extract

Instructions

Implementation Reference

• src/index.ts:217-293 (registration)

  Registration of the 'extract' tool using server.tool, including name, description, schema, and handler function.

  server.tool(
    "extract",
    "Extract information from the current page as text, html, or screenshot",
    {
      browserId: z.string().describe("Browser ID from a previous browse operation"),
      pageId: z.string().describe("Page ID from a previous browse operation"),
      type: z.enum(["text", "html", "screenshot"]).describe("Type of content to extract")
    },
    async ({ browserId, pageId, type }: { 
      browserId: string; 
      pageId: string; 
      type: "text" | "html" | "screenshot" 
    }) => {
      try {
        // Get the browser instance and page
        const instance = browserInstances.get(browserId);
        if (!instance) {
          throw new Error(`Browser instance not found: ${browserId}`);
        }
        
        const page = instance.pages.get(pageId);
        if (!page) {
          throw new Error(`Page not found: ${pageId}`);
        }
        
        let extractedContent = '';
        let screenshotPath = '';
        
        // Extract content based on requested type
        switch (type) {
          case "text":
            // Get visible text with stealth isolation
            extractedContent = await page.evaluate(`
              Array.from(document.querySelectorAll('body, body *'))
                .filter(element => {
                  const style = window.getComputedStyle(element);
                  return style.display !== 'none' && style.visibility !== 'hidden' && style.opacity !== '0';
                })
                .map(element => element.textContent)
                .filter(text => text && text.trim().length > 0)
                .join('\\n')
            `) as string;
            break;
          case "html":
            // Get HTML content
            extractedContent = await page.content();
            break;
          case "screenshot":
            // Take a screenshot
            screenshotPath = path.join(TEMP_DIR, `screenshot-${pageId}-${Date.now()}.png`);
            await page.screenshot({ path: screenshotPath });
            extractedContent = `Screenshot saved to: ${screenshotPath}`;
            break;
        }
        
        return {
          content: [
            {
              type: "text",
              text: type === "text" || type === "screenshot" 
                ? extractedContent.substring(0, 2000) + (extractedContent.length > 2000 ? '...' : '')
                : `Extracted HTML content (${extractedContent.length} characters). First 100 characters:\n${extractedContent.substring(0, 100)}...`
            }
          ]
        };
      } catch (error) {
        return {
          content: [
            {
              type: "text",
              text: `Failed to extract content: ${error}`
            }
          ]
        };
      }
    }
  );

• src/index.ts:225-292 (handler)

  Handler function for the 'extract' tool that retrieves text, HTML, or screenshot from the specified page using Patchright/Playwright.

  async ({ browserId, pageId, type }: { 
    browserId: string; 
    pageId: string; 
    type: "text" | "html" | "screenshot" 
  }) => {
    try {
      // Get the browser instance and page
      const instance = browserInstances.get(browserId);
      if (!instance) {
        throw new Error(`Browser instance not found: ${browserId}`);
      }
      
      const page = instance.pages.get(pageId);
      if (!page) {
        throw new Error(`Page not found: ${pageId}`);
      }
      
      let extractedContent = '';
      let screenshotPath = '';
      
      // Extract content based on requested type
      switch (type) {
        case "text":
          // Get visible text with stealth isolation
          extractedContent = await page.evaluate(`
            Array.from(document.querySelectorAll('body, body *'))
              .filter(element => {
                const style = window.getComputedStyle(element);
                return style.display !== 'none' && style.visibility !== 'hidden' && style.opacity !== '0';
              })
              .map(element => element.textContent)
              .filter(text => text && text.trim().length > 0)
              .join('\\n')
          `) as string;
          break;
        case "html":
          // Get HTML content
          extractedContent = await page.content();
          break;
        case "screenshot":
          // Take a screenshot
          screenshotPath = path.join(TEMP_DIR, `screenshot-${pageId}-${Date.now()}.png`);
          await page.screenshot({ path: screenshotPath });
          extractedContent = `Screenshot saved to: ${screenshotPath}`;
          break;
      }
      
      return {
        content: [
          {
            type: "text",
            text: type === "text" || type === "screenshot" 
              ? extractedContent.substring(0, 2000) + (extractedContent.length > 2000 ? '...' : '')
              : `Extracted HTML content (${extractedContent.length} characters). First 100 characters:\n${extractedContent.substring(0, 100)}...`
          }
        ]
      };
    } catch (error) {
      return {
        content: [
          {
            type: "text",
            text: `Failed to extract content: ${error}`
          }
        ]
      };
    }
  }

• src/index.ts:221-224 (schema)

  Input schema for the 'extract' tool defined using Zod, specifying browserId, pageId, and type parameters.

    browserId: z.string().describe("Browser ID from a previous browse operation"),
    pageId: z.string().describe("Page ID from a previous browse operation"),
    type: z.enum(["text", "html", "screenshot"]).describe("Type of content to extract")
  },