@ontos-ai/knowhere-sdk 0.4.0 → 0.5.0

package/README.md

@@ -47,11 +47,11 @@ console.log(`Found ${result.textChunks.length} text chunks`);
 console.log(`Found ${result.imageChunks.length} images`);
 console.log(`Found ${result.tableChunks.length} tables`);
 
-// Work with chunks
+// Work with chunks — worker metadata is in chunk.metadata
 result.textChunks.forEach((chunk) => {
   console.log(chunk.content);
-  console.log(chunk.keywords);
-  console.log(chunk.summary);
+  console.log(chunk.metadata.keywords);
+  console.log(chunk.metadata.summary);
 });
 
 // Save results to disk
@@ -188,12 +188,17 @@ if (!documentId) {
 
 console.log(documentId);
 
+// Agentic mode (LLM navigation + answer synthesis)
 const response = await client.retrieval.query({
   namespace: 'support-center',
   query: 'How do I reset Bluetooth pairing?',
   topK: 5,
+  useAgentic: true,
 });
 
+console.log(response.answerText);         // LLM-generated answer
+console.log(response.referencedChunks);   // cited evidence chunks
+
 for (const result of response.results) {
   console.log(result.content);
   console.log(result.score);

package/dist/index.d.mts

@@ -332,7 +332,6 @@ interface Manifest {
     dataId?: string;
     /** Original source file name */
     sourceFileName: string;
-    /** Processing completion date */
     /** Processing completion date (optional: only present if emitted by the worker) */
     processingDate?: Date;
     /** Worker-side processing metadata emitted by manifest v2 */
@@ -341,6 +340,13 @@ interface Manifest {
     statistics: Statistics;
     /** Legacy file index from earlier ZIP manifests */
     files?: FileIndex;
+    /**
+     * Document hierarchy emitted by the current worker.
+     *
+     * The key remains all-caps at runtime because ``keysToCamel()`` only
+     * transforms snake_case keys.
+     */
+    HIERARCHY?: Record<string, unknown>;
 }
 /**
  * Chunk relationship entry (metadata.connect_to per schema v2.1)
@@ -357,6 +363,58 @@ interface ConnectTo {
     /** Shared keywords (related only) */
     keywords?: string[];
 }
+/**
+ * A single image or table resource entry in ``doc_nav.json``.
+ */
+interface DocNavResourceItem {
+    path: string;
+    summary?: string;
+}
+/**
+ * Image and table resource summaries from ``doc_nav.json``.
+ */
+interface DocNavResources {
+    images: DocNavResourceItem[];
+    tables: DocNavResourceItem[];
+}
+/**
+ * A document section in the ``doc_nav.json`` navigation tree.
+ */
+interface DocNavSection {
+    title: string;
+    path: string;
+    level: number;
+    summary?: string;
+    chunkCount: number;
+    children: DocNavSection[];
+}
+/**
+ * Top-level document navigation structure from ``doc_nav.json``.
+ */
+interface DocNav {
+    sections: DocNavSection[];
+    resources?: DocNavResources;
+}
+/**
+ * Known worker metadata fields for a chunk.
+ *
+ * All fields are optional.  Unknown fields added by future worker
+ * versions are accessible through the index signature.
+ */
+interface ChunkMetadata {
+    length?: number;
+    pageNums?: number[];
+    tokens?: string[];
+    keywords?: string[];
+    summary?: string;
+    connectTo?: ConnectTo[];
+    filePath?: string;
+    originalName?: string;
+    tableType?: string;
+    documentTopSummary?: string;
+    /** Allow forward-compatible access to unknown fields. */
+    [key: string]: unknown;
+}
 /**
  * Base chunk properties
  */
@@ -369,50 +427,30 @@ interface BaseChunk {
     content: string;
     /** Relative path in ZIP */
     path: string;
-    /** Page numbers spanned by this chunk when provided by the backend */
-    pageNums?: number[];
+    /** Worker metadata for this chunk */
+    metadata: ChunkMetadata;
 }
 /**
- * Minimal chunk representation emitted in chunks_slim.json
+ * Minimal chunk representation emitted in chunks_slim.json (legacy).
  */
 interface SlimChunk {
     type: 'text' | 'image' | 'table';
     path: string;
     content: string;
-    summary?: string;
 }
 /**
  * Text chunk
  */
 interface TextChunk extends BaseChunk {
     type: 'text';
-    /** Content length */
-    length: number;
-    /** Extracted tokens from the current backend payload */
-    tokens?: string[];
-    /** Extracted keywords */
-    keywords?: string[];
-    /** Generated summary */
-    summary?: string;
-    /** Chunk relationships (schema v2.1: metadata.connect_to) */
-    connectTo?: ConnectTo[];
-    /**
-     * @deprecated Use connectTo instead. Retained for backward compatibility.
-     * Previously populated from metadata.relationships which is no longer emitted by the API.
-     */
-    relationships?: string[];
 }
 /**
  * Image chunk
  */
 interface ImageChunk extends BaseChunk {
     type: 'image';
-    /** Content length */
-    length: number;
     /** Relative file path in ZIP */
     filePath: string;
-    /** Generated summary */
-    summary?: string;
     /** Image data buffer */
     data: Buffer;
     /** Image format (derived from file extension) */
@@ -425,14 +463,8 @@ interface ImageChunk extends BaseChunk {
  */
 interface TableChunk extends BaseChunk {
     type: 'table';
-    /** Content length */
-    length: number;
     /** Relative file path in ZIP */
     filePath: string;
-    /** Table type */
-    tableType?: string;
-    /** Generated summary */
-    summary?: string;
     /** HTML representation */
     html: string;
     /** Save table HTML to disk */
@@ -450,20 +482,22 @@ interface ParseResult {
     manifest: Manifest;
     /** All chunks */
     chunks: Chunk[];
-    /** Minimal chunk projection from chunks_slim.json (if available) */
-    chunksSlim?: SlimChunk[];
+    /** Document navigation tree from doc_nav.json (current worker output) */
+    docNav?: DocNav;
     /** Full document as Markdown (if available) */
     fullMarkdown?: string;
-    /** Document hierarchy (if available) */
+    /** Raw ZIP buffer */
+    rawZip: Buffer;
+    /** @deprecated Current worker no longer emits chunks_slim.json */
+    chunksSlim?: SlimChunk[];
+    /** @deprecated Current worker no longer emits hierarchy.json */
     hierarchy?: unknown;
-    /** Table-of-contents hierarchy hints (if available) */
+    /** @deprecated Table-of-contents hierarchy hints (if available) */
     tocHierarchies?: unknown;
-    /** Knowledge-base CSV export (if available) */
+    /** @deprecated Knowledge-base CSV export (if available) */
     kbCsv?: string;
-    /** Pre-rendered hierarchy HTML view (if available) */
+    /** @deprecated Pre-rendered hierarchy HTML view (if available) */
     hierarchyViewHtml?: string;
-    /** Raw ZIP buffer */
-    rawZip: Buffer;
     /** Text chunks only */
     readonly textChunks: TextChunk[];
     /** Image chunks only */
@@ -611,6 +645,14 @@ interface RetrievalQueryParams {
     namespace?: string;
     /** Maximum number of results to return */
     topK?: number;
+    /**
+     * Force retrieval mode.
+     *
+     * - ``true``  — agentic (LLM navigation + answer synthesis)
+     * - ``false`` — legacy 3-channel RRF only
+     * - ``undefined`` / omitted — server default
+     */
+    useAgentic?: boolean;
     /** Chunk type filter: 1=all, 2=text, 3=image, 4=table, 5=text+image, 6=text+table */
     dataType?: 1 | 2 | 3 | 4 | 5 | 6;
     /** Path keywords for include/exclude filtering */
@@ -668,6 +710,10 @@ interface RetrievalQueryResponse {
     query: string;
     /** Retrieval router path used by the API for this query */
     routerUsed?: string;
+    /** LLM-generated natural-language answer (agentic mode only) */
+    answerText?: string | null;
+    /** Cited evidence chunks with asset URLs (agentic mode only) */
+    referencedChunks?: Array<Record<string, unknown>> | null;
     /** Ranked retrieval results */
     results: RetrievalResult[];
 }

package/dist/index.d.ts

@@ -332,7 +332,6 @@ interface Manifest {
     dataId?: string;
     /** Original source file name */
     sourceFileName: string;
-    /** Processing completion date */
     /** Processing completion date (optional: only present if emitted by the worker) */
     processingDate?: Date;
     /** Worker-side processing metadata emitted by manifest v2 */
@@ -341,6 +340,13 @@ interface Manifest {
     statistics: Statistics;
     /** Legacy file index from earlier ZIP manifests */
     files?: FileIndex;
+    /**
+     * Document hierarchy emitted by the current worker.
+     *
+     * The key remains all-caps at runtime because ``keysToCamel()`` only
+     * transforms snake_case keys.
+     */
+    HIERARCHY?: Record<string, unknown>;
 }
 /**
  * Chunk relationship entry (metadata.connect_to per schema v2.1)
@@ -357,6 +363,58 @@ interface ConnectTo {
     /** Shared keywords (related only) */
     keywords?: string[];
 }
+/**
+ * A single image or table resource entry in ``doc_nav.json``.
+ */
+interface DocNavResourceItem {
+    path: string;
+    summary?: string;
+}
+/**
+ * Image and table resource summaries from ``doc_nav.json``.
+ */
+interface DocNavResources {
+    images: DocNavResourceItem[];
+    tables: DocNavResourceItem[];
+}
+/**
+ * A document section in the ``doc_nav.json`` navigation tree.
+ */
+interface DocNavSection {
+    title: string;
+    path: string;
+    level: number;
+    summary?: string;
+    chunkCount: number;
+    children: DocNavSection[];
+}
+/**
+ * Top-level document navigation structure from ``doc_nav.json``.
+ */
+interface DocNav {
+    sections: DocNavSection[];
+    resources?: DocNavResources;
+}
+/**
+ * Known worker metadata fields for a chunk.
+ *
+ * All fields are optional.  Unknown fields added by future worker
+ * versions are accessible through the index signature.
+ */
+interface ChunkMetadata {
+    length?: number;
+    pageNums?: number[];
+    tokens?: string[];
+    keywords?: string[];
+    summary?: string;
+    connectTo?: ConnectTo[];
+    filePath?: string;
+    originalName?: string;
+    tableType?: string;
+    documentTopSummary?: string;
+    /** Allow forward-compatible access to unknown fields. */
+    [key: string]: unknown;
+}
 /**
  * Base chunk properties
  */
@@ -369,50 +427,30 @@ interface BaseChunk {
     content: string;
     /** Relative path in ZIP */
     path: string;
-    /** Page numbers spanned by this chunk when provided by the backend */
-    pageNums?: number[];
+    /** Worker metadata for this chunk */
+    metadata: ChunkMetadata;
 }
 /**
- * Minimal chunk representation emitted in chunks_slim.json
+ * Minimal chunk representation emitted in chunks_slim.json (legacy).
  */
 interface SlimChunk {
     type: 'text' | 'image' | 'table';
     path: string;
     content: string;
-    summary?: string;
 }
 /**
  * Text chunk
  */
 interface TextChunk extends BaseChunk {
     type: 'text';
-    /** Content length */
-    length: number;
-    /** Extracted tokens from the current backend payload */
-    tokens?: string[];
-    /** Extracted keywords */
-    keywords?: string[];
-    /** Generated summary */
-    summary?: string;
-    /** Chunk relationships (schema v2.1: metadata.connect_to) */
-    connectTo?: ConnectTo[];
-    /**
-     * @deprecated Use connectTo instead. Retained for backward compatibility.
-     * Previously populated from metadata.relationships which is no longer emitted by the API.
-     */
-    relationships?: string[];
 }
 /**
  * Image chunk
  */
 interface ImageChunk extends BaseChunk {
     type: 'image';
-    /** Content length */
-    length: number;
     /** Relative file path in ZIP */
     filePath: string;
-    /** Generated summary */
-    summary?: string;
     /** Image data buffer */
     data: Buffer;
     /** Image format (derived from file extension) */
@@ -425,14 +463,8 @@ interface ImageChunk extends BaseChunk {
  */
 interface TableChunk extends BaseChunk {
     type: 'table';
-    /** Content length */
-    length: number;
     /** Relative file path in ZIP */
     filePath: string;
-    /** Table type */
-    tableType?: string;
-    /** Generated summary */
-    summary?: string;
     /** HTML representation */
     html: string;
     /** Save table HTML to disk */
@@ -450,20 +482,22 @@ interface ParseResult {
     manifest: Manifest;
     /** All chunks */
     chunks: Chunk[];
-    /** Minimal chunk projection from chunks_slim.json (if available) */
-    chunksSlim?: SlimChunk[];
+    /** Document navigation tree from doc_nav.json (current worker output) */
+    docNav?: DocNav;
     /** Full document as Markdown (if available) */
     fullMarkdown?: string;
-    /** Document hierarchy (if available) */
+    /** Raw ZIP buffer */
+    rawZip: Buffer;
+    /** @deprecated Current worker no longer emits chunks_slim.json */
+    chunksSlim?: SlimChunk[];
+    /** @deprecated Current worker no longer emits hierarchy.json */
     hierarchy?: unknown;
-    /** Table-of-contents hierarchy hints (if available) */
+    /** @deprecated Table-of-contents hierarchy hints (if available) */
     tocHierarchies?: unknown;
-    /** Knowledge-base CSV export (if available) */
+    /** @deprecated Knowledge-base CSV export (if available) */
     kbCsv?: string;
-    /** Pre-rendered hierarchy HTML view (if available) */
+    /** @deprecated Pre-rendered hierarchy HTML view (if available) */
     hierarchyViewHtml?: string;
-    /** Raw ZIP buffer */
-    rawZip: Buffer;
     /** Text chunks only */
     readonly textChunks: TextChunk[];
     /** Image chunks only */
@@ -611,6 +645,14 @@ interface RetrievalQueryParams {
     namespace?: string;
     /** Maximum number of results to return */
     topK?: number;
+    /**
+     * Force retrieval mode.
+     *
+     * - ``true``  — agentic (LLM navigation + answer synthesis)
+     * - ``false`` — legacy 3-channel RRF only
+     * - ``undefined`` / omitted — server default
+     */
+    useAgentic?: boolean;
     /** Chunk type filter: 1=all, 2=text, 3=image, 4=table, 5=text+image, 6=text+table */
     dataType?: 1 | 2 | 3 | 4 | 5 | 6;
     /** Path keywords for include/exclude filtering */
@@ -668,6 +710,10 @@ interface RetrievalQueryResponse {
     query: string;
     /** Retrieval router path used by the API for this query */
     routerUsed?: string;
+    /** LLM-generated natural-language answer (agentic mode only) */
+    answerText?: string | null;
+    /** Cited evidence chunks with asset URLs (agentic mode only) */
+    referencedChunks?: Array<Record<string, unknown>> | null;
     /** Ranked retrieval results */
     results: RetrievalResult[];
 }

package/dist/index.js

@@ -844,6 +844,13 @@ async function parseResult(httpClient, resultUrl, options) {
   if (fullMdFile) {
     fullMarkdown = await fullMdFile.async("string");
   }
+  let docNav;
+  const docNavFile = zip.file("doc_nav.json");
+  if (docNavFile) {
+    const docNavContent = await docNavFile.async("string");
+    const rawDocNav = JSON.parse(docNavContent);
+    docNav = keysToCamel(rawDocNav);
+  }
   let hierarchy;
   const hierarchyFile = zip.file("hierarchy.json");
   if (hierarchyFile) {
@@ -877,13 +884,15 @@ async function parseResult(httpClient, resultUrl, options) {
   const result = {
     manifest,
     chunks,
-    chunksSlim,
+    docNav,
     fullMarkdown,
+    rawZip: zipBuffer,
+    // Legacy
+    chunksSlim,
     hierarchy,
     tocHierarchies,
     kbCsv,
     hierarchyViewHtml,
-    rawZip: zipBuffer,
     get textChunks() {
       return chunks.filter((c) => c.type === "text");
     },
@@ -905,6 +914,9 @@ async function parseResult(httpClient, resultUrl, options) {
     async save(directory) {
       await import_fs2.promises.mkdir(directory, { recursive: true });
       await import_fs2.promises.writeFile((0, import_path.join)(directory, "manifest.json"), JSON.stringify(manifest, null, 2));
+      if (docNav) {
+        await import_fs2.promises.writeFile((0, import_path.join)(directory, "doc_nav.json"), JSON.stringify(docNav, null, 2));
+      }
       await import_fs2.promises.writeFile((0, import_path.join)(directory, "chunks.json"), JSON.stringify(chunks, null, 2));
       if (chunksSlim) {
         await import_fs2.promises.writeFile(
@@ -960,56 +972,24 @@ function extractSlimChunks(payload) {
   }
   return [];
 }
-function getChunkMetadata(chunkData) {
-  if (!chunkData.metadata) {
-    return {};
-  }
-  return chunkData.metadata;
-}
 function getChunkFilePath(chunkData) {
-  const metadata = getChunkMetadata(chunkData);
-  return chunkData.filePath ?? metadata.filePath ?? chunkData.path;
-}
-function normalizePageNums(pageNums) {
-  if (!Array.isArray(pageNums)) {
-    return void 0;
-  }
-  const normalized = pageNums.filter((pageNum) => typeof pageNum === "number");
-  return normalized.length > 0 ? normalized : void 0;
-}
-function normalizeTokens(tokens) {
-  if (!Array.isArray(tokens)) {
-    return void 0;
-  }
-  if (!tokens.every((token) => typeof token === "string")) {
-    return void 0;
-  }
-  return tokens;
+  const metadata = chunkData.metadata;
+  return chunkData.filePath ?? metadata?.filePath ?? chunkData.path;
 }
-function normalizeTextChunk(chunkData) {
-  const metadata = getChunkMetadata(chunkData);
-  const connectTo = metadata.connectTo ?? chunkData.connectTo;
-  const relationships = metadata.relationships ?? chunkData.relationships;
+function buildTextChunk(chunkData) {
   return {
     chunkId: chunkData.chunkId ?? "",
     type: "text",
     content: chunkData.content ?? "",
     path: chunkData.path ?? "",
-    pageNums: normalizePageNums(metadata.pageNums ?? chunkData.pageNums),
-    length: metadata.length ?? chunkData.length ?? 0,
-    tokens: normalizeTokens(metadata.tokens ?? chunkData.tokens),
-    keywords: metadata.keywords ?? chunkData.keywords,
-    summary: metadata.summary ?? chunkData.summary,
-    ...connectTo !== void 0 && { connectTo },
-    ...relationships !== void 0 && { relationships }
+    metadata: chunkData.metadata ?? {}
   };
 }
 async function processChunk(zip, chunkData) {
   if (chunkData.type === "text") {
-    return normalizeTextChunk(chunkData);
+    return buildTextChunk(chunkData);
   }
   if (chunkData.type === "image") {
-    const metadata = getChunkMetadata(chunkData);
     const filePath = getChunkFilePath(chunkData);
     if (!filePath) {
       throw new KnowhereError(`Image chunk missing file path: ${chunkData.chunkId ?? "unknown"}`);
@@ -1025,11 +1005,9 @@ async function processChunk(zip, chunkData) {
       type: "image",
       content: chunkData.content ?? "",
       path: chunkData.path ?? "",
-      pageNums: normalizePageNums(metadata.pageNums ?? chunkData.pageNums),
-      length: metadata.length ?? chunkData.length ?? 0,
       filePath,
-      summary: metadata.summary ?? chunkData.summary,
       data: imageBuffer,
+      metadata: chunkData.metadata ?? {},
       get format() {
         return getFileExtension(this.filePath);
       },
@@ -1044,7 +1022,6 @@ async function processChunk(zip, chunkData) {
     return enrichedChunk;
   }
   if (chunkData.type === "table") {
-    const metadata = getChunkMetadata(chunkData);
     const filePath = getChunkFilePath(chunkData);
     if (!filePath) {
       throw new KnowhereError(`Table chunk missing file path: ${chunkData.chunkId ?? "unknown"}`);
@@ -1060,12 +1037,9 @@ async function processChunk(zip, chunkData) {
       type: "table",
       content: chunkData.content ?? "",
       path: chunkData.path ?? "",
-      pageNums: normalizePageNums(metadata.pageNums ?? chunkData.pageNums),
-      length: metadata.length ?? chunkData.length ?? 0,
       filePath,
-      tableType: metadata.tableType ?? chunkData.tableType,
-      summary: metadata.summary ?? chunkData.summary,
       html,
+      metadata: chunkData.metadata ?? {},
       async save(directory) {
         const outputPath = (0, import_path.join)(directory, sanitizePath(this.filePath));
         const outputDir = (0, import_path.dirname)(outputPath);
@@ -1076,7 +1050,7 @@ async function processChunk(zip, chunkData) {
     };
     return enrichedChunk;
   }
-  return normalizeTextChunk(chunkData);
+  return buildTextChunk(chunkData);
 }
 
 // src/resources/jobs.ts

package/dist/index.mjs

@@ -784,6 +784,13 @@ async function parseResult(httpClient, resultUrl, options) {
   if (fullMdFile) {
     fullMarkdown = await fullMdFile.async("string");
   }
+  let docNav;
+  const docNavFile = zip.file("doc_nav.json");
+  if (docNavFile) {
+    const docNavContent = await docNavFile.async("string");
+    const rawDocNav = JSON.parse(docNavContent);
+    docNav = keysToCamel(rawDocNav);
+  }
   let hierarchy;
   const hierarchyFile = zip.file("hierarchy.json");
   if (hierarchyFile) {
@@ -817,13 +824,15 @@ async function parseResult(httpClient, resultUrl, options) {
   const result = {
     manifest,
     chunks,
-    chunksSlim,
+    docNav,
     fullMarkdown,
+    rawZip: zipBuffer,
+    // Legacy
+    chunksSlim,
     hierarchy,
     tocHierarchies,
     kbCsv,
     hierarchyViewHtml,
-    rawZip: zipBuffer,
     get textChunks() {
       return chunks.filter((c) => c.type === "text");
     },
@@ -845,6 +854,9 @@ async function parseResult(httpClient, resultUrl, options) {
     async save(directory) {
       await fs2.mkdir(directory, { recursive: true });
       await fs2.writeFile(join(directory, "manifest.json"), JSON.stringify(manifest, null, 2));
+      if (docNav) {
+        await fs2.writeFile(join(directory, "doc_nav.json"), JSON.stringify(docNav, null, 2));
+      }
       await fs2.writeFile(join(directory, "chunks.json"), JSON.stringify(chunks, null, 2));
       if (chunksSlim) {
         await fs2.writeFile(
@@ -900,56 +912,24 @@ function extractSlimChunks(payload) {
   }
   return [];
 }
-function getChunkMetadata(chunkData) {
-  if (!chunkData.metadata) {
-    return {};
-  }
-  return chunkData.metadata;
-}
 function getChunkFilePath(chunkData) {
-  const metadata = getChunkMetadata(chunkData);
-  return chunkData.filePath ?? metadata.filePath ?? chunkData.path;
-}
-function normalizePageNums(pageNums) {
-  if (!Array.isArray(pageNums)) {
-    return void 0;
-  }
-  const normalized = pageNums.filter((pageNum) => typeof pageNum === "number");
-  return normalized.length > 0 ? normalized : void 0;
-}
-function normalizeTokens(tokens) {
-  if (!Array.isArray(tokens)) {
-    return void 0;
-  }
-  if (!tokens.every((token) => typeof token === "string")) {
-    return void 0;
-  }
-  return tokens;
+  const metadata = chunkData.metadata;
+  return chunkData.filePath ?? metadata?.filePath ?? chunkData.path;
 }
-function normalizeTextChunk(chunkData) {
-  const metadata = getChunkMetadata(chunkData);
-  const connectTo = metadata.connectTo ?? chunkData.connectTo;
-  const relationships = metadata.relationships ?? chunkData.relationships;
+function buildTextChunk(chunkData) {
   return {
     chunkId: chunkData.chunkId ?? "",
     type: "text",
     content: chunkData.content ?? "",
     path: chunkData.path ?? "",
-    pageNums: normalizePageNums(metadata.pageNums ?? chunkData.pageNums),
-    length: metadata.length ?? chunkData.length ?? 0,
-    tokens: normalizeTokens(metadata.tokens ?? chunkData.tokens),
-    keywords: metadata.keywords ?? chunkData.keywords,
-    summary: metadata.summary ?? chunkData.summary,
-    ...connectTo !== void 0 && { connectTo },
-    ...relationships !== void 0 && { relationships }
+    metadata: chunkData.metadata ?? {}
   };
 }
 async function processChunk(zip, chunkData) {
   if (chunkData.type === "text") {
-    return normalizeTextChunk(chunkData);
+    return buildTextChunk(chunkData);
   }
   if (chunkData.type === "image") {
-    const metadata = getChunkMetadata(chunkData);
     const filePath = getChunkFilePath(chunkData);
     if (!filePath) {
       throw new KnowhereError(`Image chunk missing file path: ${chunkData.chunkId ?? "unknown"}`);
@@ -965,11 +945,9 @@ async function processChunk(zip, chunkData) {
       type: "image",
       content: chunkData.content ?? "",
       path: chunkData.path ?? "",
-      pageNums: normalizePageNums(metadata.pageNums ?? chunkData.pageNums),
-      length: metadata.length ?? chunkData.length ?? 0,
       filePath,
-      summary: metadata.summary ?? chunkData.summary,
       data: imageBuffer,
+      metadata: chunkData.metadata ?? {},
       get format() {
         return getFileExtension(this.filePath);
       },
@@ -984,7 +962,6 @@ async function processChunk(zip, chunkData) {
     return enrichedChunk;
   }
   if (chunkData.type === "table") {
-    const metadata = getChunkMetadata(chunkData);
     const filePath = getChunkFilePath(chunkData);
     if (!filePath) {
       throw new KnowhereError(`Table chunk missing file path: ${chunkData.chunkId ?? "unknown"}`);
@@ -1000,12 +977,9 @@ async function processChunk(zip, chunkData) {
       type: "table",
       content: chunkData.content ?? "",
       path: chunkData.path ?? "",
-      pageNums: normalizePageNums(metadata.pageNums ?? chunkData.pageNums),
-      length: metadata.length ?? chunkData.length ?? 0,
       filePath,
-      tableType: metadata.tableType ?? chunkData.tableType,
-      summary: metadata.summary ?? chunkData.summary,
       html,
+      metadata: chunkData.metadata ?? {},
       async save(directory) {
         const outputPath = join(directory, sanitizePath(this.filePath));
         const outputDir = dirname(outputPath);
@@ -1016,7 +990,7 @@ async function processChunk(zip, chunkData) {
     };
     return enrichedChunk;
   }
-  return normalizeTextChunk(chunkData);
+  return buildTextChunk(chunkData);
 }
 
 // src/resources/jobs.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ontos-ai/knowhere-sdk",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "description": "Official Node.js SDK for Knowhere document parsing API",
   "keywords": [
     "knowhere",