Using the Undetected Browser Mode with the Docker Job Queue API

[pbtsrc]

Is it possible to use the Undetected Browser Mode with the Docker Job Queue API?

[gitsult4n]

Not as shipped

undetected mode is a constructor argument and not a config field

crawl4ai/async_crawler_strategy.py:76 takes browser_adapter and :95 is

self.adapter = browser_adapter or PlaywrightAdapter()

so the only way in is the python path from docs/md_v2/advanced/undetected-browser.md:124-134

AsyncWebCrawler(crawler_strategy=AsyncPlaywrightCrawlerStrategy(browser_config=cfg, browser_adapter=UndetectedAdapter()))

the server never goes through that
/crawl/job passes your browser_config down to handle_crawl_request deploy/docker/job.py:114 :133
and the pool builds the crawler with the default strategy deploy/docker/crawler_pool.py:113

crawler = AsyncWebCrawler(config=cfg, thread_safe=False)

BrowserConfig has no undetected flag so there is nothing to put in the payload

what you can do over the wire

{"browser_config": {"type": "BrowserConfig", "params": {"enable_stealth": true}}}

that is playwright-stealth on a normal chromium crawl4ai/browser_manager.py:763 — weaker than patchright but it is allowlisted for API bodies crawl4ai/async_configs.py:224-230

attaching to your own patchright browser is blocked too
cdp_url is in UNTRUSTED_FORBIDDEN_FIELDS crawl4ai/async_configs.py:203-208 and the job path loads with Provenance.UNTRUSTED deploy/docker/api.py:674 so sending it returns 400

so for real undetected behind the job queue you patch your own image

# deploy/docker/crawler_pool.py:113
crawler = AsyncWebCrawler(
    config=cfg,
    crawler_strategy=AsyncPlaywrightCrawlerStrategy(
        browser_config=cfg, browser_adapter=UndetectedAdapter()
    ),
    thread_safe=False,
)

that makes every pooled browser undetected
there is no per job switch since the pool key is the BrowserConfig signature crawler_pool.py:46