crawlforge-mcp-server 4.2.12 → 4.5.0

package/src/tools/basic/_fetch.js

@@ -3,28 +3,43 @@
  * Applies an AbortController timeout and a default User-Agent.
  */
 
+import { config } from '../../constants/config.js';
+import { createRequire } from 'module';
+
+// Derive User-Agent from package version so it reflects the actual release.
+const _require = createRequire(import.meta.url);
+const _pkg = _require('../../../package.json');
+const CRAWLFORGE_UA = `CrawlForge/${_pkg.version} (+https://crawlforge.dev)`;
+
 /**
- * Fetch a URL with a configurable timeout.
+ * Fetch a URL with a configurable timeout and body-size cap.
+ *
+ * Content-Length is checked before the body is read; if absent or lying, the
+ * accumulated byte count is checked during streaming.  Both checks use the
+ * configurable cap from config.fetch.maxBodySize (env MAX_FETCH_BODY_SIZE,
+ * default 25 MB).
+ *
  * @param {string} url
  * @param {{ timeout?: number, headers?: Record<string,string> }} [options]
- * @returns {Promise<Response>}
+ * @returns {Promise<Response & { _body: string }>}
  */
 export async function fetchWithTimeout(url, options = {}) {
   const { timeout = 10000, headers = {} } = options;
+  const maxBodySize = config.fetch.maxBodySize;
 
   const controller = new AbortController();
   const timeoutId = setTimeout(() => controller.abort(), timeout);
 
+  let response;
   try {
-    const response = await fetch(url, {
+    response = await fetch(url, {
       signal: controller.signal,
       headers: {
-        'User-Agent': 'CrawlForge/1.0.0',
+        'User-Agent': CRAWLFORGE_UA,
         ...headers
       }
     });
     clearTimeout(timeoutId);
-    return response;
   } catch (error) {
     clearTimeout(timeoutId);
     if (error.name === 'AbortError') {
@@ -32,4 +47,62 @@ export async function fetchWithTimeout(url, options = {}) {
     }
     throw error;
   }
+
+  // --- Body-size cap ---
+
+  // Early rejection via Content-Length (servers may omit or lie — guard below
+  // handles that case). Optional-chained so non-standard responses (e.g. test
+  // mocks) without a Headers object don't throw.
+  const contentLengthHeader = response.headers?.get?.('content-length') ?? null;
+  if (contentLengthHeader !== null) {
+    const declared = parseInt(contentLengthHeader, 10);
+    if (!isNaN(declared) && declared > maxBodySize) {
+      throw new Error(
+        `Response body too large: Content-Length ${declared} exceeds limit of ${maxBodySize} bytes`
+      );
+    }
+  }
+
+  // Only the streaming byte-count guard requires a readable body. Responses
+  // without a ReadableStream body (already-buffered responses, test mocks)
+  // are returned unchanged so callers' native .text()/.json() still work.
+  if (!response.body || typeof response.body.getReader !== 'function') {
+    return response;
+  }
+
+  // Stream the body and abort if accumulated bytes exceed the cap.
+  const reader = response.body.getReader();
+  const chunks = [];
+  let totalBytes = 0;
+
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    totalBytes += value.byteLength;
+    if (totalBytes > maxBodySize) {
+      reader.cancel();
+      throw new Error(
+        `Response body too large: exceeded limit of ${maxBodySize} bytes`
+      );
+    }
+    chunks.push(value);
+  }
+
+  // Reassemble and expose as a response-like object that callers can use.
+  const bodyText = new TextDecoder().decode(
+    chunks.reduce((acc, chunk) => {
+      const merged = new Uint8Array(acc.byteLength + chunk.byteLength);
+      merged.set(acc, 0);
+      merged.set(chunk, acc.byteLength);
+      return merged;
+    }, new Uint8Array(0))
+  );
+
+  // Attach the pre-read text so callers can call .text() on the result.
+  // We wrap it in a minimal compatible object.
+  return Object.assign(response, {
+    text: () => Promise.resolve(bodyText),
+    json: () => Promise.resolve(JSON.parse(bodyText)),
+    _body: bodyText
+  });
 }

package/src/tools/basic/extractLinks.js

@@ -41,7 +41,7 @@ export async function extractLinksHandler({ url, filter_external, base_url }) {
           isExternal = false;
         }
 
-        if (filter_external && isExternal) return;
+        if (filter_external && !isExternal) return;
 
         links.push({ href: absoluteUrl, text, is_external: isExternal, original_href: href });
       } catch {

package/src/tools/basic/extractMetadata.js

@@ -1,11 +1,64 @@
 /**
  * extract_metadata — Extract page metadata (title, description, OG tags, etc.).
  * Extracted from server.js inline handler.
+ * B1: Parse JSON-LD and microdata; stronger title fallback chain (og:title → <title> → h1).
  */
 
 import { load } from 'cheerio';
 import { fetchWithTimeout } from './_fetch.js';
 
+/**
+ * Parse all JSON-LD blocks from the document.
+ * @param {import('cheerio').CheerioAPI} $
+ * @returns {Array}
+ */
+function parseJsonLd($) {
+  const results = [];
+  $('script[type="application/ld+json"]').each((_, el) => {
+    try {
+      const raw = $(el).html();
+      if (raw) results.push(JSON.parse(raw));
+    } catch {
+      // Skip invalid blocks
+    }
+  });
+  return results;
+}
+
+/**
+ * Parse microdata items (elements with itemscope).
+ * @param {import('cheerio').CheerioAPI} $
+ * @returns {Array}
+ */
+function parseMicrodata($) {
+  const results = [];
+  $('[itemscope]').each((_, el) => {
+    const $el = $(el);
+    const item = {
+      type: $el.attr('itemtype') || null,
+      properties: {}
+    };
+    $el.find('[itemprop]').each((_, prop) => {
+      const $prop = $(prop);
+      const name = $prop.attr('itemprop');
+      if (!name) return;
+      const tag = ($prop.get(0).tagName || '').toLowerCase();
+      let value;
+      if (tag === 'meta') value = $prop.attr('content');
+      else if (tag === 'a' || tag === 'link') value = $prop.attr('href');
+      else if (tag === 'img') value = $prop.attr('src');
+      else if (tag === 'time') value = $prop.attr('datetime') || $prop.text().trim();
+      else value = $prop.text().trim();
+      if (value) {
+        if (!item.properties[name]) item.properties[name] = [];
+        item.properties[name].push(value);
+      }
+    });
+    results.push(item);
+  });
+  return results;
+}
+
 /**
  * @param {{ url: string }} params
  */
@@ -19,7 +72,13 @@ export async function extractMetadataHandler({ url }) {
     const html = await response.text();
     const $ = load(html);
 
-    const title = $('title').text().trim() || $('h1').first().text().trim();
+    // Stronger title fallback: og:title → <title> → h1
+    const title =
+      $('meta[property="og:title"]').attr('content') ||
+      $('title').text().trim() ||
+      $('h1').first().text().trim() ||
+      '';
+
     const description =
       $('meta[name="description"]').attr('content') ||
       $('meta[property="og:description"]').attr('content') || '';
@@ -47,6 +106,9 @@ export async function extractMetadataHandler({ url }) {
       $('meta[charset]').attr('charset') ||
       $('meta[http-equiv="Content-Type"]').attr('content') || '';
 
+    const jsonLd = parseJsonLd($);
+    const microdata = parseMicrodata($);
+
     return {
       content: [{
         type: 'text',
@@ -61,6 +123,8 @@ export async function extractMetadataHandler({ url }) {
           charset,
           og_tags: ogTags,
           twitter_tags: twitterTags,
+          json_ld: jsonLd,
+          microdata,
           url: response.url
         }, null, 2)
       }]

package/src/tools/basic/extractText.js

@@ -2,12 +2,56 @@
  * extract_text — Extract clean text content from HTML.
  * Extracted from server.js inline handler.
  * D3.1: Added output_format:"markdown" option backed by Turndown.
+ * B1: Preserve block structure for text mode; use Readability + GFM for markdown mode.
  */
 
 import { load } from 'cheerio';
+import { JSDOM } from 'jsdom';
+import { Readability } from '@mozilla/readability';
 import { fetchWithTimeout } from './_fetch.js';
 import { htmlToMarkdown } from '../../utils/htmlToMarkdown.js';
 
+// Block-level elements whose boundaries should become paragraph breaks
+const BLOCK_ELEMENTS = new Set([
+  'p', 'div', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6',
+  'li', 'blockquote', 'pre', 'td', 'th', 'dt', 'dd',
+  'article', 'section', 'figure', 'figcaption', 'aside',
+  'header', 'footer', 'main', 'nav', 'form', 'fieldset',
+  'table', 'tr', 'caption'
+]);
+
+/**
+ * Extract plain text from a cheerio root preserving block-element paragraph breaks.
+ * @param {import('cheerio').CheerioAPI} $ - loaded cheerio instance
+ * @returns {string}
+ */
+function extractBlockText($) {
+  const parts = [];
+
+  function walk(node) {
+    if (node.type === 'text') {
+      const t = node.data.replace(/[ \t\r\n]+/g, ' ');
+      if (t.trim()) parts.push(t);
+      return;
+    }
+    if (node.type !== 'tag') return;
+    const tag = node.tagName ? node.tagName.toLowerCase() : '';
+    const isBlock = BLOCK_ELEMENTS.has(tag);
+    if (isBlock) parts.push('\n\n');
+    for (const child of (node.children || [])) {
+      walk(child);
+    }
+    if (isBlock) parts.push('\n\n');
+  }
+
+  const body = $('body').get(0);
+  if (body) {
+    for (const child of (body.children || [])) walk(child);
+  }
+
+  return parts.join('').replace(/\n{3,}/g, '\n\n').trim();
+}
+
 /**
  * @param {{ url: string, remove_scripts?: boolean, remove_styles?: boolean, output_format?: "text"|"markdown" }} params
  */
@@ -26,20 +70,32 @@ export async function extractTextHandler({ url, remove_scripts, remove_styles, o
 
     $('nav, header, footer, aside, .advertisement, .ad, .sidebar').remove();
 
-    const text = $('body').text().replace(/\s+/g, ' ').trim();
-
     const result = {
-      word_count: text.split(/\s+/).filter(w => w.length > 0).length,
-      char_count: text.length,
       url: response.url
     };
 
     if (output_format === 'markdown') {
-      result.markdown = htmlToMarkdown($.html('body'));
+      // Run Readability first to get main content, then convert to GFM markdown
+      let articleHtml;
+      try {
+        const dom = new JSDOM(html, { url: response.url });
+        const reader = new Readability(dom.window.document);
+        const article = reader.parse();
+        articleHtml = article ? article.content : $.html('body');
+      } catch {
+        articleHtml = $.html('body');
+      }
+      result.markdown = htmlToMarkdown(articleHtml);
       result.output_format = 'markdown';
+      const plainText = result.markdown.replace(/[#*`_\[\]]/g, '').replace(/\s+/g, ' ').trim();
+      result.word_count = plainText.split(/\s+/).filter(w => w.length > 0).length;
+      result.char_count = plainText.length;
     } else {
+      const text = extractBlockText($);
       result.text = text;
       result.output_format = 'text';
+      result.word_count = text.split(/\s+/).filter(w => w.length > 0).length;
+      result.char_count = text.length;
     }
 
     return {

package/src/tools/basic/scrapeStructured.js

@@ -1,15 +1,33 @@
 /**
  * scrape_structured — Extract structured data using CSS selectors.
  * Extracted from server.js inline handler.
+ * B1: Support attribute extraction (selector@attr), add max_results,
+ *     fix elements_found to report real per-field DOM match counts.
  */
 
 import { load } from 'cheerio';
 import { fetchWithTimeout } from './_fetch.js';
 
 /**
- * @param {{ url: string, selectors: Record<string, string> }} params
+ * Parse a selector string that may include an attribute suffix: "css@attr"
+ * e.g. "a.link@href" -> { selector: "a.link", attribute: "href" }
+ *      "img@src"      -> { selector: "img",    attribute: "src" }
+ *      "h1"           -> { selector: "h1",      attribute: null }
+ * @param {string} raw
+ * @returns {{ selector: string, attribute: string|null }}
  */
-export async function scrapeStructuredHandler({ url, selectors }) {
+function parseSelectorSpec(raw) {
+  const atIdx = raw.lastIndexOf('@');
+  if (atIdx > 0) {
+    return { selector: raw.slice(0, atIdx), attribute: raw.slice(atIdx + 1) };
+  }
+  return { selector: raw, attribute: null };
+}
+
+/**
+ * @param {{ url: string, selectors: Record<string, string>, max_results?: number }} params
+ */
+export async function scrapeStructuredHandler({ url, selectors, max_results }) {
   try {
     const response = await fetchWithTimeout(url);
     if (!response.ok) {
@@ -19,22 +37,42 @@ export async function scrapeStructuredHandler({ url, selectors }) {
     const html = await response.text();
     const $ = load(html);
     const results = {};
+    const matchCounts = {};
 
-    for (const [fieldName, selector] of Object.entries(selectors)) {
+    for (const [fieldName, rawSelector] of Object.entries(selectors)) {
       try {
-        const elements = $(selector);
-        if (elements.length === 0) {
+        const { selector, attribute } = parseSelectorSpec(rawSelector);
+        let elements = $(selector);
+        const domCount = elements.length;
+        matchCounts[fieldName] = domCount;
+
+        if (domCount === 0) {
           results[fieldName] = null;
-        } else if (elements.length === 1) {
-          results[fieldName] = elements.text().trim();
         } else {
-          results[fieldName] = elements.map((_, el) => $(el).text().trim()).get();
+          // Apply max_results cap if specified
+          if (max_results != null && max_results > 0 && domCount > max_results) {
+            elements = elements.slice(0, max_results);
+          }
+
+          const extract = (el) => {
+            if (attribute) {
+              return $(el).attr(attribute) ?? null;
+            }
+            return $(el).text().trim();
+          };
+
+          if (elements.length === 1) {
+            results[fieldName] = extract(elements.get(0));
+          } else {
+            results[fieldName] = elements.map((_, el) => extract(el)).get();
+          }
         }
       } catch (selectorError) {
         results[fieldName] = {
-          error: `Invalid selector: ${selector}`,
+          error: `Invalid selector: ${rawSelector}`,
           message: selectorError.message
         };
+        matchCounts[fieldName] = 0;
       }
     }
 
@@ -44,7 +82,7 @@ export async function scrapeStructuredHandler({ url, selectors }) {
         text: JSON.stringify({
           data: results,
           selectors_used: selectors,
-          elements_found: Object.keys(results).length,
+          elements_found: matchCounts,
           url: response.url
         }, null, 2)
       }]

package/src/tools/crawl/crawlDeep.js

@@ -14,6 +14,7 @@ const CrawlDeepSchema = z.object({
   follow_external: z.boolean().optional().default(false),
   respect_robots: z.boolean().optional().default(true),
   extract_content: z.boolean().optional().default(true),
+  content_max_length: z.number().min(1).max(100000).optional().default(500),
   concurrency: z.number().min(1).max(20).optional().default(10),
   enable_link_analysis: z.boolean().optional().default(true),
   link_analysis_options: z.object({
@@ -217,7 +218,7 @@ export class CrawlDeepTool {
         errors: results.errors.length,
         duration_ms: duration,
         pages_per_second: results.urls.length / (duration / 1000),
-        results: this.formatResults(results.results, validated.extract_content),
+        results: this.formatResults(results.results, validated.extract_content, validated.content_max_length),
         errors: results.errors,
         stats: results.stats,
         site_structure: this.analyzeSiteStructure(results.urls),
@@ -240,7 +241,7 @@ export class CrawlDeepTool {
     }
   }
 
-  formatResults(results, includeContent) {
+  formatResults(results, includeContent, contentMaxLength = 500) {
     return results.map(result => {
       const formatted = {
         url: result.url,
@@ -250,12 +251,19 @@ export class CrawlDeepTool {
         content_length: result.contentLength,
         timestamp: result.timestamp
       };
-      
+
       if (includeContent) {
-        formatted.content = result.content ? result.content.substring(0, 500) + '...' : '';
+        const raw = result.content || '';
+        if (raw.length > contentMaxLength) {
+          formatted.content = raw.substring(0, contentMaxLength);
+          formatted.truncated = true;
+        } else {
+          formatted.content = raw;
+          formatted.truncated = false;
+        }
         formatted.metadata = result.metadata;
       }
-      
+
       return formatted;
     });
   }

package/src/tools/crawl/mapSite.js

@@ -3,6 +3,7 @@ import { load } from 'cheerio';
 import { DomainFilter } from '../../utils/domainFilter.js';
 import { normalizeUrl, getBaseUrl } from '../../utils/urlNormalizer.js';
 import { CacheManager } from '../../core/cache/CacheManager.js';
+import { SitemapParser } from '../../utils/sitemapParser.js';
 
 const MapSiteSchema = z.object({
   url: z.string().url(),
@@ -33,6 +34,7 @@ export class MapSiteTool {
     this.timeout = timeout;
     // Per-session result cache: avoids redundant site maps for the same root URL
     this.cache = cacheEnabled ? new CacheManager({ ttl: cacheTTL }) : null;
+    this.sitemapParser = new SitemapParser({ userAgent, timeout, enableCaching: cacheEnabled, cacheTTL });
   }
 
   async execute(params) {
@@ -131,61 +133,32 @@ export class MapSiteTool {
   }
 
   async fetchSitemapUrls(baseUrl, domainFilter = null) {
+    // Discover sitemaps via robots.txt and common paths, then parse with full
+    // SitemapParser support (sitemap-index recursion, gzip, CDATA/entities).
+    const discovered = await this.sitemapParser.discoverSitemaps(baseUrl, {
+      checkRobotsTxt: true,
+      checkCommonPaths: true,
+      checkSitemapIndex: false
+    });
+
     const urls = new Set();
-    const sitemapUrls = [
-      `${baseUrl}/sitemap.xml`,
-      `${baseUrl}/sitemap_index.xml`,
-      `${baseUrl}/sitemap-index.xml`,
-      `${baseUrl}/sitemaps.xml`
-    ];
-
-    for (const sitemapUrl of sitemapUrls) {
+    for (const sitemapUrl of discovered) {
       try {
-        const response = await this.fetchWithTimeout(sitemapUrl);
-        if (response.ok) {
-          const xml = await response.text();
-          const extractedUrls = this.parseSitemap(xml);
-          
-          // Apply domain filter if provided
-          extractedUrls.forEach(url => {
+        const parsed = await this.sitemapParser.parseSitemap(sitemapUrl, {
+          includeMetadata: false,
+          followIndexes: true
+        });
+        if (parsed.success) {
+          for (const entry of parsed.urls) {
+            const url = entry.loc || entry;
             if (!domainFilter || domainFilter.isAllowed(url).allowed) {
               urls.add(url);
             }
-          });
-          
-          // If we found a sitemap, don't try others
-          if (urls.size > 0) break;
+          }
         }
+        if (urls.size > 0) break;
       } catch {
-        // Continue to next sitemap URL
-      }
-    }
-
-    return Array.from(urls);
-  }
-
-  parseSitemap(xml) {
-    const urls = new Set();
-    
-    // Extract URLs from sitemap
-    const urlMatches = xml.match(/<loc>([^<]+)<\/loc>/g);
-    if (urlMatches) {
-      urlMatches.forEach(match => {
-        const url = match.replace(/<\/?loc>/g, '').trim();
-        if (url) urls.add(url);
-      });
-    }
-
-    // Check for nested sitemaps (sitemap index)
-    const sitemapMatches = xml.match(/<sitemap>[\s\S]*?<\/sitemap>/g);
-    if (sitemapMatches) {
-      for (const sitemapMatch of sitemapMatches) {
-        const locMatch = sitemapMatch.match(/<loc>([^<]+)<\/loc>/);
-        if (locMatch && locMatch[1]) {
-          // We could recursively fetch nested sitemaps here
-          // For now, just add the sitemap URL itself
-          urls.add(locMatch[1]);
-        }
+        // Continue to next discovered sitemap
       }
     }
 
@@ -362,7 +335,7 @@ export class MapSiteTool {
       max_depth: 0,
       average_depth: 0,
       url_lengths: {
-        min: Infinity,
+        min: null,
         max: 0,
         average: 0
       }
@@ -374,7 +347,7 @@ export class MapSiteTool {
     for (const url of urls) {
       try {
         const urlObj = new URL(url);
-        
+
         // Count secure URLs
         if (urlObj.protocol === 'https:') {
           stats.secure_urls++;
@@ -396,7 +369,7 @@ export class MapSiteTool {
         // Track URL lengths
         const length = url.length;
         totalLength += length;
-        stats.url_lengths.min = Math.min(stats.url_lengths.min, length);
+        stats.url_lengths.min = stats.url_lengths.min === null ? length : Math.min(stats.url_lengths.min, length);
         stats.url_lengths.max = Math.max(stats.url_lengths.max, length);
 
         // Track file extensions

package/src/tools/extract/analyzeContent.js

@@ -266,9 +266,9 @@ export class AnalyzeContentTool {
     };
 
     const keywordStr = keywords.join(' ').toLowerCase();
-    
+
     for (const [category, categoryKeywords] of Object.entries(categories)) {
-      const matches = categoryKeywords.filter(word => keywordStr.includes(word));
+      const matches = categoryKeywords.filter(word => new RegExp(`\\b${word}\\b`).test(keywordStr));
       if (matches.length > 0) {
         return category;
       }
@@ -394,13 +394,18 @@ export class AnalyzeContentTool {
       anticipation: ['excited', 'eager', 'looking forward', 'anticipating', 'expecting']
     };
 
-    const words = text.toLowerCase().split(/\s+/);
+    const lowerText = text.toLowerCase();
+    const words = lowerText.split(/\s+/);
     const emotions = [];
 
     for (const [emotion, emotionKeywords] of Object.entries(emotionWords)) {
-      const matches = words.filter(word => emotionKeywords.some(keyword => word.includes(keyword)));
-      if (matches.length > 0) {
-        const intensity = Math.min(1, matches.length / Math.max(words.length / 100, 1));
+      const matchCount = emotionKeywords.reduce((count, keyword) => {
+        const re = new RegExp(`\\b${keyword}\\b`, 'g');
+        const found = lowerText.match(re);
+        return count + (found ? found.length : 0);
+      }, 0);
+      if (matchCount > 0) {
+        const intensity = Math.min(1, matchCount / Math.max(words.length / 100, 1));
         emotions.push({
           emotion,
           intensity: Math.round(intensity * 100) / 100