feat: overide generate hash function for image document (#751)

Co-authored-by: Marcus Schiesser <mail@marcusschiesser.de>

examples/multimodal/load.ts

@@ -4,7 +4,6 @@ import {
   VectorStoreIndex,
   storageContextFromDefaults,
 } from "llamaindex";
-import { DocStoreStrategy } from "llamaindex/ingestion/strategies/index";
 
 import * as path from "path";
 
@@ -32,7 +31,6 @@ async function generateDatasource() {
     });
     await VectorStoreIndex.fromDocuments(documents, {
       storageContext,
-      docStoreStrategy: DocStoreStrategy.NONE,
     });
   });
   console.log(`Storage successfully generated in ${ms / 1000}s.`);

packages/core/src/Node.ts

@@ -326,6 +326,37 @@ export class ImageNode<T extends Metadata = Metadata> extends TextNode<T> {
     const absPath = path.resolve(this.id_);
     return new URL(`file://${absPath}`);
   }
+
+  // Calculates the image part of the hash
+  private generateImageHash() {
+    const hashFunction = createSHA256();
+
+    if (this.image instanceof Blob) {
+      // TODO: ideally we should use the blob's content to calculate the hash:
+      // hashFunction.update(new Uint8Array(await this.image.arrayBuffer()));
+      // as this is async, we're using the node's ID for the time being
+      hashFunction.update(this.id_);
+    } else if (this.image instanceof URL) {
+      hashFunction.update(this.image.toString());
+    } else if (typeof this.image === "string") {
+      hashFunction.update(this.image);
+    } else {
+      throw new Error(
+        `Unknown image type: ${typeof this.image}. Can't calculate hash`,
+      );
+    }
+
+    return hashFunction.digest();
+  }
+
+  generateHash() {
+    const hashFunction = createSHA256();
+    // calculates hash based on hash of both components (image and text)
+    hashFunction.update(super.generateHash());
+    hashFunction.update(this.generateImageHash());
+
+    return hashFunction.digest();
+  }
 }
 
 export class ImageDocument<T extends Metadata = Metadata> extends ImageNode<T> {

packages/core/tests/Document.test.ts

-import { Document } from "llamaindex/Node";
+import { Document, ImageDocument } from "llamaindex/Node";
 import { describe, expect, test } from "vitest";
 
 describe("Document", () => {
@@ -6,4 +6,16 @@ describe("Document", () => {
     const doc = new Document({ text: "text", id_: "docId" });
     expect(doc).toBeDefined();
   });
+
+  test("should generate different hash for different image contents", () => {
+    const imageNode1 = new ImageDocument({
+      id_: "image",
+      image: "data:image/png;base64,sample_image_content1",
+    });
+    const imageNode2 = new ImageDocument({
+      id_: "image",
+      image: "data:image/png;base64,sample_image_content2",
+    });
+    expect(imageNode1.hash).not.toBe(imageNode2.hash);
+  });
 });