feat: use crawlee for docs service

Author: Patrick-Erichsen

core/core.ts

@@ -27,6 +27,10 @@ import type { IMessenger, Message } from "./util/messenger";
 import { editConfigJson } from "./util/paths";
 import { Telemetry } from "./util/posthog";
 import { streamDiffLines } from "./util/verticalEdit";
+import {
+  installChromium,
+  isChromiumInstalled,
+} from "./indexing/docs/installChromium";
 
 export class Core {
   // implements IMessenger<ToCoreProtocol, FromCoreProtocol>
@@ -167,6 +171,14 @@ export class Core {
       (..._) => Promise.resolve([]),
     );
 
+    try {
+      if (!isChromiumInstalled()) {
+        installChromium();
+      }
+    } catch (err) {
+      console.debug(`Failed to install Chromium: ${err}`);
+    }
+
     const on = this.messenger.on.bind(this.messenger);
 
     this.messenger.onError((err) => {

core/indexing/docs/DocsService.ts

@@ -22,7 +22,7 @@ import {
 import { Telemetry } from "../../util/posthog.js";
 import TransformersJsEmbeddingsProvider from "../embeddings/TransformersJsEmbeddingsProvider.js";
 import { Article, chunkArticle, pageToArticle } from "./article.js";
-import { crawlPage } from "./crawl.js";
+import { crawlSite } from "./crawlSite.js";
 import { runLanceMigrations, runSqliteMigrations } from "./migrations.js";
 import {
   downloadFromS3,
@@ -202,11 +202,36 @@ export default class DocsService {
     let processedPages = 0;
     let maxKnownPages = 1;
 
+    // crawlSite(
+    //   new URL(startUrl),
+    //   siteIndexingConfig.maxDepth,
+    //   async function* (page) {
+    //     processedPages++;
+
+    //     const article = pageToArticle(page);
+
+    //     if (article) {
+    //       articles.push(article);
+    //     }
+
+    //     // Use a heuristic approach for progress calculation
+    //     const progress = Math.min(processedPages / maxKnownPages, 1);
+
+    //     yield {
+    //       progress, // Yield the heuristic progress
+    //       desc: `Finding subpages (${page.path})`,
+    //       status: "indexing",
+    //     };
+
+    //     // Increase maxKnownPages to delay progress reaching 100% too soon
+    //     if (processedPages === maxKnownPages) {
+    //       maxKnownPages *= 2;
+    //     }
+    //   },
+    // );
+
     // Crawl pages and retrieve info as articles
-    for await (const page of crawlPage(
-      new URL(startUrl),
-      siteIndexingConfig.maxDepth,
-    )) {
+    for await (const page of crawlSite(startUrl)) {
       processedPages++;
 
       const article = pageToArticle(page);
@@ -521,8 +546,9 @@ export default class DocsService {
   private async getLanceTableNameFromEmbeddingsProvider(
     isPreIndexedDoc: boolean,
   ) {
-    const embeddingsProvider =
-      await this.getEmbeddingsProvider(isPreIndexedDoc);
+    const embeddingsProvider = await this.getEmbeddingsProvider(
+      isPreIndexedDoc,
+    );
     const embeddingsProviderId = this.removeInvalidLanceTableNameChars(
       embeddingsProvider.id,
     );

core/indexing/docs/article.ts

@@ -2,7 +2,7 @@ import { Readability } from "@mozilla/readability";
 import { JSDOM } from "jsdom";
 import { Chunk } from "../../index.js";
 import { cleanFragment, cleanHeader } from "../chunk/markdown.js";
-import { PageData } from "./crawl.js";
+import { PageData } from "./crawlSite.js";
 
 export type ArticleComponent = {
   title: string;
@@ -148,7 +148,7 @@ export function stringToArticle(
 
 export function pageToArticle(page: PageData): Article | undefined {
   try {
-    return stringToArticle(page.url, page.html, page.path);
+    return stringToArticle(page.url, page.content, page.path);
   } catch (err) {
     console.error("Error converting URL to article components", err);
     return undefined;