@xberg-io/crawlberg-wasm 1.7.1 → 1.8.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xberg-io/crawlberg-wasm",
-  "version": "1.7.1",
+  "version": "1.8.0",
   "private": false,
   "description": "High-performance web crawling engine",
   "license": "MIT",

package/pkg/bundler/crawlberg_wasm.d.ts

@@ -153,7 +153,7 @@ export class WasmBrowserConfig {
     free(): void;
     [Symbol.dispose](): void;
     static default(): WasmBrowserConfig;
-    constructor(mode?: WasmBrowserMode | null, backend?: WasmBrowserBackend | null, timeout?: bigint | null, wait?: WasmBrowserWait | null, blockUrlPatterns?: string[] | null, captureNetworkEvents?: boolean | null, sessionAffinity?: boolean | null, endpoint?: string | null, waitSelector?: string | null, extraWait?: bigint | null, proxy?: WasmProxyConfig | null, evalScript?: string | null, robotsUserAgent?: string | null);
+    constructor(mode?: WasmBrowserMode | null, backend?: WasmBrowserBackend | null, timeout?: bigint | null, overallTimeout?: bigint | null, shutdownTimeout?: bigint | null, wait?: WasmBrowserWait | null, blockUrlPatterns?: string[] | null, captureNetworkEvents?: boolean | null, sessionAffinity?: boolean | null, endpoint?: string | null, waitSelector?: string | null, extraWait?: bigint | null, proxy?: WasmProxyConfig | null, evalScript?: string | null, robotsUserAgent?: string | null);
     get backend(): string;
     set backend(value: WasmBrowserBackend);
     blockUrlPatterns: string[];
@@ -166,11 +166,15 @@ export class WasmBrowserConfig {
     set extraWait(value: bigint | null | undefined);
     get mode(): string;
     set mode(value: WasmBrowserMode);
+    get overallTimeout(): bigint | undefined;
+    set overallTimeout(value: bigint | null | undefined);
     get proxy(): WasmProxyConfig | undefined;
     set proxy(value: WasmProxyConfig | null | undefined);
     get robotsUserAgent(): string | undefined;
     set robotsUserAgent(value: string | null | undefined);
     sessionAffinity: boolean;
+    get shutdownTimeout(): bigint | undefined;
+    set shutdownTimeout(value: bigint | null | undefined);
     get timeout(): bigint | undefined;
     set timeout(value: bigint | null | undefined);
     get waitSelector(): string | undefined;
@@ -251,8 +255,9 @@ export class WasmContentConfig {
     free(): void;
     [Symbol.dispose](): void;
     static default(): WasmContentConfig;
-    constructor(outputFormat?: string | null, preprocessingPreset?: string | null, removeNavigation?: boolean | null, removeForms?: boolean | null, stripTags?: string[] | null, preserveTags?: string[] | null, excludeSelectors?: string[] | null, skipImages?: boolean | null, wrap?: boolean | null, wrapWidth?: number | null, includeDocumentStructure?: boolean | null, maxDepth?: number | null);
+    constructor(outputFormat?: string | null, preprocessingPreset?: string | null, removeNavigation?: boolean | null, removeForms?: boolean | null, stripTags?: string[] | null, preserveTags?: string[] | null, excludeSelectors?: string[] | null, skipImages?: boolean | null, wrap?: boolean | null, wrapWidth?: number | null, includeDocumentStructure?: boolean | null, extractMetadata?: boolean | null, maxDepth?: number | null);
     excludeSelectors: string[];
+    extractMetadata: boolean;
     includeDocumentStructure: boolean;
     get maxDepth(): number | undefined;
     set maxDepth(value: number | null | undefined);
@@ -297,9 +302,12 @@ export class WasmCrawlConfig {
     free(): void;
     [Symbol.dispose](): void;
     static default(): WasmCrawlConfig;
-    constructor(crawlStrategy?: WasmCrawlStrategyKind | null, respectRobotsTxt?: boolean | null, softHttpErrors?: boolean | null, stayOnDomain?: boolean | null, allowSubdomains?: boolean | null, includePaths?: string[] | null, excludePaths?: string[] | null, customHeaders?: any | null, requestTimeout?: bigint | null, maxRedirects?: number | null, retryCount?: number | null, retryCodes?: Uint16Array | null, cookiesEnabled?: boolean | null, removeTags?: string[] | null, content?: WasmContentConfig | null, downloadAssets?: boolean | null, assetTypes?: WasmAssetCategory[] | null, browser?: WasmBrowserConfig | null, userAgents?: string[] | null, captureScreenshot?: boolean | null, followDocumentUrls?: boolean | null, downloadDocuments?: boolean | null, documentMimeTypes?: string[] | null, saveBrowserProfile?: boolean | null, ssrf?: WasmSsrfPolicy | null, maxDepth?: number | null, maxPages?: number | null, maxLinksPerPage?: number | null, maxConcurrent?: number | null, contentFilter?: WasmContentFilterKind | null, bm25Query?: string | null, bm25Threshold?: number | null, userAgent?: string | null, rateLimitMs?: bigint | null, auth?: any | null, maxBodySize?: number | null, mapLimit?: number | null, mapSearch?: string | null, maxAssetSize?: number | null, proxy?: WasmProxyConfig | null, documentUrlDepth?: number | null, documentMaxSize?: number | null, documentOutputDir?: string | null, documentContentEncoding?: WasmDocumentContentEncoding | null, warcOutput?: string | null, browserProfile?: string | null, ssrfDenyPrivateExplicit?: boolean | null);
+    constructor(crawlStrategy?: WasmCrawlStrategyKind | null, respectRobotsTxt?: boolean | null, softHttpErrors?: boolean | null, stayOnDomain?: boolean | null, allowSubdomains?: boolean | null, includePaths?: string[] | null, excludePaths?: string[] | null, pathPatternsMatchQuery?: boolean | null, dedupIncludeQuery?: boolean | null, stripTrackingParams?: boolean | null, trackingParams?: string[] | null, customHeaders?: any | null, requestTimeout?: bigint | null, maxRedirects?: number | null, retryCount?: number | null, retryCodes?: Uint16Array | null, retryInitialDelayMs?: bigint | null, retryMaxDelayMs?: bigint | null, rateLimitJitterRatio?: number | null, cookiesEnabled?: boolean | null, removeTags?: string[] | null, content?: WasmContentConfig | null, downloadAssets?: boolean | null, assetTypes?: WasmAssetCategory[] | null, browser?: WasmBrowserConfig | null, userAgents?: string[] | null, captureScreenshot?: boolean | null, followDocumentUrls?: boolean | null, downloadDocuments?: boolean | null, documentMimeTypes?: string[] | null, saveBrowserProfile?: boolean | null, ssrf?: WasmSsrfPolicy | null, maxDepth?: number | null, maxPages?: number | null, maxLinksPerPage?: number | null, maxConcurrent?: number | null, contentFilter?: WasmContentFilterKind | null, bm25Query?: string | null, bm25Threshold?: number | null, userAgent?: string | null, rateLimitMs?: bigint | null, auth?: any | null, maxBodySize?: number | null, mapLimit?: number | null, mapSearch?: string | null, maxAssetSize?: number | null, proxy?: WasmProxyConfig | null, documentUrlDepth?: number | null, documentMaxSize?: number | null, documentOutputDir?: string | null, documentContentEncoding?: WasmDocumentContentEncoding | null, warcOutput?: string | null, browserProfile?: string | null, ssrfDenyPrivateExplicit?: boolean | null);
     /**
      * Validate the configuration, returning an error if any values are invalid.
+     *
+     * Only the first violation encountered is reported, not every one, so fixing a rejected
+     * configuration can surface a further error on the next attempt.
      */
     validate(): void;
     allowSubdomains: boolean;
@@ -321,6 +329,7 @@ export class WasmCrawlConfig {
     get crawlStrategy(): string;
     set crawlStrategy(value: WasmCrawlStrategyKind);
     customHeaders: any;
+    dedupIncludeQuery: boolean;
     get documentContentEncoding(): string | undefined;
     set documentContentEncoding(value: WasmDocumentContentEncoding | null | undefined);
     get documentMaxSize(): number | undefined;
@@ -352,8 +361,10 @@ export class WasmCrawlConfig {
     get maxPages(): number | undefined;
     set maxPages(value: number | null | undefined);
     maxRedirects: number;
+    pathPatternsMatchQuery: boolean;
     get proxy(): WasmProxyConfig | undefined;
     set proxy(value: WasmProxyConfig | null | undefined);
+    rateLimitJitterRatio: number;
     get rateLimitMs(): bigint | undefined;
     set rateLimitMs(value: bigint | null | undefined);
     removeTags: string[];
@@ -362,12 +373,16 @@ export class WasmCrawlConfig {
     respectRobotsTxt: boolean;
     retryCodes: Uint16Array;
     retryCount: number;
+    retryInitialDelayMs: bigint;
+    retryMaxDelayMs: bigint;
     saveBrowserProfile: boolean;
     softHttpErrors: boolean;
     get ssrfDenyPrivateExplicit(): boolean | undefined;
     set ssrfDenyPrivateExplicit(value: boolean | null | undefined);
     ssrf: WasmSsrfPolicy;
     stayOnDomain: boolean;
+    stripTrackingParams: boolean;
+    trackingParams: string[];
     get userAgent(): string | undefined;
     set userAgent(value: string | null | undefined);
     userAgents: string[];
@@ -394,7 +409,7 @@ export class WasmCrawlPageResult {
     free(): void;
     [Symbol.dispose](): void;
     static default(): WasmCrawlPageResult;
-    constructor(url?: string | null, normalizedUrl?: string | null, statusCode?: number | null, contentType?: string | null, html?: string | null, bodySize?: number | null, metadata?: WasmPageMetadata | null, links?: WasmLinkInfo[] | null, images?: WasmImageInfo[] | null, feeds?: WasmFeedInfo[] | null, jsonLd?: WasmJsonLdEntry[] | null, depth?: number | null, stayedOnDomain?: boolean | null, wasSkipped?: boolean | null, isPdf?: boolean | null, browserUsed?: boolean | null, detectedCharset?: string | null, markdown?: WasmMarkdownResult | null, extractedData?: any | null, extractionMeta?: WasmExtractionMeta | null, downloadedDocument?: WasmDownloadedDocument | null);
+    constructor(url?: string | null, normalizedUrl?: string | null, statusCode?: number | null, contentType?: string | null, html?: string | null, bodySize?: number | null, metadata?: WasmPageMetadata | null, links?: WasmLinkInfo[] | null, images?: WasmImageInfo[] | null, feeds?: WasmFeedInfo[] | null, jsonLd?: WasmJsonLdEntry[] | null, depth?: number | null, stayedOnDomain?: boolean | null, wasSkipped?: boolean | null, isPdf?: boolean | null, browserUsed?: boolean | null, finalUrl?: string | null, redirectCount?: number | null, noindexDetected?: boolean | null, nofollowDetected?: boolean | null, detectedCharset?: string | null, markdown?: WasmMarkdownResult | null, extractedData?: any | null, extractionMeta?: WasmExtractionMeta | null, downloadedDocument?: WasmDownloadedDocument | null);
     bodySize: number;
     browserUsed: boolean;
     contentType: string;
@@ -408,6 +423,7 @@ export class WasmCrawlPageResult {
     get extractionMeta(): WasmExtractionMeta | undefined;
     set extractionMeta(value: WasmExtractionMeta | null | undefined);
     feeds: WasmFeedInfo[];
+    finalUrl: string;
     html: string;
     images: WasmImageInfo[];
     isPdf: boolean;
@@ -416,7 +432,10 @@ export class WasmCrawlPageResult {
     get markdown(): WasmMarkdownResult | undefined;
     set markdown(value: WasmMarkdownResult | null | undefined);
     metadata: WasmPageMetadata;
+    nofollowDetected: boolean;
+    noindexDetected: boolean;
     normalizedUrl: string;
+    redirectCount: number;
     statusCode: number;
     stayedOnDomain: boolean;
     url: string;