How to deal with these problems?

[cyongping66-svg]

Identified Limitations:

1，Bing performs extremely poorly in Chinese search queries; searching for "Large Language Model Product Manager" returns only dictionary definitions.
2，Boss Zhiping uses dynamic JS rendering, so content cannot be captured via fetch‑web.
3，Zhihu pages are un-scrapable.

[Aas-ee]

Thanks for the feedback. These are valid limitations, but I think they should be handled separately.

1. Bing Chinese search quality

Bing can be used with Playwright mode. If request-based Bing results are poor or look unrelated, users can try SEARCH_MODE=playwright or pass searchMode=playwright for Bing
searches.

For Chinese queries, users can also try baidu, startpage, duckduckgo, or multiple engines instead of relying only on request-based Bing.

2. Dynamic JS-rendered pages, such as Boss Zhipin

This is a limitation of the current generic fetchWebContent behavior. It mainly uses the request-based path and only falls back to browser-assisted fetching in limited cases.

I opened a follow-up feature request to make this explicit by adding a renderMode parameter to fetchWebContent, so users can directly request Playwright/browser rendering
when needed.

3. Zhihu

Zhihu is not planned for now. Its anti-scraping mechanism is quite complex and changes frequently, so it would be difficult to support reliably and maintain over time.