Merge pull request #878 from mendableai/mog/deduplicate-urls

feat(crawl): Similar URL deduplication

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

@@ -203,6 +203,7 @@ const crawlerOptions = z.object({
   allowBackwardLinks: z.boolean().default(false), // >> TODO: CHANGE THIS NAME???
   allowExternalLinks: z.boolean().default(false),
   ignoreSitemap: z.boolean().default(true),
+  deduplicateSimilarURLs: z.boolean().default(true),
 }).strict(strictMessage);
 
 // export type CrawlerOptions = {
@@ -457,6 +458,7 @@ export function toLegacyCrawlerOptions(x: CrawlerOptions) {
     allowBackwardCrawling: x.allowBackwardLinks,
     allowExternalContentLinks: x.allowExternalLinks,
     ignoreSitemap: x.ignoreSitemap,
+    deduplicateSimilarURLs: x.deduplicateSimilarURLs,
   };
 }
 
@@ -470,7 +472,7 @@ export function fromLegacyCrawlerOptions(x: any): { crawlOptions: CrawlerOptions
       allowBackwardLinks: x.allowBackwardCrawling,
       allowExternalLinks: x.allowExternalContentLinks,
       ignoreSitemap: x.ignoreSitemap,
-      // TODO: returnOnlyUrls support
+      deduplicateSimilarURLs: x.deduplicateSimilarURLs,
     }),
     internalOptions: {
       v0CrawlOnlyUrls: x.returnOnlyUrls,

apps/api/src/lib/crawl-redis.test.ts

@@ -0,0 +1,33 @@
+import { generateURLPermutations } from "./crawl-redis";
+
+describe("generateURLPermutations", () => {
+    it("generates permutations correctly", () => {
+        const bareHttps = generateURLPermutations("https://firecrawl.dev").map(x => x.href);
+        expect(bareHttps.length).toBe(4);
+        expect(bareHttps.includes("https://firecrawl.dev/")).toBe(true);
+        expect(bareHttps.includes("https://www.firecrawl.dev/")).toBe(true);
+        expect(bareHttps.includes("http://firecrawl.dev/")).toBe(true);
+        expect(bareHttps.includes("http://www.firecrawl.dev/")).toBe(true);
+
+        const bareHttp = generateURLPermutations("http://firecrawl.dev").map(x => x.href);
+        expect(bareHttp.length).toBe(4);
+        expect(bareHttp.includes("https://firecrawl.dev/")).toBe(true);
+        expect(bareHttp.includes("https://www.firecrawl.dev/")).toBe(true);
+        expect(bareHttp.includes("http://firecrawl.dev/")).toBe(true);
+        expect(bareHttp.includes("http://www.firecrawl.dev/")).toBe(true);
+
+        const wwwHttps = generateURLPermutations("https://www.firecrawl.dev").map(x => x.href);
+        expect(wwwHttps.length).toBe(4);
+        expect(wwwHttps.includes("https://firecrawl.dev/")).toBe(true);
+        expect(wwwHttps.includes("https://www.firecrawl.dev/")).toBe(true);
+        expect(wwwHttps.includes("http://firecrawl.dev/")).toBe(true);
+        expect(wwwHttps.includes("http://www.firecrawl.dev/")).toBe(true);
+
+        const wwwHttp = generateURLPermutations("http://www.firecrawl.dev").map(x => x.href);
+        expect(wwwHttp.length).toBe(4);
+        expect(wwwHttp.includes("https://firecrawl.dev/")).toBe(true);
+        expect(wwwHttp.includes("https://www.firecrawl.dev/")).toBe(true);
+        expect(wwwHttp.includes("http://firecrawl.dev/")).toBe(true);
+        expect(wwwHttp.includes("http://www.firecrawl.dev/")).toBe(true);
+    })
+});

apps/api/src/lib/crawl-redis.ts

@@ -90,6 +90,44 @@ export async function getThrottledJobs(teamId: string): Promise<string[]> {
     return await redisConnection.zrangebyscore("concurrency-limiter:" + teamId + ":throttled", Date.now(), Infinity);
 }
 
+export function normalizeURL(url: string): string {
+    const urlO = new URL(url);
+    urlO.search = "";
+    urlO.hash = "";
+    return urlO.href;
+}
+
+export function generateURLPermutations(url: string | URL): URL[] {
+    const urlO = new URL(url);
+
+    // Construct two versions, one with www., one without
+    const urlWithWWW = new URL(urlO);
+    const urlWithoutWWW = new URL(urlO);
+    if (urlO.hostname.startsWith("www.")) {
+        urlWithoutWWW.hostname = urlWithWWW.hostname.slice(4);
+    } else {
+        urlWithWWW.hostname = "www." + urlWithoutWWW.hostname;
+    }
+
+    let permutations = [urlWithWWW, urlWithoutWWW];
+
+    // Construct more versions for http/https
+    permutations = permutations.flatMap(urlO => {
+        if (!["http:", "https:"].includes(urlO.protocol)) {
+            return [urlO];
+        }
+
+        const urlWithHTTP = new URL(urlO);
+        const urlWithHTTPS = new URL(urlO);
+        urlWithHTTP.protocol = "http:";
+        urlWithHTTPS.protocol = "https:";
+
+        return [urlWithHTTP, urlWithHTTPS];
+    });
+
+    return permutations;
+}
+
 export async function lockURL(id: string, sc: StoredCrawl, url: string): Promise<boolean> {
     if (typeof sc.crawlerOptions?.limit === "number") {
         if (await redisConnection.scard("crawl:" + id + ":visited") >= sc.crawlerOptions.limit) {
@@ -97,16 +135,16 @@ export async function lockURL(id: string, sc: StoredCrawl, url: string): Promise
         }
     }
 
-    try {
-        const urlO = new URL(url);
-        urlO.search = "";
-        urlO.hash = "";
-        url = urlO.href;
-    } catch (error) {
-        logger.warn("Failed to normalize URL " + JSON.stringify(url) + ": " + error);
+    url = normalizeURL(url);
+
+    let res: boolean;
+    if (!sc.crawlerOptions.deduplicateSimilarURLs) {
+        res = (await redisConnection.sadd("crawl:" + id + ":visited", url)) !== 0
+    } else {
+        const permutations = generateURLPermutations(url);
+        res = (await redisConnection.sadd("crawl:" + id + ":visited", ...permutations.map(x => x.href))) === permutations.length;
     }
 
-    const res = (await redisConnection.sadd("crawl:" + id + ":visited", url)) !== 0
     await redisConnection.expire("crawl:" + id + ":visited", 24 * 60 * 60, "NX");
     return res;
 }

apps/api/src/services/queue-worker.ts

@@ -23,6 +23,7 @@ import {
   getCrawl,
   getCrawlJobs,
   lockURL,
+  normalizeURL,
 } from "../lib/crawl-redis";
 import { StoredCrawl } from "../lib/crawl-redis";
 import { addScrapeJob } from "./queue-jobs";
@@ -318,6 +319,11 @@ async function processJob(job: Job & { id: string }, token: string) {
 
     if (job.data.crawl_id) {
       const sc = (await getCrawl(job.data.crawl_id)) as StoredCrawl;
+    
+      if (doc.metadata.url !== undefined && doc.metadata.sourceURL !== undefined && normalizeURL(doc.metadata.url) !== normalizeURL(doc.metadata.sourceURL)) {
+        logger.debug("Was redirected, locking new URL...");
+        await lockURL(job.data.crawl_id, sc, doc.metadata.url);
+      }
 
       await logJob({
         job_id: job.id as string,

apps/js-sdk/firecrawl/src/index.ts

@@ -86,6 +86,8 @@ export interface CrawlScrapeOptions {
     country?: string;
     languages?: string[];
   };
+  skipTlsVerification?: boolean;
+  removeBase64Images?: boolean;
 }
 
 export type Action = {
@@ -151,6 +153,7 @@ export interface CrawlParams {
   ignoreSitemap?: boolean;
   scrapeOptions?: CrawlScrapeOptions;
   webhook?: string;
+  deduplicateSimilarURLs?: boolean;
 }
 
 /**