150+ Chinese Web Search Engine

Problem

Chinese web searches often span very different source types: engineering answers, academic papers, breaking news, e-commerce rules, medical, legal, and finance information all benefit from different sites. A single generic search frequently mixes low-relevance results, and it is hard for a model to judge source credibility, freshness, and whether cross-validation is needed.

cn-web-search addresses this by adding a routing, evidence-gathering, ranking, and validation flow before the model answers, rather than fetching raw pages directly.

How it works

The core mechanism is progressive intelligent routing. After reading the query, it detects domain, subtype, freshness, location, and negation signals, then selects an engine pool. The documented routes cover tech, academic, finance, news, medical, legal, tmall_ops, jd_ops, 1688_ops, and more, with P0 and P1 priority engines plus URL templates.

Execution adapts to query complexity: simple factual lookups can use one engine and one round, complex research can use parallel engines and gap analysis, and time-sensitive queries can append freshness parameters. Results then go through Rerank scoring across relevance, authority, freshness, completeness, and novelty. Source labels help distinguish official, media, community, and personal content, and suggest whether cross-validation is needed. v4.3.2 adds Tmall, JD, and 1688 e-commerce operations domains for platform rules, store operations, and campaign queries.

Boundaries

It depends on target pages being reachable. The documentation marks network availability with ✅, ⚠️, and ❌; engines that require proxies or are regionally unstable should be treated as fallbacks. It is useful for Chinese-language retrieval, multi-source verification, and structured evidence gathering, but not a universal scraper. High-risk medical, legal, and finance outputs should still be checked against official sources.