@workglow/ai 0.0.113 → 0.0.115

package/README.md

@@ -425,8 +425,8 @@ The AI package provides a comprehensive set of tasks for building RAG pipelines.
 
 | Task                    | Description                              |
 | ----------------------- | ---------------------------------------- |
-| `ChunkToVectorTask`     | Transforms chunks to vector store format |
-| `ChunkVectorUpsertTask` | Stores vectors in a repository           |
+| `ChunkToVectorTask`     | Transforms chunks to vector store format (input: `vector` + `chunks`, output: `vectors`) |
+| `ChunkVectorUpsertTask` | Stores vectors in a KnowledgeBase (input: `knowledgeBase` + `vectors`) |
 | `ChunkVectorSearchTask` | Searches vectors by similarity           |
 | `VectorQuantizeTask`    | Quantizes vectors for storage efficiency |
 
@@ -445,10 +445,13 @@ The AI package provides a comprehensive set of tasks for building RAG pipelines.
 
 ```typescript
 import { Workflow } from "@workglow/task-graph";
-import { InMemoryVectorRepository } from "@workglow/storage";
+import { createKnowledgeBase } from "@workglow/knowledge-base";
 
-const vectorRepo = new InMemoryVectorRepository();
-await vectorRepo.setupDatabase();
+// Create a KnowledgeBase (auto-registers globally as "my-kb")
+const kb = await createKnowledgeBase({
+  name: "my-kb",
+  vectorDimensions: 384, // must match your embedding model
+});
 
 // Document ingestion - fully chainable, no loops required
 await new Workflow()
@@ -470,25 +473,18 @@ await new Workflow()
     model: "Xenova/all-MiniLM-L6-v2",
   })
   .chunkToVector()
-  .vectorStoreUpsert({
-    repository: vectorRepo,
+  .chunkVectorUpsert({
+    knowledgeBase: "my-kb",
   })
   .run();
 
-// Query pipeline
-const answer = await new Workflow()
-  .queryExpander({
+// Query pipeline — ChunkRetrievalTask handles embedding + vector search end-to-end
+const result = await new Workflow()
+  .chunkRetrieval({
+    knowledgeBase: "my-kb",
     query: "What is transfer learning?",
-    method: "multi-query",
-    numVariations: 3,
-  })
-  .textEmbedding({
     model: "Xenova/all-MiniLM-L6-v2",
-  })
-  .vectorStoreSearch({
-    repository: vectorRepo,
     topK: 10,
-    scoreThreshold: 0.5,
   })
   .reranker({
     query: "What is transfer learning?",
@@ -537,7 +533,7 @@ Each task passes through what the next task needs:
 | `hierarchicalChunker` | `doc_id`                 | `chunks`, `text[]`, `count`           |
 | `textEmbedding`       | (implicit)               | `vector[]`                            |
 | `chunkToVector`       | -                        | `ids[]`, `vectors[]`, `metadata[]`    |
-| `vectorStoreUpsert`   | -                        | `count`, `ids`                        |
+| `chunkVectorUpsert`   | -                        | `count`, `ids`                        |
 
 This design eliminates the need for external loops - the entire pipeline chains together naturally.
 

package/dist/browser.js

@@ -779,7 +779,7 @@ var backgroundRemoval = (input, config) => {
 Workflow.prototype.backgroundRemoval = CreateWorkflow(BackgroundRemovalTask);
 
 // src/task/ChunkRetrievalTask.ts
-import { TypeKnowledgeBase } from "@workglow/dataset";
+import { TypeKnowledgeBase } from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow3,
   Task,
@@ -1031,7 +1031,7 @@ var chunkRetrieval = (input, config) => {
 Workflow3.prototype.chunkRetrieval = CreateWorkflow3(ChunkRetrievalTask);
 
 // src/task/ChunkToVectorTask.ts
-import { ChunkRecordSchema } from "@workglow/dataset";
+import { ChunkRecordSchema } from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow4,
   Task as Task2,
@@ -1165,7 +1165,7 @@ var chunkToVector = (input, config) => {
 Workflow4.prototype.chunkToVector = CreateWorkflow4(ChunkToVectorTask);
 
 // src/task/ChunkVectorHybridSearchTask.ts
-import { TypeKnowledgeBase as TypeKnowledgeBase2 } from "@workglow/dataset";
+import { TypeKnowledgeBase as TypeKnowledgeBase2 } from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow5,
   Task as Task3,
@@ -1339,7 +1339,7 @@ var hybridSearch = async (input, config) => {
 Workflow5.prototype.hybridSearch = CreateWorkflow5(ChunkVectorHybridSearchTask);
 
 // src/task/ChunkVectorSearchTask.ts
-import { TypeKnowledgeBase as TypeKnowledgeBase3 } from "@workglow/dataset";
+import { TypeKnowledgeBase as TypeKnowledgeBase3 } from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow6,
   Task as Task4,
@@ -1439,7 +1439,7 @@ class ChunkVectorSearchTask extends Task4 {
   static outputSchema() {
     return outputSchema4;
   }
-  async execute(input, context) {
+  async execute(input, _context) {
     const { knowledgeBase, query, topK = 10, filter, scoreThreshold = 0 } = input;
     const kb = knowledgeBase;
     const results = await kb.similaritySearch(query, {
@@ -1462,7 +1462,7 @@ var vectorStoreSearch = (input, config) => {
 Workflow6.prototype.vectorStoreSearch = CreateWorkflow6(ChunkVectorSearchTask);
 
 // src/task/ChunkVectorUpsertTask.ts
-import { TypeKnowledgeBase as TypeKnowledgeBase4 } from "@workglow/dataset";
+import { TypeKnowledgeBase as TypeKnowledgeBase4 } from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow7,
   Task as Task5,
@@ -1584,7 +1584,7 @@ var chunkVectorUpsert = (input, config) => {
 Workflow7.prototype.chunkVectorUpsert = CreateWorkflow7(ChunkVectorUpsertTask);
 
 // src/task/ContextBuilderTask.ts
-import { estimateTokens } from "@workglow/dataset";
+import { estimateTokens } from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow9,
   Task as Task6,
@@ -1942,7 +1942,7 @@ Workflow9.prototype.contextBuilder = CreateWorkflow9(ContextBuilderTask);
 import {
   getChildren,
   hasChildren
-} from "@workglow/dataset";
+} from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow12,
   Task as Task7,
@@ -3009,7 +3009,7 @@ import {
   estimateTokens as estimateTokens2,
   getChildren as getChildren2,
   hasChildren as hasChildren2
-} from "@workglow/dataset";
+} from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow18,
   Task as Task8,
@@ -3245,7 +3245,7 @@ Workflow18.prototype.hierarchicalChunker = CreateWorkflow18(HierarchicalChunkerT
 import {
   ChunkRecordArraySchema,
   TypeKnowledgeBase as TypeKnowledgeBase5
-} from "@workglow/dataset";
+} from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow19,
   Task as Task9,
@@ -4506,7 +4506,7 @@ var reranker = (input, config) => {
 Workflow29.prototype.reranker = CreateWorkflow29(RerankerTask);
 
 // src/task/StructuralParserTask.ts
-import { StructuralParser } from "@workglow/dataset";
+import { StructuralParser } from "@workglow/knowledge-base";
 import {
   CreateWorkflow as CreateWorkflow30,
   Task as Task12,
@@ -6348,4 +6348,4 @@ export {
   AiJob
 };
 
-//# debugId=0E9DD67701C6DF9264756E2164756E21
+//# debugId=168B101D45A05FA664756E2164756E21