Merge pull request #808 from mendableai/feat/skipTlsVerification

feat: skipTlsVerification

apps/api/src/controllers/v1/crawl.ts

@@ -78,7 +78,7 @@ export async function crawlController(
   const crawler = crawlToCrawler(id, sc);
 
   try {
-    sc.robots = await crawler.getRobotsTxt();
+    sc.robots = await crawler.getRobotsTxt(pageOptions.skipTlsVerification);
   } catch (e) {
     Logger.debug(
       `[Crawl] Failed to get robots.txt (this is probably fine!): ${JSON.stringify(

apps/api/src/controllers/v1/types.ts

@@ -117,6 +117,7 @@ export const scrapeOptions = z.object({
       }
     ).transform(val => val ? val.toUpperCase() : 'US')
   }).optional(),
+  skipTlsVerification: z.boolean().default(false),
 }).strict(strictMessage)
 
 
@@ -433,6 +434,7 @@ export function legacyScrapeOptions(x: ScrapeOptions): PageOptions {
     parsePDF: x.parsePDF,
     actions: x.actions as Action[], // no strict null checking grrrr - mogery
     geolocation: x.geolocation,
+    skipTlsVerification: x.skipTlsVerification
   };
 }
 

apps/api/src/lib/entities.ts

@@ -54,6 +54,7 @@ export type PageOptions = {
   geolocation?: {
     country?: string;
   };
+  skipTlsVerification?: boolean;
 };
 
 export type ExtractorOptions = {

apps/api/src/scraper/WebScraper/crawler.ts

@@ -9,7 +9,7 @@ import robotsParser from "robots-parser";
 import { getURLDepth } from "./utils/maxDepthUtils";
 import { axiosTimeout } from "../../../src/lib/timeout";
 import { Logger } from "../../../src/lib/logger";
-
+import https from "https";
 export class WebCrawler {
   private jobId: string;
   private initialUrl: string;
@@ -145,8 +145,14 @@ export class WebCrawler {
       .slice(0, limit);
   }
 
-  public async getRobotsTxt(): Promise<string> {
-    const response = await axios.get(this.robotsTxtUrl, { timeout: axiosTimeout });
+  public async getRobotsTxt(skipTlsVerification = false): Promise<string> {
+    let extraArgs = {};
+    if(skipTlsVerification) {
+      extraArgs["httpsAgent"] = new https.Agent({
+        rejectUnauthorized: false
+      });
+    }
+    const response = await axios.get(this.robotsTxtUrl, { timeout: axiosTimeout, ...extraArgs });
     return response.data;
   }
 

apps/api/src/scraper/WebScraper/index.ts

@@ -594,6 +594,7 @@ export class WebScraperDataProvider {
       atsv: options.pageOptions?.atsv ?? false,
       actions: options.pageOptions?.actions ?? undefined,
       geolocation: options.pageOptions?.geolocation ?? undefined,
+      skipTlsVerification: options.pageOptions?.skipTlsVerification ?? false,
     };
     this.extractorOptions = options.extractorOptions ?? { mode: "markdown" };
     this.replaceAllPathsWithAbsolutePaths =

apps/api/src/scraper/WebScraper/scrapers/fireEngine.ts

@@ -28,7 +28,7 @@ export async function scrapWithFireEngine({
   waitFor = 0,
   screenshot = false,
   fullPageScreenshot = false,
-  pageOptions = { parsePDF: true, atsv: false, useFastMode: false, disableJsDom: false, geolocation: { country: "US" } },
+  pageOptions = { parsePDF: true, atsv: false, useFastMode: false, disableJsDom: false, geolocation: { country: "US" }, skipTlsVerification: false },
   fireEngineOptions = {},
   headers,
   options,
@@ -40,7 +40,7 @@ export async function scrapWithFireEngine({
   waitFor?: number;
   screenshot?: boolean;
   fullPageScreenshot?: boolean;
-  pageOptions?: { scrollXPaths?: string[]; parsePDF?: boolean, atsv?: boolean, useFastMode?: boolean, disableJsDom?: boolean, geolocation?: { country?: string } };
+  pageOptions?: { scrollXPaths?: string[]; parsePDF?: boolean, atsv?: boolean, useFastMode?: boolean, disableJsDom?: boolean, geolocation?: { country?: string }, skipTlsVerification?: boolean };
   fireEngineOptions?: FireEngineOptions;
   headers?: Record<string, string>;
   options?: any;
@@ -119,6 +119,7 @@ export async function scrapWithFireEngine({
           atsv: pageOptions?.atsv ?? false,
           scrollXPaths: pageOptions?.scrollXPaths ?? [],
           geolocation: pageOptions?.geolocation,
+          skipTlsVerification: pageOptions?.skipTlsVerification ?? false,
           actions: actions,
         },
         {

apps/api/src/scraper/WebScraper/single_url.ts

@@ -157,6 +157,7 @@ export async function scrapSingleUrl(
     atsv: pageOptions.atsv ?? false,
     actions: pageOptions.actions ?? undefined,
     geolocation: pageOptions.geolocation ?? undefined,
+    skipTlsVerification: pageOptions.skipTlsVerification ?? false,
   }
 
   if (extractorOptions) {