n}\n\nexport interface PageShape {\n  pageNumber: number;\n  content: string;\n  tables?: Table[];\n  isBlank?: boolean;\n}\n\n// Structural view of a binding ExtractedDocument. The binding types its own fields\n// through unexported `Js*` aliases; this interface uses the clean exported types plus\n// local chunk/page shapes so the mapper is fully typed and testable without the addon. ~keep\nexport interface ExtractedDoc {\n  content?: string;\n  mimeType?: string;\n  metadata?: Metadata;\n  tables?: Table[];\n  counts?: DocumentCounts;\n  detectedLanguages?: string[];\n  chunks?: ChunkShape[];\n  pages?: PageShape[];\n  extractedKeywords?: Keyword[];\n  qualityScore?: number;\n  processingWarnings?: ProcessingWarning[];\n}\n\nexport interface ResultEnvelope {\n  errors?: Array<{ source: string; message: string }>;\n  results?: ExtractedDoc[];\n}\n\nexport interface SplitOptions {\n  chunking: boolean;\n  perPage: boolean;\n}\n\nexport function isChunkingEnabled(config?: ExtractionConfig | null): boolean {\n  return config?.chunking != null;\n}\n\nexport function isPerPageEnabled(config?: ExtractionConfig | null): boolean {\n  return Boolean(config?.pages?.extractPages);\n}\n\nexport function flattenMetadata(metadata?: Metadata): Record<string, unknown> {\n  const flat: Record<string, unknown> = {};\n  if (!metadata) {\n    return flat;\n  }\n\n  for (const [snakeKey, camelKey] of METADATA_FIELDS) {\n    const value = metadata[camelKey];\n    if (value != null) {\n      flat[snakeKey] = value;\n    }\n  }\n\n  const additional = metadata.additional;\n  if (additional && Object.keys(additional).length > 0) {\n    flat.additional = { ...additional };\n  }\n\n  return flat;\n}\n\nexport function assembleContent(content: string | undefined, tables: Table[] | undefined): string {\n  const text = content ?? \"\";\n  if (!tables || tables.length === 0) {\n    return text;\n  }\n  const parts = tables.map((table) => table.markdown ?? \"\").filter((markdown) => markdown.length > 0);\n  if (parts.length === 0) {\n    return text;\n  }\n  return [text, ...parts].join(CONTENT_SEPARATOR);\n}\n\nexport function buildMetadata(document: ExtractedDoc, source: string): Record<string, unknown> {\n  const metadata = flattenMetadata(document.metadata);\n\n  metadata.mime_type = document.mimeType;\n  if (document.qualityScore != null) {\n    metadata.quality_score = document.qualityScore;\n  }\n  if (document.detectedLanguages && document.detectedLanguages.length > 0) {\n    metadata.detected_languages = document.detectedLanguages;\n  }\n  if (document.counts != null) {\n    metadata.page_count = document.counts.pages;\n  }\n\n  if (document.extractedKeywords && document.extractedKeywords.length > 0) {\n    metadata.extracted_keywords = document.extractedKeywords.map((keyword) => ({\n      text: keyword.text,\n      score: keyword.score,\n      algorithm: String(keyword.algorithm),\n    }));\n  }\n\n  const tables = document.tables ?? [];\n  metadata.table_count = tables.length;\n  if (tables.length > 0) {\n    metadata.tables = tables.map((table) => ({\n      cells: table.cells,\n      markdown: table.markdown,\n      page_number: table.pageNumber,\n    }));\n  }\n\n  if (document.processingWarnings && document.processingWarnings.length > 0) {\n    metadata.processing_warnings = document.processingWarnings.map((warning) => ({\n      source: warning.source,\n      message: warning.message,\n    }));\n  }\n\n  metadata.source = source;\n  return metadata;\n}\n\nfunction chunksToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const chunk of document.chunks ?? []) {\n    const metadata: Record<string, unknown> = { ...baseMetadata };\n    const chunkMetadata = chunk.metadata;\n    metadata.chunk_index = chunkMetadata.chunkIndex;\n    metadata.total_chunks = chunkMetadata.totalChunks;\n    metadata.chunk_type = String(chunk.chunkType);\n    if (chunkMetadata.headingPath && chunkMetadata.headingPath.length > 0) {\n      metadata.heading_path = [...chunkMetadata.headingPath];\n    }\n    if (chunkMetadata.tokenCount != null) {\n      metadata.token_count = chunkMetadata.tokenCount;\n    }\n    if (chunkMetadata.firstPage != null) {\n      // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n      metadata.page = chunkMetadata.firstPage - 1;\n      metadata.first_page = chunkMetadata.firstPage;\n    }\n    if (chunkMetadata.lastPage != null) {\n      metadata.last_page = chunkMetadata.lastPage;\n    }\n    documents.push(new Document({ pageContent: chunk.content, metadata }));\n  }\n\n  return documents;\n}\n\nfunction pagesToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const page of document.pages ?? []) {\n    const metadata: Record<string, unknown> = { ...baseMetadata };\n    // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n    metadata.page = page.pageNumber - 1;\n    if (page.isBlank != null) {\n      metadata.is_blank = page.isBlank;\n    }\n    const pageContent = assembleContent(page.content, page.tables);\n    documents.push(new Document({ pageContent, metadata }));\n  }\n\n  return documents;\n}\n\nexport function documentToDocuments(document: ExtractedDoc, source: string, options: SplitOptions): Document[] {\n  if (options.chunking && document.chunks && document.chunks.length > 0) {\n    return chunksToDocuments(document, source);\n  }\n  if (options.perPage && document.pages && document.pages.length > 0) {\n    return pagesToDocuments(document, source);\n  }\n  const metadata = buildMetadata(document, source);\n  const pageContent = assembleContent(document.content, document.tables);\n  return [new Document({ pageContent, metadata })];\n}\n\nexport function resultToDocuments(result: ResultEnvelope, sources: string[], options: SplitOptions): Document[] {\n  if (result.errors && result.errors.length > 0) {\n    const error = result.errors[0];\n    throw new Error(`Failed to extract '${error.source}': ${error.message}`);\n  }\n\n  const documents: Document[] = [];\n  const results = result.results ?? [];\n  results.forEach((document, index) => {\n    const source = sources[index] ?? (sources.length > 0 ? sources[sources.length - 1] : \"\");\n    documents.push(...documentToDocuments(document, source, options));\n  });\n  return documents;\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,sBAAqB;AACrB,kBAAmC;AAEnC,mBAAwD;AAExD,uBAAqB;;;ACLrB,uBAAyB;AAOzB,IAAM,kBAAoE;AAAA,EACxE,CAAC,SAAS,OAAO;AAAA,EACjB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,QAAQ,MAAM;AAAA,EACf,CAAC,oBAAoB,iBAAiB;AAAA,EACtC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,YAAY,SAAS;AAAA,EACtB,CAAC,0BAA0B,sBAAsB;AACnD;AAEA,IAAM,oBAAoB;AAmDnB,SAAS,kBAAkB,QAA2C;AAC3E,SAAO,QAAQ,YAAY;AAC7B;AAEO,SAAS,iBAAiB,QAA2C;AAC1E,SAAO,QAAQ,QAAQ,OAAO,YAAY;AAC5C;AAEO,SAAS,gBAAgB,UAA8C;AAC5E,QAAM,OAAgC,CAAC;AACvC,MAAI,CAAC,UAAU;AACb,WAAO;AAAA,EACT;AAEA,aAAW,CAAC,UAAU,QAAQ,KAAK,iBAAiB;AAClD,UAAM,QAAQ,SAAS,QAAQ;AAC/B,QAAI,SAAS,MAAM;AACjB,WAAK,QAAQ,IAAI;AAAA,IACnB;AAAA,EACF;AAEA,QAAM,aAAa,SAAS;AAC5B,MAAI,cAAc,OAAO,KAAK,UAAU,EAAE,SAAS,GAAG;AACpD,SAAK,aAAa,EAAE,GAAG,WAAW;AAAA,EACpC;AAEA,SAAO;AACT;AAEO,SAAS,gBAAgB,SAA6B,QAAqC;AAChG,QAAM,OAAO,WAAW;AACxB,MAAI,CAAC,UAAU,OAAO,WAAW,GAAG;AAClC,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,OAAO,IAAI,CAAC,UAAU,MAAM,YAAY,EAAE,EAAE,OAAO,CAAC,aAAa,SAAS,SAAS,CAAC;AAClG,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO;AAAA,EACT;AACA,SAAO,CAAC,MAAM,GAAG,KAAK,EAAE,KAAK,iBAAiB;AAChD;AAEO,SAAS,cAAc,UAAwB,QAAyC;AAC7F,QAAM,WAAW,gBAAgB,SAAS,QAAQ;AAElD,WAAS,YAAY,SAAS;AAC9B,MAAI,SAAS,gBAAgB,MAAM;AACjC,aAAS,gBAAgB,SAAS;AAAA,EACpC;AACA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS;AAAA,EACzC;AACA,MAAI,SAAS,UAAU,MAAM;AAC3B,aAAS,aAAa,SAAS,OAAO;AAAA,EACxC;AAEA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS,kBAAkB,IAAI,CAAC,aAAa;AAAA,MACzE,MAAM,QAAQ;AAAA,MACd,OAAO,QAAQ;AAAA,MACf,WAAW,OAAO,QAAQ,SAAS;AAAA,IACrC,EAAE;AAAA,EACJ;AAEA,QAAM,SAAS,SAAS,UAAU,CAAC;AACnC,WAAS,cAAc,OAAO;AAC9B,MAAI,OAAO,SAAS,GAAG;AACrB,aAAS,SAAS,OAAO,IAAI,CAAC,WAAW;AAAA,MACvC,OAAO,MAAM;AAAA,MACb,UAAU,MAAM;AAAA,MAChB,aAAa,MAAM;AAAA,IACrB,EAAE;AAAA,EACJ;AAEA,MAAI,SAAS,sBAAsB,SAAS,mBAAmB,SAAS,GAAG;AACzE,aAAS,sBAAsB,SAAS,mBAAmB,IAAI,CAAC,aAAa;AAAA,MAC3E,QAAQ,QAAQ;AAAA,MAChB,SAAS,QAAQ;AAAA,IACnB,EAAE;AAAA,EACJ;AAEA,WAAS,SAAS;AAClB,SAAO;AACT;AAEA,SAAS,kBAAkB,UAAwB,QAA4B;AAC7E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,SAAS,SAAS,UAAU,CAAC,GAAG;AACzC,UAAM,WAAoC,EAAE,GAAG,aAAa;AAC5D,UAAM,gBAAgB,MAAM;AAC5B,aAAS,cAAc,cAAc;AACrC,aAAS,eAAe,cAAc;AACtC,aAAS,aAAa,OAAO,MAAM,SAAS;AAC5C,QAAI,cAAc,eAAe,cAAc,YAAY,SAAS,GAAG;AACrE,eAAS,eAAe,CAAC,GAAG,cAAc,WAAW;AAAA,IACvD;AACA,QAAI,cAAc,cAAc,MAAM;AACpC,eAAS,cAAc,cAAc;AAAA,IACvC;AACA,QAAI,cAAc,aAAa,MAAM;AAEnC,eAAS,OAAO,cAAc,YAAY;AAC1C,eAAS,aAAa,cAAc;AAAA,IACtC;AACA,QAAI,cAAc,YAAY,MAAM;AAClC,eAAS,YAAY,cAAc;AAAA,IACrC;AACA,cAAU,KAAK,IAAI,0BAAS,EAAE,aAAa,MAAM,SAAS,SAAS,CAAC,CAAC;AAAA,EACvE;AAEA,SAAO;AACT;AAEA,SAAS,iBAAiB,UAAwB,QAA4B;AAC5E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,QAAQ,SAAS,SAAS,CAAC,GAAG;AACvC,UAAM,WAAoC,EAAE,GAAG,aAAa;AAE5D,aAAS,OAAO,KAAK,aAAa;AAClC,QAAI,KAAK,WAAW,MAAM;AACxB,eAAS,WAAW,KAAK;AAAA,IAC3B;AACA,UAAM,cAAc,gBAAgB,KAAK,SAAS,KAAK,MAAM;AAC7D,cAAU,KAAK,IAAI,0BAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AAAA,EACxD;AAEA,SAAO;AACT;AAEO,SAAS,oBAAoB,UAAwB,QAAgB,SAAmC;AAC7G,MAAI,QAAQ,YAAY,SAAS,UAAU,SAAS,OAAO,SAAS,GAAG;AACrE,WAAO,kBAAkB,UAAU,MAAM;AAAA,EAC3C;AACA,MAAI,QAAQ,WAAW,SAAS,SAAS,SAAS,MAAM,SAAS,GAAG;AAClE,WAAO,iBAAiB,UAAU,MAAM;AAAA,EAC1C;AACA,QAAM,WAAW,cAAc,UAAU,MAAM;AAC/C,QAAM,cAAc,gBAAgB,SAAS,SAAS,SAAS,MAAM;AACrE,SAAO,CAAC,IAAI,0BAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AACjD;AAEO,SAAS,kBAAkB,QAAwB,SAAmB,SAAmC;AAC9G,MAAI,OAAO,UAAU,OAAO,OAAO,SAAS,GAAG;AAC7C,UAAM,QAAQ,OAAO,OAAO,CAAC;AAC7B,UAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,MAAM,MAAM,OAAO,EAAE;AAAA,EACzE;AAEA,QAAM,YAAwB,CAAC;AAC/B,QAAM,UAAU,OAAO,WAAW,CAAC;AACnC,UAAQ,QAAQ,CAAC,UAAU,UAAU;AACnC,UAAM,SAAS,QAAQ,KAAK,MAAM,QAAQ,SAAS,IAAI,QAAQ,QAAQ,SAAS,CAAC,IAAI;AACrF,cAAU,KAAK,GAAG,oBAAoB,UAAU,QAAQ,OAAO,CAAC;AAAA,EAClE,CAAC;AACD,SAAO;AACT;;;ADlOA,IAAM,eAAe;AAuBd,IAAM,cAAN,cAA0B,+BAAmB;AAAA,EACjC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEjB,YAAY,SAA6B;AACvC,UAAM;AACN,UAAM,EAAE,UAAU,MAAM,UAAU,MAAM,OAAO,IAAI;AACnD,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AACA,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AACA,QAAI,SAAS,UAAa,aAAa,QAAW;AAChD,YAAM,IAAI,MAAM,2CAA2C;AAAA,IAC7D;AAEA,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,OAA4B;AAChC,UAAM,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,KAAK,YAAY;AAC1D,QAAI,OAAO,WAAW,GAAG;AACvB,aAAO,CAAC;AAAA,IACV;AAEA,QAAI;AACJ,QAAI;AACF,eAAS,QAAQ,UAAM,2BAAa,QAAQ,KAAK,UAAU,IAAI,IAAI,UAAM,sBAAQ,OAAO,CAAC,GAAG,KAAK,UAAU,IAAI;AAAA,IACjH,SAAS,OAAO;AACd,YAAM,SAAS,QAAQ,CAAC,KAAK;AAC7B,YAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,aAAa,KAAK,CAAC,EAAE;AAAA,IACzE;AAEA,WAAO,kBAAkB,QAAqC,SAAS;AAAA,MACrE,UAAU,kBAAkB,KAAK,MAAM;AAAA,MACvC,SAAS,iBAAiB,KAAK,MAAM;AAAA,IACvC,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,cAAsF;AAClG,QAAI,KAAK,SAAS,QAAW;AAC3B,YAAM,SAAS,WAAW,KAAK,QAAQ;AACvC,YAAM,QAAsB,EAAE,MAAM,8BAAiB,OAAO,OAAO,KAAK,MAAM,UAAU,KAAK,SAAS;AACtG,aAAO,EAAE,QAAQ,CAAC,KAAK,GAAG,SAAS,CAAC,MAAM,GAAG,OAAO,MAAM;AAAA,IAC5D;AAEA,UAAM,EAAE,OAAO,MAAM,IAAI,MAAM,KAAK,aAAa;AACjD,UAAM,SAAyB,MAAM,IAAI,CAAC,UAAU;AAAA,MAClD,MAAM,8BAAiB;AAAA,MACvB,KAAK;AAAA,MACL,UAAU,KAAK;AAAA,IACjB,EAAE;AACF,WAAO,EAAE,QAAQ,SAAS,OAAO,MAAM;AAAA,EACzC;AAAA,EAEA,MAAc,eAA6D;AACzE,UAAM,WAAW,KAAK;AACtB,QAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,aAAO,EAAE,OAAO,UAAU,OAAO,KAAK;AAAA,IACxC;AACA,QAAI,OAAO,aAAa,UAAU;AAChC,UAAI,MAAM,YAAY,QAAQ,GAAG;AAC/B,cAAM,UAAU,KAAK,QAAQ;AAC7B,cAAM,UAAU,UAAM,iBAAAA,SAAS,SAAS,EAAE,KAAK,UAAU,WAAW,MAAM,UAAU,KAAK,CAAC;AAC1F,eAAO,EAAE,OAAO,QAAQ,KAAK,GAAG,OAAO,KAAK;AAAA,MAC9C;AACA,aAAO,EAAE,OAAO,CAAC,QAAQ,GAAG,OAAO,MAAM;AAAA,IAC3C;AACA,WAAO,EAAE,OAAO,CAAC,GAAG,OAAO,MAAM;AAAA,EACnC;AACF;AAEA,eAAe,YAAY,MAAgC;AACzD,MAAI;AACF,UAAM,QAAQ,UAAM,sBAAK,IAAI;AAC7B,WAAO,MAAM,YAAY;AAAA,EAC3B,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEA,SAAS,aAAa,OAAwB;AAC5C,SAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAC9D;","names":["fastGlob"]}
+{"version":3,"sources":["../src/index.ts","../src/loader.ts","../src/mapping.ts"],"sourcesContent":["export { XbergLoader, type XbergLoaderOptions } from \"./loader\";\n","import { stat } from \"node:fs/promises\";\nimport { BaseDocumentLoader } from \"@langchain/core/document_loaders/base\";\nimport type { Document } from \"@langchain/core/documents\";\nimport { extract, extractBatch, ExtractInputKind } from \"@xberg-io/xberg\";\nimport type { ExtractInput, ExtractionConfig } from \"@xberg-io/xberg\";\nimport fastGlob from \"fast-glob\";\nimport { isChunkingEnabled, isPerPageEnabled, resultToDocuments, type ResultEnvelope } from \"./mapping\";\n\nconst DEFAULT_GLOB = \"**/*\";\n\nexport interface XbergLoaderOptions {\n  /** File path, list of file paths, or directory path to load. */\n  filePath?: string | string[];\n  /** Raw bytes to extract from. Mutually exclusive with `filePath`. */\n  data?: Uint8Array;\n  /** MIME type hint. Required when using `data`, optional for `filePath`. */\n  mimeType?: string;\n  /** Glob pattern for directory mode. Defaults to matching all files. */\n  glob?: string;\n  /** Xberg extraction configuration controlling output format, OCR, pages, chunking, etc. */\n  config?: ExtractionConfig;\n}\n\n/**\n * Load documents using Xberg, supporting 107 file formats with true async extraction.\n *\n * By default each source becomes one Document. Enable `chunking` on the\n * `ExtractionConfig` to emit one Document per chunk, or `pages` for one Document\n * per page. Multiple paths (a list or a directory glob) are extracted with a single\n * `extractBatch` call so concurrency happens Rust-side.\n */\nexport class XbergLoader extends BaseDocumentLoader {\n  private readonly filePath?: string | string[];\n  private readonly data?: Uint8Array;\n  private readonly mimeType?: string;\n  private readonly glob?: string;\n  private readonly config?: ExtractionConfig;\n\n  constructor(options: XbergLoaderOptions) {\n    super();\n    const { filePath, data, mimeType, glob, config } = options;\n    if (filePath === undefined && data === undefined) {\n      throw new Error(\"Either 'filePath' or 'data' must be provided.\");\n    }\n    if (filePath !== undefined && data !== undefined) {\n      throw new Error(\"Cannot specify both 'filePath' and 'data'. Use one or the other.\");\n    }\n    if (data !== undefined && mimeType === undefined) {\n      throw new Error(\"'mimeType' is required when using 'data'.\");\n    }\n\n    this.filePath = filePath;\n    this.data = data;\n    this.mimeType = mimeType;\n    this.glob = glob;\n    this.config = config;\n  }\n\n  async load(): Promise<Document[]> {\n    const { inputs, sources, batch } = await this.buildInputs();\n    if (inputs.length === 0) {\n      return [];\n    }\n\n    let result;\n    try {\n      result = batch ? await extractBatch(inputs, this.config ?? null) : await extract(inputs[0], this.config ?? null);\n    } catch (error) {\n      const source = sources[0] ?? \"input\";\n      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`, { cause: error });\n    }\n\n    return resultToDocuments(result as unknown as ResultEnvelope, sources, {\n      chunking: isChunkingEnabled(this.config),\n      perPage: isPerPageEnabled(this.config),\n    });\n  }\n\n  private async buildInputs(): Promise<{ inputs: ExtractInput[]; sources: string[]; batch: boolean }> {\n    if (this.data !== undefined) {\n      const source = `bytes://${this.mimeType}`;\n      const input: ExtractInput = { kind: ExtractInputKind.Bytes, bytes: this.data, mimeType: this.mimeType };\n      return { inputs: [input], sources: [source], batch: false };\n    }\n\n    const { paths, batch } = await this.resolvePaths();\n    const inputs: ExtractInput[] = paths.map((path) => ({\n      kind: ExtractInputKind.Uri,\n      uri: path,\n      mimeType: this.mimeType,\n    }));\n    return { inputs, sources: paths, batch };\n  }\n\n  private async resolvePaths(): Promise<{ paths: string[]; batch: boolean }> {\n    const filePath = this.filePath;\n    if (Array.isArray(filePath)) {\n      return { paths: filePath, batch: true };\n    }\n    if (typeof filePath === \"string\") {\n      if (await isDirectory(filePath)) {\n        const pattern = this.glob ?? DEFAULT_GLOB;\n        const matches = await fastGlob(pattern, { cwd: filePath, onlyFiles: true, absolute: true });\n        return { paths: matches.sort(), batch: true };\n      }\n      return { paths: [filePath], batch: false };\n    }\n    return { paths: [], batch: false };\n  }\n}\n\nasync function isDirectory(path: string): Promise<boolean> {\n  try {\n    const stats = await stat(path);\n    return stats.isDirectory();\n  } catch {\n    return false;\n  }\n}\n\nfunction errorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n","import { Document } from \"@langchain/core/documents\";\nimport type { DocumentCounts, ExtractionConfig, Keyword, Metadata, ProcessingWarning, Table } from \"@xberg-io/xberg\";\n\n// Metadata fields (from Xberg Metadata) that carry JSON-friendly scalar/list values,\n// mapped from the binding's camelCase keys to the snake_case output keys used by the\n// Python adapter. Opaque nested fields (pages, format, imagePreprocessing, jsonSchema,\n// error) are skipped because they are native objects, not plain data. ~keep\nconst METADATA_FIELDS: ReadonlyArray<readonly [string, keyof Metadata]> = [\n  [\"title\", \"title\"],\n  [\"subject\", \"subject\"],\n  [\"authors\", \"authors\"],\n  [\"keywords\", \"keywords\"],\n  [\"language\", \"language\"],\n  [\"created_at\", \"createdAt\"],\n  [\"modified_at\", \"modifiedAt\"],\n  [\"created_by\", \"createdBy\"],\n  [\"modified_by\", \"modifiedBy\"],\n  [\"category\", \"category\"],\n  [\"tags\", \"tags\"],\n  [\"document_version\", \"documentVersion\"],\n  [\"abstract_text\", \"abstractText\"],\n  [\"output_format\", \"outputFormat\"],\n  [\"ocr_used\", \"ocrUsed\"],\n  [\"extraction_duration_ms\", \"extractionDurationMs\"],\n];\n\nconst CONTENT_SEPARATOR = \"\\n\\n\";\n\nexport interface ChunkMetadataShape {\n  chunkIndex: number;\n  totalChunks: number;\n  headingPath?: string[];\n  tokenCount?: number;\n  firstPage?: number;\n  lastPage?: number;\n}\n\nexport interface ChunkShape {\n  content: string;\n  chunkType: unknown;\n  metadata: ChunkMetadataShape;\n}\n\nexport interface PageShape {\n  pageNumber: number;\n  content: string;\n  tables?: Table[];\n  isBlank?: boolean;\n}\n\n// Structural view of a binding ExtractedDocument. The binding types its own fields\n// through unexported `Js*` aliases; this interface uses the clean exported types plus\n// local chunk/page shapes so the mapper is fully typed and testable without the addon. ~keep\nexport interface ExtractedDoc {\n  content?: string;\n  mimeType?: string;\n  metadata?: Metadata;\n  tables?: Table[];\n  counts?: DocumentCounts;\n  detectedLanguages?: string[];\n  chunks?: ChunkShape[];\n  pages?: PageShape[];\n  extractedKeywords?: Keyword[];\n  qualityScore?: number;\n  processingWarnings?: ProcessingWarning[];\n}\n\nexport interface ResultEnvelope {\n  errors?: Array<{ source: string; message: string }>;\n  results?: ExtractedDoc[];\n}\n\nexport interface SplitOptions {\n  chunking: boolean;\n  perPage: boolean;\n}\n\nexport function isChunkingEnabled(config?: ExtractionConfig | null): boolean {\n  return config?.chunking != null;\n}\n\nexport function isPerPageEnabled(config?: ExtractionConfig | null): boolean {\n  return Boolean(config?.pages?.extractPages);\n}\n\nexport function flattenMetadata(metadata?: Metadata): Record<string, unknown> {\n  const flat: Record<string, unknown> = {};\n  if (!metadata) {\n    return flat;\n  }\n\n  for (const [snakeKey, camelKey] of METADATA_FIELDS) {\n    const value = metadata[camelKey];\n    if (value != null) {\n      flat[snakeKey] = value;\n    }\n  }\n\n  const additional = metadata.additional;\n  if (additional && Object.keys(additional).length > 0) {\n    flat.additional = { ...additional };\n  }\n\n  return flat;\n}\n\nexport function assembleContent(content: string | undefined, tables: Table[] | undefined): string {\n  const text = content ?? \"\";\n  if (!tables || tables.length === 0) {\n    return text;\n  }\n  const parts = tables.map((table) => table.markdown ?? \"\").filter((markdown) => markdown.length > 0);\n  if (parts.length === 0) {\n    return text;\n  }\n  return [text, ...parts].join(CONTENT_SEPARATOR);\n}\n\nexport function buildMetadata(document: ExtractedDoc, source: string): Record<string, unknown> {\n  const metadata = flattenMetadata(document.metadata);\n\n  metadata.mime_type = document.mimeType;\n  if (document.qualityScore != null) {\n    metadata.quality_score = document.qualityScore;\n  }\n  if (document.detectedLanguages && document.detectedLanguages.length > 0) {\n    metadata.detected_languages = document.detectedLanguages;\n  }\n  if (document.counts != null) {\n    metadata.page_count = document.counts.pages;\n  }\n\n  if (document.extractedKeywords && document.extractedKeywords.length > 0) {\n    metadata.extracted_keywords = document.extractedKeywords.map((keyword) => ({\n      text: keyword.text,\n      score: keyword.score,\n      algorithm: String(keyword.algorithm),\n    }));\n  }\n\n  const tables = document.tables ?? [];\n  metadata.table_count = tables.length;\n  if (tables.length > 0) {\n    metadata.tables = tables.map((table) => ({\n      cells: table.cells,\n      markdown: table.markdown,\n      page_number: table.pageNumber,\n    }));\n  }\n\n  if (document.processingWarnings && document.processingWarnings.length > 0) {\n    metadata.processing_warnings = document.processingWarnings.map((warning) => ({\n      source: warning.source,\n      message: warning.message,\n    }));\n  }\n\n  metadata.source = source;\n  return metadata;\n}\n\nfunction chunksToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const chunk of document.chunks ?? []) {\n    const metadata: Record<string, unknown> = { ...baseMetadata };\n    const chunkMetadata = chunk.metadata;\n    metadata.chunk_index = chunkMetadata.chunkIndex;\n    metadata.total_chunks = chunkMetadata.totalChunks;\n    metadata.chunk_type = String(chunk.chunkType);\n    if (chunkMetadata.headingPath && chunkMetadata.headingPath.length > 0) {\n      metadata.heading_path = [...chunkMetadata.headingPath];\n    }\n    if (chunkMetadata.tokenCount != null) {\n      metadata.token_count = chunkMetadata.tokenCount;\n    }\n    if (chunkMetadata.firstPage != null) {\n      // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n      metadata.page = chunkMetadata.firstPage - 1;\n      metadata.first_page = chunkMetadata.firstPage;\n    }\n    if (chunkMetadata.lastPage != null) {\n      metadata.last_page = chunkMetadata.lastPage;\n    }\n    documents.push(new Document({ pageContent: chunk.content, metadata }));\n  }\n\n  return documents;\n}\n\nfunction pagesToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const page of document.pages ?? []) {\n    // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n    const metadata: Record<string, unknown> = { ...baseMetadata, page: page.pageNumber - 1 };\n    if (page.isBlank != null) {\n      metadata.is_blank = page.isBlank;\n    }\n    const pageContent = assembleContent(page.content, page.tables);\n    documents.push(new Document({ pageContent, metadata }));\n  }\n\n  return documents;\n}\n\nexport function documentToDocuments(document: ExtractedDoc, source: string, options: SplitOptions): Document[] {\n  if (options.chunking && document.chunks && document.chunks.length > 0) {\n    return chunksToDocuments(document, source);\n  }\n  if (options.perPage && document.pages && document.pages.length > 0) {\n    return pagesToDocuments(document, source);\n  }\n  const metadata = buildMetadata(document, source);\n  const pageContent = assembleContent(document.content, document.tables);\n  return [new Document({ pageContent, metadata })];\n}\n\nexport function resultToDocuments(result: ResultEnvelope, sources: string[], options: SplitOptions): Document[] {\n  if (result.errors && result.errors.length > 0) {\n    const error = result.errors[0];\n    throw new Error(`Failed to extract '${error.source}': ${error.message}`);\n  }\n\n  const documents: Document[] = [];\n  const results = result.results ?? [];\n  results.forEach((document, index) => {\n    const source = sources[index] ?? (sources.length > 0 ? sources[sources.length - 1] : \"\");\n    documents.push(...documentToDocuments(document, source, options));\n  });\n  return documents;\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,sBAAqB;AACrB,kBAAmC;AAEnC,mBAAwD;AAExD,uBAAqB;;;ACLrB,uBAAyB;AAOzB,IAAM,kBAAoE;AAAA,EACxE,CAAC,SAAS,OAAO;AAAA,EACjB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,QAAQ,MAAM;AAAA,EACf,CAAC,oBAAoB,iBAAiB;AAAA,EACtC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,YAAY,SAAS;AAAA,EACtB,CAAC,0BAA0B,sBAAsB;AACnD;AAEA,IAAM,oBAAoB;AAmDnB,SAAS,kBAAkB,QAA2C;AAC3E,SAAO,QAAQ,YAAY;AAC7B;AAEO,SAAS,iBAAiB,QAA2C;AAC1E,SAAO,QAAQ,QAAQ,OAAO,YAAY;AAC5C;AAEO,SAAS,gBAAgB,UAA8C;AAC5E,QAAM,OAAgC,CAAC;AACvC,MAAI,CAAC,UAAU;AACb,WAAO;AAAA,EACT;AAEA,aAAW,CAAC,UAAU,QAAQ,KAAK,iBAAiB;AAClD,UAAM,QAAQ,SAAS,QAAQ;AAC/B,QAAI,SAAS,MAAM;AACjB,WAAK,QAAQ,IAAI;AAAA,IACnB;AAAA,EACF;AAEA,QAAM,aAAa,SAAS;AAC5B,MAAI,cAAc,OAAO,KAAK,UAAU,EAAE,SAAS,GAAG;AACpD,SAAK,aAAa,EAAE,GAAG,WAAW;AAAA,EACpC;AAEA,SAAO;AACT;AAEO,SAAS,gBAAgB,SAA6B,QAAqC;AAChG,QAAM,OAAO,WAAW;AACxB,MAAI,CAAC,UAAU,OAAO,WAAW,GAAG;AAClC,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,OAAO,IAAI,CAAC,UAAU,MAAM,YAAY,EAAE,EAAE,OAAO,CAAC,aAAa,SAAS,SAAS,CAAC;AAClG,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO;AAAA,EACT;AACA,SAAO,CAAC,MAAM,GAAG,KAAK,EAAE,KAAK,iBAAiB;AAChD;AAEO,SAAS,cAAc,UAAwB,QAAyC;AAC7F,QAAM,WAAW,gBAAgB,SAAS,QAAQ;AAElD,WAAS,YAAY,SAAS;AAC9B,MAAI,SAAS,gBAAgB,MAAM;AACjC,aAAS,gBAAgB,SAAS;AAAA,EACpC;AACA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS;AAAA,EACzC;AACA,MAAI,SAAS,UAAU,MAAM;AAC3B,aAAS,aAAa,SAAS,OAAO;AAAA,EACxC;AAEA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS,kBAAkB,IAAI,CAAC,aAAa;AAAA,MACzE,MAAM,QAAQ;AAAA,MACd,OAAO,QAAQ;AAAA,MACf,WAAW,OAAO,QAAQ,SAAS;AAAA,IACrC,EAAE;AAAA,EACJ;AAEA,QAAM,SAAS,SAAS,UAAU,CAAC;AACnC,WAAS,cAAc,OAAO;AAC9B,MAAI,OAAO,SAAS,GAAG;AACrB,aAAS,SAAS,OAAO,IAAI,CAAC,WAAW;AAAA,MACvC,OAAO,MAAM;AAAA,MACb,UAAU,MAAM;AAAA,MAChB,aAAa,MAAM;AAAA,IACrB,EAAE;AAAA,EACJ;AAEA,MAAI,SAAS,sBAAsB,SAAS,mBAAmB,SAAS,GAAG;AACzE,aAAS,sBAAsB,SAAS,mBAAmB,IAAI,CAAC,aAAa;AAAA,MAC3E,QAAQ,QAAQ;AAAA,MAChB,SAAS,QAAQ;AAAA,IACnB,EAAE;AAAA,EACJ;AAEA,WAAS,SAAS;AAClB,SAAO;AACT;AAEA,SAAS,kBAAkB,UAAwB,QAA4B;AAC7E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,SAAS,SAAS,UAAU,CAAC,GAAG;AACzC,UAAM,WAAoC,EAAE,GAAG,aAAa;AAC5D,UAAM,gBAAgB,MAAM;AAC5B,aAAS,cAAc,cAAc;AACrC,aAAS,eAAe,cAAc;AACtC,aAAS,aAAa,OAAO,MAAM,SAAS;AAC5C,QAAI,cAAc,eAAe,cAAc,YAAY,SAAS,GAAG;AACrE,eAAS,eAAe,CAAC,GAAG,cAAc,WAAW;AAAA,IACvD;AACA,QAAI,cAAc,cAAc,MAAM;AACpC,eAAS,cAAc,cAAc;AAAA,IACvC;AACA,QAAI,cAAc,aAAa,MAAM;AAEnC,eAAS,OAAO,cAAc,YAAY;AAC1C,eAAS,aAAa,cAAc;AAAA,IACtC;AACA,QAAI,cAAc,YAAY,MAAM;AAClC,eAAS,YAAY,cAAc;AAAA,IACrC;AACA,cAAU,KAAK,IAAI,0BAAS,EAAE,aAAa,MAAM,SAAS,SAAS,CAAC,CAAC;AAAA,EACvE;AAEA,SAAO;AACT;AAEA,SAAS,iBAAiB,UAAwB,QAA4B;AAC5E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,QAAQ,SAAS,SAAS,CAAC,GAAG;AAEvC,UAAM,WAAoC,EAAE,GAAG,cAAc,MAAM,KAAK,aAAa,EAAE;AACvF,QAAI,KAAK,WAAW,MAAM;AACxB,eAAS,WAAW,KAAK;AAAA,IAC3B;AACA,UAAM,cAAc,gBAAgB,KAAK,SAAS,KAAK,MAAM;AAC7D,cAAU,KAAK,IAAI,0BAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AAAA,EACxD;AAEA,SAAO;AACT;AAEO,SAAS,oBAAoB,UAAwB,QAAgB,SAAmC;AAC7G,MAAI,QAAQ,YAAY,SAAS,UAAU,SAAS,OAAO,SAAS,GAAG;AACrE,WAAO,kBAAkB,UAAU,MAAM;AAAA,EAC3C;AACA,MAAI,QAAQ,WAAW,SAAS,SAAS,SAAS,MAAM,SAAS,GAAG;AAClE,WAAO,iBAAiB,UAAU,MAAM;AAAA,EAC1C;AACA,QAAM,WAAW,cAAc,UAAU,MAAM;AAC/C,QAAM,cAAc,gBAAgB,SAAS,SAAS,SAAS,MAAM;AACrE,SAAO,CAAC,IAAI,0BAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AACjD;AAEO,SAAS,kBAAkB,QAAwB,SAAmB,SAAmC;AAC9G,MAAI,OAAO,UAAU,OAAO,OAAO,SAAS,GAAG;AAC7C,UAAM,QAAQ,OAAO,OAAO,CAAC;AAC7B,UAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,MAAM,MAAM,OAAO,EAAE;AAAA,EACzE;AAEA,QAAM,YAAwB,CAAC;AAC/B,QAAM,UAAU,OAAO,WAAW,CAAC;AACnC,UAAQ,QAAQ,CAAC,UAAU,UAAU;AACnC,UAAM,SAAS,QAAQ,KAAK,MAAM,QAAQ,SAAS,IAAI,QAAQ,QAAQ,SAAS,CAAC,IAAI;AACrF,cAAU,KAAK,GAAG,oBAAoB,UAAU,QAAQ,OAAO,CAAC;AAAA,EAClE,CAAC;AACD,SAAO;AACT;;;ADjOA,IAAM,eAAe;AAuBd,IAAM,cAAN,cAA0B,+BAAmB;AAAA,EACjC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEjB,YAAY,SAA6B;AACvC,UAAM;AACN,UAAM,EAAE,UAAU,MAAM,UAAU,MAAM,OAAO,IAAI;AACnD,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AACA,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AACA,QAAI,SAAS,UAAa,aAAa,QAAW;AAChD,YAAM,IAAI,MAAM,2CAA2C;AAAA,IAC7D;AAEA,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,OAA4B;AAChC,UAAM,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,KAAK,YAAY;AAC1D,QAAI,OAAO,WAAW,GAAG;AACvB,aAAO,CAAC;AAAA,IACV;AAEA,QAAI;AACJ,QAAI;AACF,eAAS,QAAQ,UAAM,2BAAa,QAAQ,KAAK,UAAU,IAAI,IAAI,UAAM,sBAAQ,OAAO,CAAC,GAAG,KAAK,UAAU,IAAI;AAAA,IACjH,SAAS,OAAO;AACd,YAAM,SAAS,QAAQ,CAAC,KAAK;AAC7B,YAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,aAAa,KAAK,CAAC,IAAI,EAAE,OAAO,MAAM,CAAC;AAAA,IAC3F;AAEA,WAAO,kBAAkB,QAAqC,SAAS;AAAA,MACrE,UAAU,kBAAkB,KAAK,MAAM;AAAA,MACvC,SAAS,iBAAiB,KAAK,MAAM;AAAA,IACvC,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,cAAsF;AAClG,QAAI,KAAK,SAAS,QAAW;AAC3B,YAAM,SAAS,WAAW,KAAK,QAAQ;AACvC,YAAM,QAAsB,EAAE,MAAM,8BAAiB,OAAO,OAAO,KAAK,MAAM,UAAU,KAAK,SAAS;AACtG,aAAO,EAAE,QAAQ,CAAC,KAAK,GAAG,SAAS,CAAC,MAAM,GAAG,OAAO,MAAM;AAAA,IAC5D;AAEA,UAAM,EAAE,OAAO,MAAM,IAAI,MAAM,KAAK,aAAa;AACjD,UAAM,SAAyB,MAAM,IAAI,CAAC,UAAU;AAAA,MAClD,MAAM,8BAAiB;AAAA,MACvB,KAAK;AAAA,MACL,UAAU,KAAK;AAAA,IACjB,EAAE;AACF,WAAO,EAAE,QAAQ,SAAS,OAAO,MAAM;AAAA,EACzC;AAAA,EAEA,MAAc,eAA6D;AACzE,UAAM,WAAW,KAAK;AACtB,QAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,aAAO,EAAE,OAAO,UAAU,OAAO,KAAK;AAAA,IACxC;AACA,QAAI,OAAO,aAAa,UAAU;AAChC,UAAI,MAAM,YAAY,QAAQ,GAAG;AAC/B,cAAM,UAAU,KAAK,QAAQ;AAC7B,cAAM,UAAU,UAAM,iBAAAA,SAAS,SAAS,EAAE,KAAK,UAAU,WAAW,MAAM,UAAU,KAAK,CAAC;AAC1F,eAAO,EAAE,OAAO,QAAQ,KAAK,GAAG,OAAO,KAAK;AAAA,MAC9C;AACA,aAAO,EAAE,OAAO,CAAC,QAAQ,GAAG,OAAO,MAAM;AAAA,IAC3C;AACA,WAAO,EAAE,OAAO,CAAC,GAAG,OAAO,MAAM;AAAA,EACnC;AACF;AAEA,eAAe,YAAY,MAAgC;AACzD,MAAI;AACF,UAAM,QAAQ,UAAM,sBAAK,IAAI;AAC7B,WAAO,MAAM,YAAY;AAAA,EAC3B,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEA,SAAS,aAAa,OAAwB;AAC5C,SAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAC9D;","names":["fastGlob"]}

package/dist/index.d.cts

@@ -15,7 +15,7 @@ interface XbergLoaderOptions {
     config?: ExtractionConfig;
 }
 /**
- * Load documents using Xberg, supporting 101 file formats with true async extraction.
+ * Load documents using Xberg, supporting 107 file formats with true async extraction.
  *
  * By default each source becomes one Document. Enable `chunking` on the
  * `ExtractionConfig` to emit one Document per chunk, or `pages` for one Document

package/dist/index.d.ts

@@ -15,7 +15,7 @@ interface XbergLoaderOptions {
     config?: ExtractionConfig;
 }
 /**
- * Load documents using Xberg, supporting 101 file formats with true async extraction.
+ * Load documents using Xberg, supporting 107 file formats with true async extraction.
  *
  * By default each source becomes one Document. Enable `chunking` on the
  * `ExtractionConfig` to emit one Document per chunk, or `pages` for one Document

package/dist/index.js

@@ -126,8 +126,7 @@ function pagesToDocuments(document, source) {
   const baseMetadata = buildMetadata(document, source);
   const documents = [];
   for (const page of document.pages ?? []) {
-    const metadata = { ...baseMetadata };
-    metadata.page = page.pageNumber - 1;
+    const metadata = { ...baseMetadata, page: page.pageNumber - 1 };
     if (page.isBlank != null) {
       metadata.is_blank = page.isBlank;
     }
@@ -197,7 +196,7 @@ var XbergLoader = class extends BaseDocumentLoader {
       result = batch ? await extractBatch(inputs, this.config ?? null) : await extract(inputs[0], this.config ?? null);
     } catch (error) {
       const source = sources[0] ?? "input";
-      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`);
+      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`, { cause: error });
     }
     return resultToDocuments(result, sources, {
       chunking: isChunkingEnabled(this.config),

package/dist/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/loader.ts","../src/mapping.ts"],"sourcesContent":["import { stat } from \"node:fs/promises\";\nimport { BaseDocumentLoader } from \"@langchain/core/document_loaders/base\";\nimport type { Document } from \"@langchain/core/documents\";\nimport { extract, extractBatch, ExtractInputKind } from \"@xberg-io/xberg\";\nimport type { ExtractInput, ExtractionConfig } from \"@xberg-io/xberg\";\nimport fastGlob from \"fast-glob\";\nimport { isChunkingEnabled, isPerPageEnabled, resultToDocuments, type ResultEnvelope } from \"./mapping\";\n\nconst DEFAULT_GLOB = \"**/*\";\n\nexport interface XbergLoaderOptions {\n  /** File path, list of file paths, or directory path to load. */\n  filePath?: string | string[];\n  /** Raw bytes to extract from. Mutually exclusive with `filePath`. */\n  data?: Uint8Array;\n  /** MIME type hint. Required when using `data`, optional for `filePath`. */\n  mimeType?: string;\n  /** Glob pattern for directory mode. Defaults to matching all files. */\n  glob?: string;\n  /** Xberg extraction configuration controlling output format, OCR, pages, chunking, etc. */\n  config?: ExtractionConfig;\n}\n\n/**\n * Load documents using Xberg, supporting 101 file formats with true async extraction.\n *\n * By default each source becomes one Document. Enable `chunking` on the\n * `ExtractionConfig` to emit one Document per chunk, or `pages` for one Document\n * per page. Multiple paths (a list or a directory glob) are extracted with a single\n * `extractBatch` call so concurrency happens Rust-side.\n */\nexport class XbergLoader extends BaseDocumentLoader {\n  private readonly filePath?: string | string[];\n  private readonly data?: Uint8Array;\n  private readonly mimeType?: string;\n  private readonly glob?: string;\n  private readonly config?: ExtractionConfig;\n\n  constructor(options: XbergLoaderOptions) {\n    super();\n    const { filePath, data, mimeType, glob, config } = options;\n    if (filePath === undefined && data === undefined) {\n      throw new Error(\"Either 'filePath' or 'data' must be provided.\");\n    }\n    if (filePath !== undefined && data !== undefined) {\n      throw new Error(\"Cannot specify both 'filePath' and 'data'. Use one or the other.\");\n    }\n    if (data !== undefined && mimeType === undefined) {\n      throw new Error(\"'mimeType' is required when using 'data'.\");\n    }\n\n    this.filePath = filePath;\n    this.data = data;\n    this.mimeType = mimeType;\n    this.glob = glob;\n    this.config = config;\n  }\n\n  async load(): Promise<Document[]> {\n    const { inputs, sources, batch } = await this.buildInputs();\n    if (inputs.length === 0) {\n      return [];\n    }\n\n    let result;\n    try {\n      result = batch ? await extractBatch(inputs, this.config ?? null) : await extract(inputs[0], this.config ?? null);\n    } catch (error) {\n      const source = sources[0] ?? \"input\";\n      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`);\n    }\n\n    return resultToDocuments(result as unknown as ResultEnvelope, sources, {\n      chunking: isChunkingEnabled(this.config),\n      perPage: isPerPageEnabled(this.config),\n    });\n  }\n\n  private async buildInputs(): Promise<{ inputs: ExtractInput[]; sources: string[]; batch: boolean }> {\n    if (this.data !== undefined) {\n      const source = `bytes://${this.mimeType}`;\n      const input: ExtractInput = { kind: ExtractInputKind.Bytes, bytes: this.data, mimeType: this.mimeType };\n      return { inputs: [input], sources: [source], batch: false };\n    }\n\n    const { paths, batch } = await this.resolvePaths();\n    const inputs: ExtractInput[] = paths.map((path) => ({\n      kind: ExtractInputKind.Uri,\n      uri: path,\n      mimeType: this.mimeType,\n    }));\n    return { inputs, sources: paths, batch };\n  }\n\n  private async resolvePaths(): Promise<{ paths: string[]; batch: boolean }> {\n    const filePath = this.filePath;\n    if (Array.isArray(filePath)) {\n      return { paths: filePath, batch: true };\n    }\n    if (typeof filePath === \"string\") {\n      if (await isDirectory(filePath)) {\n        const pattern = this.glob ?? DEFAULT_GLOB;\n        const matches = await fastGlob(pattern, { cwd: filePath, onlyFiles: true, absolute: true });\n        return { paths: matches.sort(), batch: true };\n      }\n      return { paths: [filePath], batch: false };\n    }\n    return { paths: [], batch: false };\n  }\n}\n\nasync function isDirectory(path: string): Promise<boolean> {\n  try {\n    const stats = await stat(path);\n    return stats.isDirectory();\n  } catch {\n    return false;\n  }\n}\n\nfunction errorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n","import { Document } from \"@langchain/core/documents\";\nimport type { DocumentCounts, ExtractionConfig, Keyword, Metadata, ProcessingWarning, Table } from \"@xberg-io/xberg\";\n\n// Metadata fields (from Xberg Metadata) that carry JSON-friendly scalar/list values,\n// mapped from the binding's camelCase keys to the snake_case output keys used by the\n// Python adapter. Opaque nested fields (pages, format, imagePreprocessing, jsonSchema,\n// error) are skipped because they are native objects, not plain data. ~keep\nconst METADATA_FIELDS: ReadonlyArray<readonly [string, keyof Metadata]> = [\n  [\"title\", \"title\"],\n  [\"subject\", \"subject\"],\n  [\"authors\", \"authors\"],\n  [\"keywords\", \"keywords\"],\n  [\"language\", \"language\"],\n  [\"created_at\", \"createdAt\"],\n  [\"modified_at\", \"modifiedAt\"],\n  [\"created_by\", \"createdBy\"],\n  [\"modified_by\", \"modifiedBy\"],\n  [\"category\", \"category\"],\n  [\"tags\", \"tags\"],\n  [\"document_version\", \"documentVersion\"],\n  [\"abstract_text\", \"abstractText\"],\n  [\"output_format\", \"outputFormat\"],\n  [\"ocr_used\", \"ocrUsed\"],\n  [\"extraction_duration_ms\", \"extractionDurationMs\"],\n];\n\nconst CONTENT_SEPARATOR = \"\\n\\n\";\n\nexport interface ChunkMetadataShape {\n  chunkIndex: number;\n  totalChunks: number;\n  headingPath?: string[];\n  tokenCount?: number;\n  firstPage?: number;\n  lastPage?: number;\n}\n\nexport interface ChunkShape {\n  content: string;\n  chunkType: unknown;\n  metadata: ChunkMetadataShape;\n}\n\nexport interface PageShape {\n  pageNumber: number;\n  content: string;\n  tables?: Table[];\n  isBlank?: boolean;\n}\n\n// Structural view of a binding ExtractedDocument. The binding types its own fields\n// through unexported `Js*` aliases; this interface uses the clean exported types plus\n// local chunk/page shapes so the mapper is fully typed and testable without the addon. ~keep\nexport interface ExtractedDoc {\n  content?: string;\n  mimeType?: string;\n  metadata?: Metadata;\n  tables?: Table[];\n  counts?: DocumentCounts;\n  detectedLanguages?: string[];\n  chunks?: ChunkShape[];\n  pages?: PageShape[];\n  extractedKeywords?: Keyword[];\n  qualityScore?: number;\n  processingWarnings?: ProcessingWarning[];\n}\n\nexport interface ResultEnvelope {\n  errors?: Array<{ source: string; message: string }>;\n  results?: ExtractedDoc[];\n}\n\nexport interface SplitOptions {\n  chunking: boolean;\n  perPage: boolean;\n}\n\nexport function isChunkingEnabled(config?: ExtractionConfig | null): boolean {\n  return config?.chunking != null;\n}\n\nexport function isPerPageEnabled(config?: ExtractionConfig | null): boolean {\n  return Boolean(config?.pages?.extractPages);\n}\n\nexport function flattenMetadata(metadata?: Metadata): Record<string, unknown> {\n  const flat: Record<string, unknown> = {};\n  if (!metadata) {\n    return flat;\n  }\n\n  for (const [snakeKey, camelKey] of METADATA_FIELDS) {\n    const value = metadata[camelKey];\n    if (value != null) {\n      flat[snakeKey] = value;\n    }\n  }\n\n  const additional = metadata.additional;\n  if (additional && Object.keys(additional).length > 0) {\n    flat.additional = { ...additional };\n  }\n\n  return flat;\n}\n\nexport function assembleContent(content: string | undefined, tables: Table[] | undefined): string {\n  const text = content ?? \"\";\n  if (!tables || tables.length === 0) {\n    return text;\n  }\n  const parts = tables.map((table) => table.markdown ?? \"\").filter((markdown) => markdown.length > 0);\n  if (parts.length === 0) {\n    return text;\n  }\n  return [text, ...parts].join(CONTENT_SEPARATOR);\n}\n\nexport function buildMetadata(document: ExtractedDoc, source: string): Record<string, unknown> {\n  const metadata = flattenMetadata(document.metadata);\n\n  metadata.mime_type = document.mimeType;\n  if (document.qualityScore != null) {\n    metadata.quality_score = document.qualityScore;\n  }\n  if (document.detectedLanguages && document.detectedLanguages.length > 0) {\n    metadata.detected_languages = document.detectedLanguages;\n  }\n  if (document.counts != null) {\n    metadata.page_count = document.counts.pages;\n  }\n\n  if (document.extractedKeywords && document.extractedKeywords.length > 0) {\n    metadata.extracted_keywords = document.extractedKeywords.map((keyword) => ({\n      text: keyword.text,\n      score: keyword.score,\n      algorithm: String(keyword.algorithm),\n    }));\n  }\n\n  const tables = document.tables ?? [];\n  metadata.table_count = tables.length;\n  if (tables.length > 0) {\n    metadata.tables = tables.map((table) => ({\n      cells: table.cells,\n      markdown: table.markdown,\n      page_number: table.pageNumber,\n    }));\n  }\n\n  if (document.processingWarnings && document.processingWarnings.length > 0) {\n    metadata.processing_warnings = document.processingWarnings.map((warning) => ({\n      source: warning.source,\n      message: warning.message,\n    }));\n  }\n\n  metadata.source = source;\n  return metadata;\n}\n\nfunction chunksToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const chunk of document.chunks ?? []) {\n    const metadata: Record<string, unknown> = { ...baseMetadata };\n    const chunkMetadata = chunk.metadata;\n    metadata.chunk_index = chunkMetadata.chunkIndex;\n    metadata.total_chunks = chunkMetadata.totalChunks;\n    metadata.chunk_type = String(chunk.chunkType);\n    if (chunkMetadata.headingPath && chunkMetadata.headingPath.length > 0) {\n      metadata.heading_path = [...chunkMetadata.headingPath];\n    }\n    if (chunkMetadata.tokenCount != null) {\n      metadata.token_count = chunkMetadata.tokenCount;\n    }\n    if (chunkMetadata.firstPage != null) {\n      // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n      metadata.page = chunkMetadata.firstPage - 1;\n      metadata.first_page = chunkMetadata.firstPage;\n    }\n    if (chunkMetadata.lastPage != null) {\n      metadata.last_page = chunkMetadata.lastPage;\n    }\n    documents.push(new Document({ pageContent: chunk.content, metadata }));\n  }\n\n  return documents;\n}\n\nfunction pagesToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const page of document.pages ?? []) {\n    const metadata: Record<string, unknown> = { ...baseMetadata };\n    // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n    metadata.page = page.pageNumber - 1;\n    if (page.isBlank != null) {\n      metadata.is_blank = page.isBlank;\n    }\n    const pageContent = assembleContent(page.content, page.tables);\n    documents.push(new Document({ pageContent, metadata }));\n  }\n\n  return documents;\n}\n\nexport function documentToDocuments(document: ExtractedDoc, source: string, options: SplitOptions): Document[] {\n  if (options.chunking && document.chunks && document.chunks.length > 0) {\n    return chunksToDocuments(document, source);\n  }\n  if (options.perPage && document.pages && document.pages.length > 0) {\n    return pagesToDocuments(document, source);\n  }\n  const metadata = buildMetadata(document, source);\n  const pageContent = assembleContent(document.content, document.tables);\n  return [new Document({ pageContent, metadata })];\n}\n\nexport function resultToDocuments(result: ResultEnvelope, sources: string[], options: SplitOptions): Document[] {\n  if (result.errors && result.errors.length > 0) {\n    const error = result.errors[0];\n    throw new Error(`Failed to extract '${error.source}': ${error.message}`);\n  }\n\n  const documents: Document[] = [];\n  const results = result.results ?? [];\n  results.forEach((document, index) => {\n    const source = sources[index] ?? (sources.length > 0 ? sources[sources.length - 1] : \"\");\n    documents.push(...documentToDocuments(document, source, options));\n  });\n  return documents;\n}\n"],"mappings":";AAAA,SAAS,YAAY;AACrB,SAAS,0BAA0B;AAEnC,SAAS,SAAS,cAAc,wBAAwB;AAExD,OAAO,cAAc;;;ACLrB,SAAS,gBAAgB;AAOzB,IAAM,kBAAoE;AAAA,EACxE,CAAC,SAAS,OAAO;AAAA,EACjB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,QAAQ,MAAM;AAAA,EACf,CAAC,oBAAoB,iBAAiB;AAAA,EACtC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,YAAY,SAAS;AAAA,EACtB,CAAC,0BAA0B,sBAAsB;AACnD;AAEA,IAAM,oBAAoB;AAmDnB,SAAS,kBAAkB,QAA2C;AAC3E,SAAO,QAAQ,YAAY;AAC7B;AAEO,SAAS,iBAAiB,QAA2C;AAC1E,SAAO,QAAQ,QAAQ,OAAO,YAAY;AAC5C;AAEO,SAAS,gBAAgB,UAA8C;AAC5E,QAAM,OAAgC,CAAC;AACvC,MAAI,CAAC,UAAU;AACb,WAAO;AAAA,EACT;AAEA,aAAW,CAAC,UAAU,QAAQ,KAAK,iBAAiB;AAClD,UAAM,QAAQ,SAAS,QAAQ;AAC/B,QAAI,SAAS,MAAM;AACjB,WAAK,QAAQ,IAAI;AAAA,IACnB;AAAA,EACF;AAEA,QAAM,aAAa,SAAS;AAC5B,MAAI,cAAc,OAAO,KAAK,UAAU,EAAE,SAAS,GAAG;AACpD,SAAK,aAAa,EAAE,GAAG,WAAW;AAAA,EACpC;AAEA,SAAO;AACT;AAEO,SAAS,gBAAgB,SAA6B,QAAqC;AAChG,QAAM,OAAO,WAAW;AACxB,MAAI,CAAC,UAAU,OAAO,WAAW,GAAG;AAClC,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,OAAO,IAAI,CAAC,UAAU,MAAM,YAAY,EAAE,EAAE,OAAO,CAAC,aAAa,SAAS,SAAS,CAAC;AAClG,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO;AAAA,EACT;AACA,SAAO,CAAC,MAAM,GAAG,KAAK,EAAE,KAAK,iBAAiB;AAChD;AAEO,SAAS,cAAc,UAAwB,QAAyC;AAC7F,QAAM,WAAW,gBAAgB,SAAS,QAAQ;AAElD,WAAS,YAAY,SAAS;AAC9B,MAAI,SAAS,gBAAgB,MAAM;AACjC,aAAS,gBAAgB,SAAS;AAAA,EACpC;AACA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS;AAAA,EACzC;AACA,MAAI,SAAS,UAAU,MAAM;AAC3B,aAAS,aAAa,SAAS,OAAO;AAAA,EACxC;AAEA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS,kBAAkB,IAAI,CAAC,aAAa;AAAA,MACzE,MAAM,QAAQ;AAAA,MACd,OAAO,QAAQ;AAAA,MACf,WAAW,OAAO,QAAQ,SAAS;AAAA,IACrC,EAAE;AAAA,EACJ;AAEA,QAAM,SAAS,SAAS,UAAU,CAAC;AACnC,WAAS,cAAc,OAAO;AAC9B,MAAI,OAAO,SAAS,GAAG;AACrB,aAAS,SAAS,OAAO,IAAI,CAAC,WAAW;AAAA,MACvC,OAAO,MAAM;AAAA,MACb,UAAU,MAAM;AAAA,MAChB,aAAa,MAAM;AAAA,IACrB,EAAE;AAAA,EACJ;AAEA,MAAI,SAAS,sBAAsB,SAAS,mBAAmB,SAAS,GAAG;AACzE,aAAS,sBAAsB,SAAS,mBAAmB,IAAI,CAAC,aAAa;AAAA,MAC3E,QAAQ,QAAQ;AAAA,MAChB,SAAS,QAAQ;AAAA,IACnB,EAAE;AAAA,EACJ;AAEA,WAAS,SAAS;AAClB,SAAO;AACT;AAEA,SAAS,kBAAkB,UAAwB,QAA4B;AAC7E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,SAAS,SAAS,UAAU,CAAC,GAAG;AACzC,UAAM,WAAoC,EAAE,GAAG,aAAa;AAC5D,UAAM,gBAAgB,MAAM;AAC5B,aAAS,cAAc,cAAc;AACrC,aAAS,eAAe,cAAc;AACtC,aAAS,aAAa,OAAO,MAAM,SAAS;AAC5C,QAAI,cAAc,eAAe,cAAc,YAAY,SAAS,GAAG;AACrE,eAAS,eAAe,CAAC,GAAG,cAAc,WAAW;AAAA,IACvD;AACA,QAAI,cAAc,cAAc,MAAM;AACpC,eAAS,cAAc,cAAc;AAAA,IACvC;AACA,QAAI,cAAc,aAAa,MAAM;AAEnC,eAAS,OAAO,cAAc,YAAY;AAC1C,eAAS,aAAa,cAAc;AAAA,IACtC;AACA,QAAI,cAAc,YAAY,MAAM;AAClC,eAAS,YAAY,cAAc;AAAA,IACrC;AACA,cAAU,KAAK,IAAI,SAAS,EAAE,aAAa,MAAM,SAAS,SAAS,CAAC,CAAC;AAAA,EACvE;AAEA,SAAO;AACT;AAEA,SAAS,iBAAiB,UAAwB,QAA4B;AAC5E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,QAAQ,SAAS,SAAS,CAAC,GAAG;AACvC,UAAM,WAAoC,EAAE,GAAG,aAAa;AAE5D,aAAS,OAAO,KAAK,aAAa;AAClC,QAAI,KAAK,WAAW,MAAM;AACxB,eAAS,WAAW,KAAK;AAAA,IAC3B;AACA,UAAM,cAAc,gBAAgB,KAAK,SAAS,KAAK,MAAM;AAC7D,cAAU,KAAK,IAAI,SAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AAAA,EACxD;AAEA,SAAO;AACT;AAEO,SAAS,oBAAoB,UAAwB,QAAgB,SAAmC;AAC7G,MAAI,QAAQ,YAAY,SAAS,UAAU,SAAS,OAAO,SAAS,GAAG;AACrE,WAAO,kBAAkB,UAAU,MAAM;AAAA,EAC3C;AACA,MAAI,QAAQ,WAAW,SAAS,SAAS,SAAS,MAAM,SAAS,GAAG;AAClE,WAAO,iBAAiB,UAAU,MAAM;AAAA,EAC1C;AACA,QAAM,WAAW,cAAc,UAAU,MAAM;AAC/C,QAAM,cAAc,gBAAgB,SAAS,SAAS,SAAS,MAAM;AACrE,SAAO,CAAC,IAAI,SAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AACjD;AAEO,SAAS,kBAAkB,QAAwB,SAAmB,SAAmC;AAC9G,MAAI,OAAO,UAAU,OAAO,OAAO,SAAS,GAAG;AAC7C,UAAM,QAAQ,OAAO,OAAO,CAAC;AAC7B,UAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,MAAM,MAAM,OAAO,EAAE;AAAA,EACzE;AAEA,QAAM,YAAwB,CAAC;AAC/B,QAAM,UAAU,OAAO,WAAW,CAAC;AACnC,UAAQ,QAAQ,CAAC,UAAU,UAAU;AACnC,UAAM,SAAS,QAAQ,KAAK,MAAM,QAAQ,SAAS,IAAI,QAAQ,QAAQ,SAAS,CAAC,IAAI;AACrF,cAAU,KAAK,GAAG,oBAAoB,UAAU,QAAQ,OAAO,CAAC;AAAA,EAClE,CAAC;AACD,SAAO;AACT;;;ADlOA,IAAM,eAAe;AAuBd,IAAM,cAAN,cAA0B,mBAAmB;AAAA,EACjC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEjB,YAAY,SAA6B;AACvC,UAAM;AACN,UAAM,EAAE,UAAU,MAAM,UAAU,MAAM,OAAO,IAAI;AACnD,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AACA,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AACA,QAAI,SAAS,UAAa,aAAa,QAAW;AAChD,YAAM,IAAI,MAAM,2CAA2C;AAAA,IAC7D;AAEA,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,OAA4B;AAChC,UAAM,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,KAAK,YAAY;AAC1D,QAAI,OAAO,WAAW,GAAG;AACvB,aAAO,CAAC;AAAA,IACV;AAEA,QAAI;AACJ,QAAI;AACF,eAAS,QAAQ,MAAM,aAAa,QAAQ,KAAK,UAAU,IAAI,IAAI,MAAM,QAAQ,OAAO,CAAC,GAAG,KAAK,UAAU,IAAI;AAAA,IACjH,SAAS,OAAO;AACd,YAAM,SAAS,QAAQ,CAAC,KAAK;AAC7B,YAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,aAAa,KAAK,CAAC,EAAE;AAAA,IACzE;AAEA,WAAO,kBAAkB,QAAqC,SAAS;AAAA,MACrE,UAAU,kBAAkB,KAAK,MAAM;AAAA,MACvC,SAAS,iBAAiB,KAAK,MAAM;AAAA,IACvC,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,cAAsF;AAClG,QAAI,KAAK,SAAS,QAAW;AAC3B,YAAM,SAAS,WAAW,KAAK,QAAQ;AACvC,YAAM,QAAsB,EAAE,MAAM,iBAAiB,OAAO,OAAO,KAAK,MAAM,UAAU,KAAK,SAAS;AACtG,aAAO,EAAE,QAAQ,CAAC,KAAK,GAAG,SAAS,CAAC,MAAM,GAAG,OAAO,MAAM;AAAA,IAC5D;AAEA,UAAM,EAAE,OAAO,MAAM,IAAI,MAAM,KAAK,aAAa;AACjD,UAAM,SAAyB,MAAM,IAAI,CAAC,UAAU;AAAA,MAClD,MAAM,iBAAiB;AAAA,MACvB,KAAK;AAAA,MACL,UAAU,KAAK;AAAA,IACjB,EAAE;AACF,WAAO,EAAE,QAAQ,SAAS,OAAO,MAAM;AAAA,EACzC;AAAA,EAEA,MAAc,eAA6D;AACzE,UAAM,WAAW,KAAK;AACtB,QAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,aAAO,EAAE,OAAO,UAAU,OAAO,KAAK;AAAA,IACxC;AACA,QAAI,OAAO,aAAa,UAAU;AAChC,UAAI,MAAM,YAAY,QAAQ,GAAG;AAC/B,cAAM,UAAU,KAAK,QAAQ;AAC7B,cAAM,UAAU,MAAM,SAAS,SAAS,EAAE,KAAK,UAAU,WAAW,MAAM,UAAU,KAAK,CAAC;AAC1F,eAAO,EAAE,OAAO,QAAQ,KAAK,GAAG,OAAO,KAAK;AAAA,MAC9C;AACA,aAAO,EAAE,OAAO,CAAC,QAAQ,GAAG,OAAO,MAAM;AAAA,IAC3C;AACA,WAAO,EAAE,OAAO,CAAC,GAAG,OAAO,MAAM;AAAA,EACnC;AACF;AAEA,eAAe,YAAY,MAAgC;AACzD,MAAI;AACF,UAAM,QAAQ,MAAM,KAAK,IAAI;AAC7B,WAAO,MAAM,YAAY;AAAA,EAC3B,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEA,SAAS,aAAa,OAAwB;AAC5C,SAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAC9D;","names":[]}
+{"version":3,"sources":["../src/loader.ts","../src/mapping.ts"],"sourcesContent":["import { stat } from \"node:fs/promises\";\nimport { BaseDocumentLoader } from \"@langchain/core/document_loaders/base\";\nimport type { Document } from \"@langchain/core/documents\";\nimport { extract, extractBatch, ExtractInputKind } from \"@xberg-io/xberg\";\nimport type { ExtractInput, ExtractionConfig } from \"@xberg-io/xberg\";\nimport fastGlob from \"fast-glob\";\nimport { isChunkingEnabled, isPerPageEnabled, resultToDocuments, type ResultEnvelope } from \"./mapping\";\n\nconst DEFAULT_GLOB = \"**/*\";\n\nexport interface XbergLoaderOptions {\n  /** File path, list of file paths, or directory path to load. */\n  filePath?: string | string[];\n  /** Raw bytes to extract from. Mutually exclusive with `filePath`. */\n  data?: Uint8Array;\n  /** MIME type hint. Required when using `data`, optional for `filePath`. */\n  mimeType?: string;\n  /** Glob pattern for directory mode. Defaults to matching all files. */\n  glob?: string;\n  /** Xberg extraction configuration controlling output format, OCR, pages, chunking, etc. */\n  config?: ExtractionConfig;\n}\n\n/**\n * Load documents using Xberg, supporting 107 file formats with true async extraction.\n *\n * By default each source becomes one Document. Enable `chunking` on the\n * `ExtractionConfig` to emit one Document per chunk, or `pages` for one Document\n * per page. Multiple paths (a list or a directory glob) are extracted with a single\n * `extractBatch` call so concurrency happens Rust-side.\n */\nexport class XbergLoader extends BaseDocumentLoader {\n  private readonly filePath?: string | string[];\n  private readonly data?: Uint8Array;\n  private readonly mimeType?: string;\n  private readonly glob?: string;\n  private readonly config?: ExtractionConfig;\n\n  constructor(options: XbergLoaderOptions) {\n    super();\n    const { filePath, data, mimeType, glob, config } = options;\n    if (filePath === undefined && data === undefined) {\n      throw new Error(\"Either 'filePath' or 'data' must be provided.\");\n    }\n    if (filePath !== undefined && data !== undefined) {\n      throw new Error(\"Cannot specify both 'filePath' and 'data'. Use one or the other.\");\n    }\n    if (data !== undefined && mimeType === undefined) {\n      throw new Error(\"'mimeType' is required when using 'data'.\");\n    }\n\n    this.filePath = filePath;\n    this.data = data;\n    this.mimeType = mimeType;\n    this.glob = glob;\n    this.config = config;\n  }\n\n  async load(): Promise<Document[]> {\n    const { inputs, sources, batch } = await this.buildInputs();\n    if (inputs.length === 0) {\n      return [];\n    }\n\n    let result;\n    try {\n      result = batch ? await extractBatch(inputs, this.config ?? null) : await extract(inputs[0], this.config ?? null);\n    } catch (error) {\n      const source = sources[0] ?? \"input\";\n      throw new Error(`Failed to extract '${source}': ${errorMessage(error)}`, { cause: error });\n    }\n\n    return resultToDocuments(result as unknown as ResultEnvelope, sources, {\n      chunking: isChunkingEnabled(this.config),\n      perPage: isPerPageEnabled(this.config),\n    });\n  }\n\n  private async buildInputs(): Promise<{ inputs: ExtractInput[]; sources: string[]; batch: boolean }> {\n    if (this.data !== undefined) {\n      const source = `bytes://${this.mimeType}`;\n      const input: ExtractInput = { kind: ExtractInputKind.Bytes, bytes: this.data, mimeType: this.mimeType };\n      return { inputs: [input], sources: [source], batch: false };\n    }\n\n    const { paths, batch } = await this.resolvePaths();\n    const inputs: ExtractInput[] = paths.map((path) => ({\n      kind: ExtractInputKind.Uri,\n      uri: path,\n      mimeType: this.mimeType,\n    }));\n    return { inputs, sources: paths, batch };\n  }\n\n  private async resolvePaths(): Promise<{ paths: string[]; batch: boolean }> {\n    const filePath = this.filePath;\n    if (Array.isArray(filePath)) {\n      return { paths: filePath, batch: true };\n    }\n    if (typeof filePath === \"string\") {\n      if (await isDirectory(filePath)) {\n        const pattern = this.glob ?? DEFAULT_GLOB;\n        const matches = await fastGlob(pattern, { cwd: filePath, onlyFiles: true, absolute: true });\n        return { paths: matches.sort(), batch: true };\n      }\n      return { paths: [filePath], batch: false };\n    }\n    return { paths: [], batch: false };\n  }\n}\n\nasync function isDirectory(path: string): Promise<boolean> {\n  try {\n    const stats = await stat(path);\n    return stats.isDirectory();\n  } catch {\n    return false;\n  }\n}\n\nfunction errorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n","import { Document } from \"@langchain/core/documents\";\nimport type { DocumentCounts, ExtractionConfig, Keyword, Metadata, ProcessingWarning, Table } from \"@xberg-io/xberg\";\n\n// Metadata fields (from Xberg Metadata) that carry JSON-friendly scalar/list values,\n// mapped from the binding's camelCase keys to the snake_case output keys used by the\n// Python adapter. Opaque nested fields (pages, format, imagePreprocessing, jsonSchema,\n// error) are skipped because they are native objects, not plain data. ~keep\nconst METADATA_FIELDS: ReadonlyArray<readonly [string, keyof Metadata]> = [\n  [\"title\", \"title\"],\n  [\"subject\", \"subject\"],\n  [\"authors\", \"authors\"],\n  [\"keywords\", \"keywords\"],\n  [\"language\", \"language\"],\n  [\"created_at\", \"createdAt\"],\n  [\"modified_at\", \"modifiedAt\"],\n  [\"created_by\", \"createdBy\"],\n  [\"modified_by\", \"modifiedBy\"],\n  [\"category\", \"category\"],\n  [\"tags\", \"tags\"],\n  [\"document_version\", \"documentVersion\"],\n  [\"abstract_text\", \"abstractText\"],\n  [\"output_format\", \"outputFormat\"],\n  [\"ocr_used\", \"ocrUsed\"],\n  [\"extraction_duration_ms\", \"extractionDurationMs\"],\n];\n\nconst CONTENT_SEPARATOR = \"\\n\\n\";\n\nexport interface ChunkMetadataShape {\n  chunkIndex: number;\n  totalChunks: number;\n  headingPath?: string[];\n  tokenCount?: number;\n  firstPage?: number;\n  lastPage?: number;\n}\n\nexport interface ChunkShape {\n  content: string;\n  chunkType: unknown;\n  metadata: ChunkMetadataShape;\n}\n\nexport interface PageShape {\n  pageNumber: number;\n  content: string;\n  tables?: Table[];\n  isBlank?: boolean;\n}\n\n// Structural view of a binding ExtractedDocument. The binding types its own fields\n// through unexported `Js*` aliases; this interface uses the clean exported types plus\n// local chunk/page shapes so the mapper is fully typed and testable without the addon. ~keep\nexport interface ExtractedDoc {\n  content?: string;\n  mimeType?: string;\n  metadata?: Metadata;\n  tables?: Table[];\n  counts?: DocumentCounts;\n  detectedLanguages?: string[];\n  chunks?: ChunkShape[];\n  pages?: PageShape[];\n  extractedKeywords?: Keyword[];\n  qualityScore?: number;\n  processingWarnings?: ProcessingWarning[];\n}\n\nexport interface ResultEnvelope {\n  errors?: Array<{ source: string; message: string }>;\n  results?: ExtractedDoc[];\n}\n\nexport interface SplitOptions {\n  chunking: boolean;\n  perPage: boolean;\n}\n\nexport function isChunkingEnabled(config?: ExtractionConfig | null): boolean {\n  return config?.chunking != null;\n}\n\nexport function isPerPageEnabled(config?: ExtractionConfig | null): boolean {\n  return Boolean(config?.pages?.extractPages);\n}\n\nexport function flattenMetadata(metadata?: Metadata): Record<string, unknown> {\n  const flat: Record<string, unknown> = {};\n  if (!metadata) {\n    return flat;\n  }\n\n  for (const [snakeKey, camelKey] of METADATA_FIELDS) {\n    const value = metadata[camelKey];\n    if (value != null) {\n      flat[snakeKey] = value;\n    }\n  }\n\n  const additional = metadata.additional;\n  if (additional && Object.keys(additional).length > 0) {\n    flat.additional = { ...additional };\n  }\n\n  return flat;\n}\n\nexport function assembleContent(content: string | undefined, tables: Table[] | undefined): string {\n  const text = content ?? \"\";\n  if (!tables || tables.length === 0) {\n    return text;\n  }\n  const parts = tables.map((table) => table.markdown ?? \"\").filter((markdown) => markdown.length > 0);\n  if (parts.length === 0) {\n    return text;\n  }\n  return [text, ...parts].join(CONTENT_SEPARATOR);\n}\n\nexport function buildMetadata(document: ExtractedDoc, source: string): Record<string, unknown> {\n  const metadata = flattenMetadata(document.metadata);\n\n  metadata.mime_type = document.mimeType;\n  if (document.qualityScore != null) {\n    metadata.quality_score = document.qualityScore;\n  }\n  if (document.detectedLanguages && document.detectedLanguages.length > 0) {\n    metadata.detected_languages = document.detectedLanguages;\n  }\n  if (document.counts != null) {\n    metadata.page_count = document.counts.pages;\n  }\n\n  if (document.extractedKeywords && document.extractedKeywords.length > 0) {\n    metadata.extracted_keywords = document.extractedKeywords.map((keyword) => ({\n      text: keyword.text,\n      score: keyword.score,\n      algorithm: String(keyword.algorithm),\n    }));\n  }\n\n  const tables = document.tables ?? [];\n  metadata.table_count = tables.length;\n  if (tables.length > 0) {\n    metadata.tables = tables.map((table) => ({\n      cells: table.cells,\n      markdown: table.markdown,\n      page_number: table.pageNumber,\n    }));\n  }\n\n  if (document.processingWarnings && document.processingWarnings.length > 0) {\n    metadata.processing_warnings = document.processingWarnings.map((warning) => ({\n      source: warning.source,\n      message: warning.message,\n    }));\n  }\n\n  metadata.source = source;\n  return metadata;\n}\n\nfunction chunksToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const chunk of document.chunks ?? []) {\n    const metadata: Record<string, unknown> = { ...baseMetadata };\n    const chunkMetadata = chunk.metadata;\n    metadata.chunk_index = chunkMetadata.chunkIndex;\n    metadata.total_chunks = chunkMetadata.totalChunks;\n    metadata.chunk_type = String(chunk.chunkType);\n    if (chunkMetadata.headingPath && chunkMetadata.headingPath.length > 0) {\n      metadata.heading_path = [...chunkMetadata.headingPath];\n    }\n    if (chunkMetadata.tokenCount != null) {\n      metadata.token_count = chunkMetadata.tokenCount;\n    }\n    if (chunkMetadata.firstPage != null) {\n      // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n      metadata.page = chunkMetadata.firstPage - 1;\n      metadata.first_page = chunkMetadata.firstPage;\n    }\n    if (chunkMetadata.lastPage != null) {\n      metadata.last_page = chunkMetadata.lastPage;\n    }\n    documents.push(new Document({ pageContent: chunk.content, metadata }));\n  }\n\n  return documents;\n}\n\nfunction pagesToDocuments(document: ExtractedDoc, source: string): Document[] {\n  const baseMetadata = buildMetadata(document, source);\n  const documents: Document[] = [];\n\n  for (const page of document.pages ?? []) {\n    // Xberg uses 1-indexed pages; LangChain convention is 0-indexed. ~keep\n    const metadata: Record<string, unknown> = { ...baseMetadata, page: page.pageNumber - 1 };\n    if (page.isBlank != null) {\n      metadata.is_blank = page.isBlank;\n    }\n    const pageContent = assembleContent(page.content, page.tables);\n    documents.push(new Document({ pageContent, metadata }));\n  }\n\n  return documents;\n}\n\nexport function documentToDocuments(document: ExtractedDoc, source: string, options: SplitOptions): Document[] {\n  if (options.chunking && document.chunks && document.chunks.length > 0) {\n    return chunksToDocuments(document, source);\n  }\n  if (options.perPage && document.pages && document.pages.length > 0) {\n    return pagesToDocuments(document, source);\n  }\n  const metadata = buildMetadata(document, source);\n  const pageContent = assembleContent(document.content, document.tables);\n  return [new Document({ pageContent, metadata })];\n}\n\nexport function resultToDocuments(result: ResultEnvelope, sources: string[], options: SplitOptions): Document[] {\n  if (result.errors && result.errors.length > 0) {\n    const error = result.errors[0];\n    throw new Error(`Failed to extract '${error.source}': ${error.message}`);\n  }\n\n  const documents: Document[] = [];\n  const results = result.results ?? [];\n  results.forEach((document, index) => {\n    const source = sources[index] ?? (sources.length > 0 ? sources[sources.length - 1] : \"\");\n    documents.push(...documentToDocuments(document, source, options));\n  });\n  return documents;\n}\n"],"mappings":";AAAA,SAAS,YAAY;AACrB,SAAS,0BAA0B;AAEnC,SAAS,SAAS,cAAc,wBAAwB;AAExD,OAAO,cAAc;;;ACLrB,SAAS,gBAAgB;AAOzB,IAAM,kBAAoE;AAAA,EACxE,CAAC,SAAS,OAAO;AAAA,EACjB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,WAAW,SAAS;AAAA,EACrB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,cAAc,WAAW;AAAA,EAC1B,CAAC,eAAe,YAAY;AAAA,EAC5B,CAAC,YAAY,UAAU;AAAA,EACvB,CAAC,QAAQ,MAAM;AAAA,EACf,CAAC,oBAAoB,iBAAiB;AAAA,EACtC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,iBAAiB,cAAc;AAAA,EAChC,CAAC,YAAY,SAAS;AAAA,EACtB,CAAC,0BAA0B,sBAAsB;AACnD;AAEA,IAAM,oBAAoB;AAmDnB,SAAS,kBAAkB,QAA2C;AAC3E,SAAO,QAAQ,YAAY;AAC7B;AAEO,SAAS,iBAAiB,QAA2C;AAC1E,SAAO,QAAQ,QAAQ,OAAO,YAAY;AAC5C;AAEO,SAAS,gBAAgB,UAA8C;AAC5E,QAAM,OAAgC,CAAC;AACvC,MAAI,CAAC,UAAU;AACb,WAAO;AAAA,EACT;AAEA,aAAW,CAAC,UAAU,QAAQ,KAAK,iBAAiB;AAClD,UAAM,QAAQ,SAAS,QAAQ;AAC/B,QAAI,SAAS,MAAM;AACjB,WAAK,QAAQ,IAAI;AAAA,IACnB;AAAA,EACF;AAEA,QAAM,aAAa,SAAS;AAC5B,MAAI,cAAc,OAAO,KAAK,UAAU,EAAE,SAAS,GAAG;AACpD,SAAK,aAAa,EAAE,GAAG,WAAW;AAAA,EACpC;AAEA,SAAO;AACT;AAEO,SAAS,gBAAgB,SAA6B,QAAqC;AAChG,QAAM,OAAO,WAAW;AACxB,MAAI,CAAC,UAAU,OAAO,WAAW,GAAG;AAClC,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,OAAO,IAAI,CAAC,UAAU,MAAM,YAAY,EAAE,EAAE,OAAO,CAAC,aAAa,SAAS,SAAS,CAAC;AAClG,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO;AAAA,EACT;AACA,SAAO,CAAC,MAAM,GAAG,KAAK,EAAE,KAAK,iBAAiB;AAChD;AAEO,SAAS,cAAc,UAAwB,QAAyC;AAC7F,QAAM,WAAW,gBAAgB,SAAS,QAAQ;AAElD,WAAS,YAAY,SAAS;AAC9B,MAAI,SAAS,gBAAgB,MAAM;AACjC,aAAS,gBAAgB,SAAS;AAAA,EACpC;AACA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS;AAAA,EACzC;AACA,MAAI,SAAS,UAAU,MAAM;AAC3B,aAAS,aAAa,SAAS,OAAO;AAAA,EACxC;AAEA,MAAI,SAAS,qBAAqB,SAAS,kBAAkB,SAAS,GAAG;AACvE,aAAS,qBAAqB,SAAS,kBAAkB,IAAI,CAAC,aAAa;AAAA,MACzE,MAAM,QAAQ;AAAA,MACd,OAAO,QAAQ;AAAA,MACf,WAAW,OAAO,QAAQ,SAAS;AAAA,IACrC,EAAE;AAAA,EACJ;AAEA,QAAM,SAAS,SAAS,UAAU,CAAC;AACnC,WAAS,cAAc,OAAO;AAC9B,MAAI,OAAO,SAAS,GAAG;AACrB,aAAS,SAAS,OAAO,IAAI,CAAC,WAAW;AAAA,MACvC,OAAO,MAAM;AAAA,MACb,UAAU,MAAM;AAAA,MAChB,aAAa,MAAM;AAAA,IACrB,EAAE;AAAA,EACJ;AAEA,MAAI,SAAS,sBAAsB,SAAS,mBAAmB,SAAS,GAAG;AACzE,aAAS,sBAAsB,SAAS,mBAAmB,IAAI,CAAC,aAAa;AAAA,MAC3E,QAAQ,QAAQ;AAAA,MAChB,SAAS,QAAQ;AAAA,IACnB,EAAE;AAAA,EACJ;AAEA,WAAS,SAAS;AAClB,SAAO;AACT;AAEA,SAAS,kBAAkB,UAAwB,QAA4B;AAC7E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,SAAS,SAAS,UAAU,CAAC,GAAG;AACzC,UAAM,WAAoC,EAAE,GAAG,aAAa;AAC5D,UAAM,gBAAgB,MAAM;AAC5B,aAAS,cAAc,cAAc;AACrC,aAAS,eAAe,cAAc;AACtC,aAAS,aAAa,OAAO,MAAM,SAAS;AAC5C,QAAI,cAAc,eAAe,cAAc,YAAY,SAAS,GAAG;AACrE,eAAS,eAAe,CAAC,GAAG,cAAc,WAAW;AAAA,IACvD;AACA,QAAI,cAAc,cAAc,MAAM;AACpC,eAAS,cAAc,cAAc;AAAA,IACvC;AACA,QAAI,cAAc,aAAa,MAAM;AAEnC,eAAS,OAAO,cAAc,YAAY;AAC1C,eAAS,aAAa,cAAc;AAAA,IACtC;AACA,QAAI,cAAc,YAAY,MAAM;AAClC,eAAS,YAAY,cAAc;AAAA,IACrC;AACA,cAAU,KAAK,IAAI,SAAS,EAAE,aAAa,MAAM,SAAS,SAAS,CAAC,CAAC;AAAA,EACvE;AAEA,SAAO;AACT;AAEA,SAAS,iBAAiB,UAAwB,QAA4B;AAC5E,QAAM,eAAe,cAAc,UAAU,MAAM;AACnD,QAAM,YAAwB,CAAC;AAE/B,aAAW,QAAQ,SAAS,SAAS,CAAC,GAAG;AAEvC,UAAM,WAAoC,EAAE,GAAG,cAAc,MAAM,KAAK,aAAa,EAAE;AACvF,QAAI,KAAK,WAAW,MAAM;AACxB,eAAS,WAAW,KAAK;AAAA,IAC3B;AACA,UAAM,cAAc,gBAAgB,KAAK,SAAS,KAAK,MAAM;AAC7D,cAAU,KAAK,IAAI,SAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AAAA,EACxD;AAEA,SAAO;AACT;AAEO,SAAS,oBAAoB,UAAwB,QAAgB,SAAmC;AAC7G,MAAI,QAAQ,YAAY,SAAS,UAAU,SAAS,OAAO,SAAS,GAAG;AACrE,WAAO,kBAAkB,UAAU,MAAM;AAAA,EAC3C;AACA,MAAI,QAAQ,WAAW,SAAS,SAAS,SAAS,MAAM,SAAS,GAAG;AAClE,WAAO,iBAAiB,UAAU,MAAM;AAAA,EAC1C;AACA,QAAM,WAAW,cAAc,UAAU,MAAM;AAC/C,QAAM,cAAc,gBAAgB,SAAS,SAAS,SAAS,MAAM;AACrE,SAAO,CAAC,IAAI,SAAS,EAAE,aAAa,SAAS,CAAC,CAAC;AACjD;AAEO,SAAS,kBAAkB,QAAwB,SAAmB,SAAmC;AAC9G,MAAI,OAAO,UAAU,OAAO,OAAO,SAAS,GAAG;AAC7C,UAAM,QAAQ,OAAO,OAAO,CAAC;AAC7B,UAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,MAAM,MAAM,OAAO,EAAE;AAAA,EACzE;AAEA,QAAM,YAAwB,CAAC;AAC/B,QAAM,UAAU,OAAO,WAAW,CAAC;AACnC,UAAQ,QAAQ,CAAC,UAAU,UAAU;AACnC,UAAM,SAAS,QAAQ,KAAK,MAAM,QAAQ,SAAS,IAAI,QAAQ,QAAQ,SAAS,CAAC,IAAI;AACrF,cAAU,KAAK,GAAG,oBAAoB,UAAU,QAAQ,OAAO,CAAC;AAAA,EAClE,CAAC;AACD,SAAO;AACT;;;ADjOA,IAAM,eAAe;AAuBd,IAAM,cAAN,cAA0B,mBAAmB;AAAA,EACjC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEjB,YAAY,SAA6B;AACvC,UAAM;AACN,UAAM,EAAE,UAAU,MAAM,UAAU,MAAM,OAAO,IAAI;AACnD,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AACA,QAAI,aAAa,UAAa,SAAS,QAAW;AAChD,YAAM,IAAI,MAAM,kEAAkE;AAAA,IACpF;AACA,QAAI,SAAS,UAAa,aAAa,QAAW;AAChD,YAAM,IAAI,MAAM,2CAA2C;AAAA,IAC7D;AAEA,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,OAAO;AACZ,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAM,OAA4B;AAChC,UAAM,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,KAAK,YAAY;AAC1D,QAAI,OAAO,WAAW,GAAG;AACvB,aAAO,CAAC;AAAA,IACV;AAEA,QAAI;AACJ,QAAI;AACF,eAAS,QAAQ,MAAM,aAAa,QAAQ,KAAK,UAAU,IAAI,IAAI,MAAM,QAAQ,OAAO,CAAC,GAAG,KAAK,UAAU,IAAI;AAAA,IACjH,SAAS,OAAO;AACd,YAAM,SAAS,QAAQ,CAAC,KAAK;AAC7B,YAAM,IAAI,MAAM,sBAAsB,MAAM,MAAM,aAAa,KAAK,CAAC,IAAI,EAAE,OAAO,MAAM,CAAC;AAAA,IAC3F;AAEA,WAAO,kBAAkB,QAAqC,SAAS;AAAA,MACrE,UAAU,kBAAkB,KAAK,MAAM;AAAA,MACvC,SAAS,iBAAiB,KAAK,MAAM;AAAA,IACvC,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,cAAsF;AAClG,QAAI,KAAK,SAAS,QAAW;AAC3B,YAAM,SAAS,WAAW,KAAK,QAAQ;AACvC,YAAM,QAAsB,EAAE,MAAM,iBAAiB,OAAO,OAAO,KAAK,MAAM,UAAU,KAAK,SAAS;AACtG,aAAO,EAAE,QAAQ,CAAC,KAAK,GAAG,SAAS,CAAC,MAAM,GAAG,OAAO,MAAM;AAAA,IAC5D;AAEA,UAAM,EAAE,OAAO,MAAM,IAAI,MAAM,KAAK,aAAa;AACjD,UAAM,SAAyB,MAAM,IAAI,CAAC,UAAU;AAAA,MAClD,MAAM,iBAAiB;AAAA,MACvB,KAAK;AAAA,MACL,UAAU,KAAK;AAAA,IACjB,EAAE;AACF,WAAO,EAAE,QAAQ,SAAS,OAAO,MAAM;AAAA,EACzC;AAAA,EAEA,MAAc,eAA6D;AACzE,UAAM,WAAW,KAAK;AACtB,QAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,aAAO,EAAE,OAAO,UAAU,OAAO,KAAK;AAAA,IACxC;AACA,QAAI,OAAO,aAAa,UAAU;AAChC,UAAI,MAAM,YAAY,QAAQ,GAAG;AAC/B,cAAM,UAAU,KAAK,QAAQ;AAC7B,cAAM,UAAU,MAAM,SAAS,SAAS,EAAE,KAAK,UAAU,WAAW,MAAM,UAAU,KAAK,CAAC;AAC1F,eAAO,EAAE,OAAO,QAAQ,KAAK,GAAG,OAAO,KAAK;AAAA,MAC9C;AACA,aAAO,EAAE,OAAO,CAAC,QAAQ,GAAG,OAAO,MAAM;AAAA,IAC3C;AACA,WAAO,EAAE,OAAO,CAAC,GAAG,OAAO,MAAM;AAAA,EACnC;AACF;AAEA,eAAe,YAAY,MAAgC;AACzD,MAAI;AACF,UAAM,QAAQ,MAAM,KAAK,IAAI;AAC7B,WAAO,MAAM,YAAY;AAAA,EAC3B,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEA,SAAS,aAAa,OAAwB;AAC5C,SAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAC9D;","names":[]}

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@xberg-io/langchain-xberg",
-  "version": "1.0.14",
-  "description": "LangChain.js document loader for Xberg — extract text, tables, and metadata from 101 document formats with optional OCR.",
+  "version": "1.1.0",
+  "description": "LangChain.js document loader for Xberg — extract text, tables, and metadata from 107 document formats with optional OCR.",
   "keywords": [
     "langchain",
     "langchainjs",
@@ -38,16 +38,14 @@
       "require": "./dist/index.cjs"
     }
   },
-  "files": [
-    "dist"
-  ],
+  "files": ["dist"],
   "scripts": {
     "build": "tsup",
     "typecheck": "tsc --noEmit",
     "test": "vitest run"
   },
   "dependencies": {
-    "@xberg-io/xberg": "1.0.14",
+    "@xberg-io/xberg": "1.1.0",
     "fast-glob": "^3.3.2"
   },
   "peerDependencies": {