StagehandCrawler + Stagehand browser plugin

[vdusek]

Summary

This is a feature parity issue — Crawlee JS already has the @crawlee/stagehand package that integrates Stagehand (the AI Browser Automation Framework by Browserbase) with Crawlee.

Note: This is specifically about browser-based AI automation (Stagehand + Playwright), not about AI/LLM-based HTML parsing for HTTP clients — that is tracked separately in #1593.

Crawlee JS references

• PR: feat: add @crawlee/stagehand package for AI-powered browser automation crawlee#3331 — feat: add @crawlee/stagehand package for AI-powered browser automation
• Issue: Stagehand guide crawlee#3064 — Stagehand guide

What was implemented in Crawlee JS

• StagehandCrawler extending BrowserCrawler
• Stagehand BrowserPlugin wrapping Stagehand for BrowserPool integration
• AI methods on the page object:
  • page.act() — Natural language browser interactions
  • page.extract() — Structured data extraction with Zod schemas
  • page.observe() — Get available page actions
  • page.agent() — Multi-step autonomous agents
• Full anti-blocking support via BrowserPool integration
• Browser fingerprinting automatically applied
• Support for LOCAL and BROWSERBASE environments
• Session-based fingerprint caching
• Automatic proxy rotation on blocking

Current state in Crawlee Python

In #1278 we explored Stagehand integration and added a documentation guide showing how to use Stagehand with PlaywrightCrawler. However, this is just a guide — there is no dedicated StagehandCrawler or Stagehand browser plugin in the codebase itself.

• Python Stagehand binding: https://github.com/browserbase/stagehand-python

Goal

To align with the JS implementation, we should have a dedicated StagehandCrawler and the corresponding Stagehand browser plugin directly in the Crawlee Python codebase — extending BrowserCrawler and integrating Stagehand through the BrowserPool / browser plugin system, rather than relying only on the documentation guide.