@workglow/ai 0.0.111 → 0.0.113

package/dist/bun.js

@@ -780,7 +780,7 @@ var backgroundRemoval = (input, config) => {
 Workflow.prototype.backgroundRemoval = CreateWorkflow(BackgroundRemovalTask);
 
 // src/task/ChunkRetrievalTask.ts
-import { TypeDocumentChunkDataset } from "@workglow/dataset";
+import { TypeKnowledgeBase } from "@workglow/dataset";
 import {
   CreateWorkflow as CreateWorkflow3,
   Task,
@@ -843,9 +843,9 @@ Workflow2.prototype.textEmbedding = CreateWorkflow2(TextEmbeddingTask);
 var inputSchema = {
   type: "object",
   properties: {
-    dataset: TypeDocumentChunkDataset({
-      title: "Document Chunk Vector Repository",
-      description: "The document chunk vector repository instance to search in"
+    knowledgeBase: TypeKnowledgeBase({
+      title: "Knowledge Base",
+      description: "The knowledge base instance to search in"
     }),
     query: TypeSingleOrArray({
       oneOf: [
@@ -889,14 +889,14 @@ var inputSchema = {
       default: false
     }
   },
-  required: ["dataset", "query"],
+  required: ["knowledgeBase", "query"],
   if: {
     properties: {
       query: { type: "string" }
     }
   },
   then: {
-    required: ["dataset", "query", "model"]
+    required: ["knowledgeBase", "query", "model"]
   },
   else: {},
   additionalProperties: false
@@ -945,9 +945,20 @@ var outputSchema = {
       type: "number",
       title: "Count",
       description: "Number of results returned"
-    }
+    },
+    query: TypeSingleOrArray({
+      oneOf: [
+        { type: "string" },
+        TypedArraySchema2({
+          title: "Query Vector",
+          description: "Pre-computed query vector"
+        })
+      ],
+      title: "Query",
+      description: "The query used for retrieval (pass-through)"
+    })
   },
-  required: ["chunks", "chunk_ids", "metadata", "scores", "count"],
+  required: ["chunks", "chunk_ids", "metadata", "scores", "count", "query"],
   additionalProperties: false
 };
 
@@ -965,7 +976,7 @@ class ChunkRetrievalTask extends Task {
   }
   async execute(input, context) {
     const {
-      dataset,
+      knowledgeBase,
       query,
       topK = 5,
       filter,
@@ -973,7 +984,7 @@ class ChunkRetrievalTask extends Task {
       scoreThreshold = 0,
       returnVectors = false
     } = input;
-    const repo = dataset;
+    const kb = knowledgeBase;
     let queryVectors;
     if (typeof query === "string" || Array.isArray(query) && query.every((q) => typeof q === "string")) {
       if (!model) {
@@ -990,7 +1001,7 @@ class ChunkRetrievalTask extends Task {
     const searchVectors = queryVectors.map((v) => v instanceof Float32Array ? v : new Float32Array(v));
     const results = [];
     for (const searchVector of searchVectors) {
-      const res = await repo.similaritySearch(searchVector, {
+      const res = await kb.similaritySearch(searchVector, {
         topK,
         filter,
         scoreThreshold
@@ -999,14 +1010,15 @@ class ChunkRetrievalTask extends Task {
     }
     const chunks = results.map((r) => {
       const meta = r.metadata;
-      return meta.text || meta.content || meta.chunk || JSON.stringify(meta);
+      return meta.text || JSON.stringify(meta);
     });
     const output = {
       chunks,
       chunk_ids: results.map((r) => r.chunk_id),
       metadata: results.map((r) => r.metadata),
       scores: results.map((r) => r.score),
-      count: results.length
+      count: results.length,
+      query
     };
     if (returnVectors) {
       output.vectors = results.map((r) => r.vector);
@@ -1020,7 +1032,7 @@ var chunkRetrieval = (input, config) => {
 Workflow3.prototype.chunkRetrieval = CreateWorkflow3(ChunkRetrievalTask);
 
 // src/task/ChunkToVectorTask.ts
-import { ChunkNodeSchema } from "@workglow/dataset";
+import { ChunkRecordSchema } from "@workglow/dataset";
 import {
   CreateWorkflow as CreateWorkflow4,
   Task as Task2,
@@ -1044,11 +1056,11 @@ var inputSchema2 = {
     },
     chunks: {
       type: "array",
-      items: ChunkNodeSchema(),
+      items: ChunkRecordSchema(),
       title: "Chunks",
-      description: "Array of chunk nodes"
+      description: "Array of chunk records"
     },
-    vectors: {
+    vector: {
       type: "array",
       items: TypedArraySchema3({
         title: "Vector",
@@ -1058,7 +1070,7 @@ var inputSchema2 = {
       description: "Embeddings from TextEmbeddingTask"
     }
   },
-  required: ["chunks", "vectors"],
+  required: ["chunks", "vector"],
   additionalProperties: false
 };
 var outputSchema2 = {
@@ -1113,13 +1125,13 @@ class ChunkToVectorTask extends Task2 {
     return outputSchema2;
   }
   async execute(input, context) {
-    const { chunks, vectors, doc_title } = input;
+    const { chunks, vector, doc_title } = input;
     const chunkArray = chunks;
-    if (!chunkArray || !vectors) {
+    if (!chunkArray || !vector) {
       throw new Error("Both chunks and vector are required");
     }
-    if (chunkArray.length !== vectors.length) {
-      throw new Error(`Mismatch: ${chunkArray.length} chunks but ${vectors.length} vectors`);
+    if (chunkArray.length !== vector.length) {
+      throw new Error(`Mismatch: ${chunkArray.length} chunks but ${vector.length} vectors`);
     }
     const ids = [];
     const metadata = [];
@@ -1136,12 +1148,13 @@ class ChunkToVectorTask extends Task2 {
         text: chunk.text,
         nodePath: chunk.nodePath,
         ...doc_title ? { doc_title } : {},
-        ...chunk.enrichment || {}
+        ...chunk.summary ? { summary: chunk.summary } : {},
+        ...chunk.entities ? { entities: chunk.entities } : {}
       });
     }
     return {
       ids,
-      vectors,
+      vectors: vector,
       metadata,
       texts
     };
@@ -1153,7 +1166,7 @@ var chunkToVector = (input, config) => {
 Workflow4.prototype.chunkToVector = CreateWorkflow4(ChunkToVectorTask);
 
 // src/task/ChunkVectorHybridSearchTask.ts
-import { TypeDocumentChunkDataset as TypeDocumentChunkDataset2 } from "@workglow/dataset";
+import { TypeKnowledgeBase as TypeKnowledgeBase2 } from "@workglow/dataset";
 import {
   CreateWorkflow as CreateWorkflow5,
   Task as Task3,
@@ -1165,9 +1178,9 @@ import {
 var inputSchema3 = {
   type: "object",
   properties: {
-    dataset: TypeDocumentChunkDataset2({
-      title: "Document Chunk Vector Repository",
-      description: "The document chunk vector repository instance to search in (must support hybridSearch)"
+    knowledgeBase: TypeKnowledgeBase2({
+      title: "Knowledge Base",
+      description: "The knowledge base instance to search in (must support hybridSearch)"
     }),
     queryVector: TypedArraySchema4({
       title: "Query Vector",
@@ -1213,7 +1226,7 @@ var inputSchema3 = {
       default: false
     }
   },
-  required: ["dataset", "queryVector", "queryText"],
+  required: ["knowledgeBase", "queryVector", "queryText"],
   additionalProperties: false
 };
 var outputSchema3 = {
@@ -1225,10 +1238,10 @@ var outputSchema3 = {
       title: "Text Chunks",
       description: "Retrieved text chunks"
     },
-    ids: {
+    chunk_ids: {
       type: "array",
       items: { type: "string" },
-      title: "IDs",
+      title: "Chunk IDs",
       description: "IDs of retrieved chunks"
     },
     metadata: {
@@ -1260,9 +1273,14 @@ var outputSchema3 = {
       type: "number",
       title: "Count",
       description: "Number of results returned"
+    },
+    query: {
+      type: "string",
+      title: "Query",
+      description: "The text query used for search (pass-through)"
     }
   },
-  required: ["chunks", "ids", "metadata", "scores", "count"],
+  required: ["chunks", "chunk_ids", "metadata", "scores", "count", "query"],
   additionalProperties: false
 };
 
@@ -1280,7 +1298,7 @@ class ChunkVectorHybridSearchTask extends Task3 {
   }
   async execute(input, context) {
     const {
-      dataset,
+      knowledgeBase,
       queryVector,
       queryText,
       topK = 10,
@@ -1289,12 +1307,9 @@ class ChunkVectorHybridSearchTask extends Task3 {
       vectorWeight = 0.7,
       returnVectors = false
     } = input;
-    const repo = dataset;
-    if (!repo.hybridSearch) {
-      throw new Error("Dataset does not support hybrid search.");
-    }
+    const kb = knowledgeBase;
     const searchVector = queryVector instanceof Float32Array ? queryVector : new Float32Array(queryVector);
-    const results = await repo.hybridSearch(searchVector, {
+    const results = await kb.hybridSearch(searchVector, {
       textQuery: queryText,
       topK,
       filter,
@@ -1303,14 +1318,15 @@ class ChunkVectorHybridSearchTask extends Task3 {
     });
     const chunks = results.map((r) => {
       const meta = r.metadata;
-      return meta.text || meta.content || meta.chunk || JSON.stringify(meta);
+      return meta.text || JSON.stringify(meta);
     });
     const output = {
       chunks,
-      ids: results.map((r) => r.chunk_id),
+      chunk_ids: results.map((r) => r.chunk_id),
       metadata: results.map((r) => r.metadata),
       scores: results.map((r) => r.score),
-      count: results.length
+      count: results.length,
+      query: queryText
     };
     if (returnVectors) {
       output.vectors = results.map((r) => r.vector);
@@ -1324,7 +1340,7 @@ var hybridSearch = async (input, config) => {
 Workflow5.prototype.hybridSearch = CreateWorkflow5(ChunkVectorHybridSearchTask);
 
 // src/task/ChunkVectorSearchTask.ts
-import { TypeDocumentChunkDataset as TypeDocumentChunkDataset3 } from "@workglow/dataset";
+import { TypeKnowledgeBase as TypeKnowledgeBase3 } from "@workglow/dataset";
 import {
   CreateWorkflow as CreateWorkflow6,
   Task as Task4,
@@ -1336,9 +1352,9 @@ import {
 var inputSchema4 = {
   type: "object",
   properties: {
-    dataset: TypeDocumentChunkDataset3({
-      title: "Vector Repository",
-      description: "The vector repository instance to search in"
+    knowledgeBase: TypeKnowledgeBase3({
+      title: "Knowledge Base",
+      description: "The knowledge base instance to search in"
     }),
     query: TypedArraySchema5({
       title: "Query Vector",
@@ -1365,7 +1381,7 @@ var inputSchema4 = {
       default: 0
     }
   },
-  required: ["dataset", "query"],
+  required: ["knowledgeBase", "query"],
   additionalProperties: false
 };
 var outputSchema4 = {
@@ -1416,7 +1432,7 @@ class ChunkVectorSearchTask extends Task4 {
   static type = "ChunkVectorSearchTask";
   static category = "Vector Store";
   static title = "Vector Store Search";
-  static description = "Search for similar vectors in a document chunk dataset";
+  static description = "Search for similar vectors in a knowledge base";
   static cacheable = true;
   static inputSchema() {
     return inputSchema4;
@@ -1425,9 +1441,9 @@ class ChunkVectorSearchTask extends Task4 {
     return outputSchema4;
   }
   async execute(input, context) {
-    const { dataset, query, topK = 10, filter, scoreThreshold = 0 } = input;
-    const repo = dataset;
-    const results = await repo.similaritySearch(query, {
+    const { knowledgeBase, query, topK = 10, filter, scoreThreshold = 0 } = input;
+    const kb = knowledgeBase;
+    const results = await kb.similaritySearch(query, {
       topK,
       filter,
       scoreThreshold
@@ -1447,7 +1463,7 @@ var vectorStoreSearch = (input, config) => {
 Workflow6.prototype.vectorStoreSearch = CreateWorkflow6(ChunkVectorSearchTask);
 
 // src/task/ChunkVectorUpsertTask.ts
-import { TypeDocumentChunkDataset as TypeDocumentChunkDataset4 } from "@workglow/dataset";
+import { TypeKnowledgeBase as TypeKnowledgeBase4 } from "@workglow/dataset";
 import {
   CreateWorkflow as CreateWorkflow7,
   Task as Task5,
@@ -1459,9 +1475,9 @@ import {
 var inputSchema5 = {
   type: "object",
   properties: {
-    dataset: TypeDocumentChunkDataset4({
-      title: "Document Chunk Vector Repository",
-      description: "The document chunk vector repository instance to store vectors in"
+    knowledgeBase: TypeKnowledgeBase4({
+      title: "Knowledge Base",
+      description: "The knowledge base instance to store vectors in"
     }),
     doc_id: {
       type: "string",
@@ -1479,7 +1495,7 @@ var inputSchema5 = {
       additionalProperties: true
     })
   },
-  required: ["dataset", "doc_id", "vectors", "metadata"],
+  required: ["knowledgeBase", "doc_id", "vectors", "metadata"],
   additionalProperties: false
 };
 var outputSchema5 = {
@@ -1510,7 +1526,7 @@ class ChunkVectorUpsertTask extends Task5 {
   static type = "ChunkVectorUpsertTask";
   static category = "Vector Store";
   static title = "Add to Vector Store";
-  static description = "Store vector embeddings with metadata in a document chunk dataset";
+  static description = "Store vector embeddings with metadata in a knowledge base";
   static cacheable = false;
   static inputSchema() {
     return inputSchema5;
@@ -1519,10 +1535,10 @@ class ChunkVectorUpsertTask extends Task5 {
     return outputSchema5;
   }
   async execute(input, context) {
-    const { dataset, doc_id, vectors, metadata } = input;
+    const { knowledgeBase, doc_id, vectors, metadata } = input;
     const vectorArray = Array.isArray(vectors) ? vectors : [vectors];
     const metadataArray = Array.isArray(metadata) ? metadata : Array(vectorArray.length).fill(metadata);
-    const repo = dataset;
+    const kb = knowledgeBase;
     await context.updateProgress(1, "Upserting vectors");
     if (vectorArray.length > 1) {
       if (vectorArray.length !== metadataArray.length) {
@@ -1536,7 +1552,7 @@ class ChunkVectorUpsertTask extends Task5 {
           metadata: metadataItem
         };
       });
-      const results = await repo.putBulk(entities);
+      const results = await kb.upsertChunksBulk(entities);
       const chunk_ids = results.map((r) => r.chunk_id);
       return {
         doc_id,
@@ -1545,7 +1561,7 @@ class ChunkVectorUpsertTask extends Task5 {
       };
     } else if (vectorArray.length === 1) {
       const metadataItem = metadataArray[0];
-      const result = await repo.put({
+      const result = await kb.upsertChunk({
         doc_id,
         vector: vectorArray[0],
         metadata: metadataItem
@@ -2990,7 +3006,7 @@ Workflow17.prototype.handLandmarker = CreateWorkflow17(HandLandmarkerTask);
 
 // src/task/HierarchicalChunkerTask.ts
 import {
-  ChunkNodeSchema as ChunkNodeSchema2,
+  ChunkRecordSchema as ChunkRecordSchema2,
   estimateTokens as estimateTokens2,
   getChildren as getChildren2,
   hasChildren as hasChildren2
@@ -3060,9 +3076,9 @@ var outputSchema8 = {
     },
     chunks: {
       type: "array",
-      items: ChunkNodeSchema2(),
+      items: ChunkRecordSchema2(),
       title: "Chunks",
-      description: "Array of chunk nodes"
+      description: "Array of chunk records"
     },
     text: {
       type: "array",
@@ -3228,9 +3244,8 @@ Workflow18.prototype.hierarchicalChunker = CreateWorkflow18(HierarchicalChunkerT
 
 // src/task/HierarchyJoinTask.ts
 import {
-  ChunkMetadataArraySchema,
-  EnrichedChunkMetadataArraySchema,
-  TypeDocumentDataset
+  ChunkRecordArraySchema,
+  TypeKnowledgeBase as TypeKnowledgeBase5
 } from "@workglow/dataset";
 import {
   CreateWorkflow as CreateWorkflow19,
@@ -3240,9 +3255,9 @@ import {
 var inputSchema9 = {
   type: "object",
   properties: {
-    documents: TypeDocumentDataset({
-      title: "Documents",
-      description: "The documents dataset to query for hierarchy"
+    knowledgeBase: TypeKnowledgeBase5({
+      title: "Knowledge Base",
+      description: "The knowledge base to query for hierarchy"
     }),
     chunks: {
       type: "array",
@@ -3256,7 +3271,7 @@ var inputSchema9 = {
       title: "Chunk IDs",
       description: "IDs of retrieved chunks"
     },
-    metadata: ChunkMetadataArraySchema,
+    metadata: ChunkRecordArraySchema,
     scores: {
       type: "array",
       items: { type: "number" },
@@ -3276,7 +3291,7 @@ var inputSchema9 = {
       default: true
     }
   },
-  required: ["documents", "chunks", "chunk_ids", "metadata", "scores"],
+  required: ["knowledgeBase", "chunks", "chunk_ids", "metadata", "scores"],
   additionalProperties: false
 };
 var outputSchema9 = {
@@ -3294,7 +3309,7 @@ var outputSchema9 = {
       title: "Chunk IDs",
       description: "IDs of retrieved chunks"
     },
-    metadata: EnrichedChunkMetadataArraySchema,
+    metadata: ChunkRecordArraySchema,
     scores: {
       type: "array",
       items: { type: "number" },
@@ -3325,7 +3340,7 @@ class HierarchyJoinTask extends Task9 {
   }
   async execute(input, context) {
     const {
-      documents,
+      knowledgeBase,
       chunks,
       chunk_ids,
       metadata,
@@ -3333,7 +3348,7 @@ class HierarchyJoinTask extends Task9 {
       includeParentSummaries = true,
       includeEntities = true
     } = input;
-    const repo = documents;
+    const kb = knowledgeBase;
     const enrichedMetadata = [];
     for (let i = 0;i < chunk_ids.length; i++) {
       const chunkId = chunk_ids[i];
@@ -3349,7 +3364,7 @@ class HierarchyJoinTask extends Task9 {
         continue;
       }
       try {
-        const ancestors = await repo.getAncestors(doc_id, leafNodeId);
+        const ancestors = await kb.getAncestors(doc_id, leafNodeId);
         const enriched = { ...originalMetadata };
         if (includeParentSummaries && ancestors.length > 0) {
           const parentSummaries = [];
@@ -3358,7 +3373,7 @@ class HierarchyJoinTask extends Task9 {
             if (ancestor.enrichment?.summary) {
               parentSummaries.push(ancestor.enrichment.summary);
             }
-            if (ancestor.kind === "section" && ancestor.title) {
+            if (ancestor.kind === "section" && "title" in ancestor) {
               sectionTitles.push(ancestor.title);
             }
           }
@@ -4009,7 +4024,7 @@ var inputSchema10 = {
 var outputSchema10 = {
   type: "object",
   properties: {
-    queries: {
+    query: {
       type: "array",
       items: { type: "string" },
       title: "Expanded Queries",
@@ -4031,7 +4046,7 @@ var outputSchema10 = {
       description: "Number of queries generated"
     }
   },
-  required: ["queries", "originalQuery", "method", "count"],
+  required: ["query", "originalQuery", "method", "count"],
   additionalProperties: false
 };
 
@@ -4069,7 +4084,7 @@ class QueryExpanderTask extends Task10 {
       queries.unshift(query);
     }
     return {
-      queries,
+      query: queries,
       originalQuery: query,
       method,
       count: queries.length
@@ -6334,4 +6349,4 @@ export {
   AiJob
 };
 
-//# debugId=0AADE5DC48AE458464756E2164756E21
+//# debugId=8E8187274333361364756E2164756E21