Skip to content

Supercharge your AI agents with data from the web and beyond. Building the library for superintelligence. 🔥

fix(crawler): relative URL handling on non-start pages (#893)

Gergő Móriczfbabc779f59c
diff --git a/apps/api/src/lib/crawl-redis.ts b/apps/api/src/lib/crawl-redis.ts
index b5936ad6d..bd79a86d9 100644
--- a/apps/api/src/lib/crawl-redis.ts
+++ b/apps/api/src/lib/crawl-redis.ts
@@ -166,10 +166,10 @@ export async function lockURLs(id: string, sc: StoredCrawl, urls: string[]): Pro
return res;
}
-export function crawlToCrawler(id: string, sc: StoredCrawl): WebCrawler {
+export function crawlToCrawler(id: string, sc: StoredCrawl, initialUrl?: string): WebCrawler {
const crawler = new WebCrawler({
jobId: id,
- initialUrl: sc.originUrl!,
+ initialUrl: initialUrl ?? sc.originUrl!,
includes: sc.crawlerOptions?.includes ?? [],
excludes: sc.crawlerOptions?.excludes ?? [],
maxCrawledLinks: sc.crawlerOptions?.maxCrawledLinks ?? 1000,
diff --git a/apps/api/src/scraper/WebScraper/crawler.ts b/apps/api/src/scraper/WebScraper/crawler.ts
index 9e3f7cd21..e5a25f37a 100644
--- a/apps/api/src/scraper/WebScraper/crawler.ts
+++ b/apps/api/src/scraper/WebScraper/crawler.ts
@@ -171,7 +171,7 @@ export class WebCrawler {
let fullUrl = href;
if (!href.startsWith("http")) {
try {
- fullUrl = new URL(href, this.baseUrl).toString();
+ fullUrl = new URL(href, url).toString();
} catch (_) {
return null;
}
diff --git a/apps/api/src/services/queue-worker.ts b/apps/api/src/services/queue-worker.ts
index 428e7e01f..5a0b28dbf 100644
--- a/apps/api/src/services/queue-worker.ts
+++ b/apps/api/src/services/queue-worker.ts
@@ -352,10 +352,10 @@ async function processJob(job: Job & { id: string }, token: string) {
if (!job.data.sitemapped && job.data.crawlerOptions !== null) {
if (!sc.cancelled) {
- const crawler = crawlToCrawler(job.data.crawl_id, sc);
+ const crawler = crawlToCrawler(job.data.crawl_id, sc, doc.metadata?.url ?? doc.metadata?.sourceURL ?? undefined);
const links = crawler.filterLinks(
- crawler.extractLinksFromHTML(rawHtml ?? "", sc.originUrl as string),
+ crawler.extractLinksFromHTML(rawHtml ?? "", doc.metadata?.url ?? doc.metadata?.sourceURL ?? sc.originUrl as string),
Infinity,
sc.crawlerOptions?.maxDepth ?? 10
);