feat: add embedchain javascript package (#576)

embedchain-js/embedchain/chunkers/BaseChunker.ts

@@ -0,0 +1,44 @@
+import { createHash } from 'crypto';
+import type { RecursiveCharacterTextSplitter } from 'langchain/text_splitter';
+
+import type { BaseLoader } from '../loaders';
+import type { Input, LoaderResult } from '../models';
+import type { ChunkResult } from '../models/ChunkResult';
+
+class BaseChunker {
+  textSplitter: RecursiveCharacterTextSplitter;
+
+  constructor(textSplitter: RecursiveCharacterTextSplitter) {
+    this.textSplitter = textSplitter;
+  }
+
+  async createChunks(loader: BaseLoader, url: Input): Promise<ChunkResult> {
+    const documents: ChunkResult['documents'] = [];
+    const ids: ChunkResult['ids'] = [];
+    const datas: LoaderResult = await loader.loadData(url);
+    const metadatas: ChunkResult['metadatas'] = [];
+
+    const dataPromises = datas.map(async (data) => {
+      const { content, metaData } = data;
+      const chunks: string[] = await this.textSplitter.splitText(content);
+      chunks.forEach((chunk) => {
+        const chunkId = createHash('sha256')
+          .update(chunk + metaData.url)
+          .digest('hex');
+        ids.push(chunkId);
+        documents.push(chunk);
+        metadatas.push(metaData);
+      });
+    });
+
+    await Promise.all(dataPromises);
+
+    return {
+      documents,
+      ids,
+      metadatas,
+    };
+  }
+}
+
+export { BaseChunker };

embedchain-js/embedchain/chunkers/PdfFile.ts

@@ -0,0 +1,26 @@
+import { RecursiveCharacterTextSplitter } from 'langchain/text_splitter';
+
+import { BaseChunker } from './BaseChunker';
+
+interface TextSplitterChunkParams {
+  chunkSize: number;
+  chunkOverlap: number;
+  keepSeparator: boolean;
+}
+
+const TEXT_SPLITTER_CHUNK_PARAMS: TextSplitterChunkParams = {
+  chunkSize: 1000,
+  chunkOverlap: 0,
+  keepSeparator: false,
+};
+
+class PdfFileChunker extends BaseChunker {
+  constructor() {
+    const textSplitter = new RecursiveCharacterTextSplitter(
+      TEXT_SPLITTER_CHUNK_PARAMS
+    );
+    super(textSplitter);
+  }
+}
+
+export { PdfFileChunker };

embedchain-js/embedchain/chunkers/QnaPair.ts

@@ -0,0 +1,26 @@
+import { RecursiveCharacterTextSplitter } from 'langchain/text_splitter';
+
+import { BaseChunker } from './BaseChunker';
+
+interface TextSplitterChunkParams {
+  chunkSize: number;
+  chunkOverlap: number;
+  keepSeparator: boolean;
+}
+
+const TEXT_SPLITTER_CHUNK_PARAMS: TextSplitterChunkParams = {
+  chunkSize: 300,
+  chunkOverlap: 0,
+  keepSeparator: false,
+};
+
+class QnaPairChunker extends BaseChunker {
+  constructor() {
+    const textSplitter = new RecursiveCharacterTextSplitter(
+      TEXT_SPLITTER_CHUNK_PARAMS
+    );
+    super(textSplitter);
+  }
+}
+
+export { QnaPairChunker };

embedchain-js/embedchain/chunkers/WebPage.ts

@@ -0,0 +1,26 @@
+import { RecursiveCharacterTextSplitter } from 'langchain/text_splitter';
+
+import { BaseChunker } from './BaseChunker';
+
+interface TextSplitterChunkParams {
+  chunkSize: number;
+  chunkOverlap: number;
+  keepSeparator: boolean;
+}
+
+const TEXT_SPLITTER_CHUNK_PARAMS: TextSplitterChunkParams = {
+  chunkSize: 500,
+  chunkOverlap: 0,
+  keepSeparator: false,
+};
+
+class WebPageChunker extends BaseChunker {
+  constructor() {
+    const textSplitter = new RecursiveCharacterTextSplitter(
+      TEXT_SPLITTER_CHUNK_PARAMS
+    );
+    super(textSplitter);
+  }
+}
+
+export { WebPageChunker };

embedchain-js/embedchain/chunkers/index.ts

@@ -0,0 +1,6 @@
+import { BaseChunker } from './BaseChunker';
+import { PdfFileChunker } from './PdfFile';
+import { QnaPairChunker } from './QnaPair';
+import { WebPageChunker } from './WebPage';
+
+export { BaseChunker, PdfFileChunker, QnaPairChunker, WebPageChunker };