Skip to main content

fetchlib

A compliance-first page-fetch waterfall for the repo's scraping skills: tiered backends that escalate only when the current tier is genuinely blocked — L1 curl_cffi (TLS/JA3 impersonation, no JS) → L2 r.jina.ai (JS render to clean markdown) → L3 pluggable nodriver/browser → L4 opt-in paid unblocker — wrapped in api-pacer + AIMD per-domain rate control, a per-domain circuit breaker, JSONL fetch telemetry (url/tier/status/blocked/bytes/ms), robots.txt honoring, and optional Thompson-sampling backend learning (learn=True) because the best tier varies by site AND by IP. Use when the user wants to fetch or scrape pages without getting 403'd, needs one shared fetch layer instead of ad-hoc curl in every script, or asks which fetch method to use for a specific hard site and whether it's time to pay for an unblocker. Triggers on "fetchlib", "抓取被封了", "403 抓不到", "换个方式抓这个站", "统一抓取层", "要不要上付费代理/unblocker", "curl_cffi 还是 Jina", "抓竞品页面老失败", "scrape without getting blocked", "fetch waterfall", "which fetcher for this site".

Jump to install

Source facts

Repository
noique/cross-border-ecommerce-skills
Last source activity
July 26, 2026 at 17:01
Detected SKILL.md language
Chinese
Stars
38
Forks
6

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.