proxy-benchmark

Find out what's blocking your scraper: proxy, browser, host

3 followers

Find out what's blocking your scraper: proxy, browser, host

3 followers

An open-source harness that separates the causes of a scraping failure - exit IP, browser, TLS handshake, the machine you run it on, the target - by varying one at a time in the same hour. Every attempt is a JSONL row in the repo, and every table is generated from those rows. Built by Nodemaven for developers who are working in scraping, browser automation, anti-bot systems and others related fields.

Free

Launch tags:Open Source•Developer Tools•GitHub

Launch Team / Built With

Customer.ioAutomate Messaging Everywhere — Startups Get 12 Months Free

Promoted

OpenAI

GPT-6 Astra Challenge

What became possible in your product with Astra that was not practical before?

Maker

Astra made it practical to turn thousands of raw benchmark rows into an auditable research assistant. It can compare controlled runs, surface confounders, explain why experiments disagree, and trace every conclusion back to the exact underlying evidence - instead of producing a simplistic proxy or browser leaderboard

Report

2mo ago

Maker

📌

We built this because a scraping failure gets blamed on the proxy first, and usually nobody has checked. Three layers can refuse you - the exit address, the browser, the TLS handshake - plus the machine you are running on and the target itself. Debugging fails when you inspect the one above whichever is actually refusing you. So the harness varies one at a time, interleaved inside the same hour, because the hour turned out to be the largest confound we found. Two results that surprised us: - On Amazon the unmodified browser finished in the leading group. Stock Chromium 96% (419/436) against 63% (288/457) for the lowest anti-detect engine, over 3530 judged attempts. Six engines sit within four points at the top, a Fisher exact separates none of them, and the control is one of the six. - Chrome spends 43 MB per fresh profile talking to Google before you ask it for anything - 43.2 MB of a 43.4 MB window with the browser parked on about:blank. On a metered residential exit that is billed traffic for a file no target ever sees. Some of our earlier conclusions did not survive re-measurement, and both versions are in the notebook rather than the wrong one being quietly edited. 18,763 attempt rows, run ids and the generated tables are all committed, so any number here can be recomputed rather than taken. Built at NodeMaven, who sell proxies. No account needed - a gateway is a TOML file, so you point it at whatever you already have. Would like feedback from people working on scraping, browser automation, anti-bot or proxy infrastructure - especially anyone who can break the Amazon result.

Report

2mo ago

Pros

Cons

Reviews

OpenAI

GPT-6 Astra Challenge

What became possible in your product with Astra that was not practical before?

Maker

Report

2mo ago

Maker

📌

Report

2mo ago