Skip to content

v1.1.0 — Ghost Mode + Production Bulk Scraping

Latest
Latest

Choose a tag to compare

View all tags
@ai-naymul ai-naymul released this 05 May 18:34
· 7 commits to main since this release
v1.1.0
524ebae

Ghost Mode — Production-Grade Stealth

BrowserPilot now passes every major bot detection benchmark on the internet. No noise injection, no brittle hacks — real Chromium with patchright, real GPU fingerprints, human-like behavior.

Stealth Benchmark Results

Benchmark Score
Sannysoft 29/29 Passed
Pixelscan 105/105 Clear (including CDP + DevTools)
Rebrowser Bot Detector 9/10 Pass
BrowserScan All Normal
DeviceAndBrowserInfo "You are human!"
BrowserLeaks WebRTC No IP Leak

Tier S Anti-Bot Bypass

Tested against the hardest commercial anti-bot systems — the ones that block 99% of automation tools:

Anti-Bot System Tier Sites Tested Result
DataDome S Foot Locker, Leboncoin, Vinted, Booking.com ✅ Loaded
Akamai A Nike, New Balance, Zalando ✅ Loaded
PerimeterX A Wayfair, Ticketmaster ✅ Loaded
Cloudflare Enterprise A Stake.com, LinkedIn ✅ Loaded

Production Bulk Scraping Engine

Not just single-page demos — BrowserPilot now handles production-scale bulk scraping across protected sites:

  • Concurrent workers — up to 10 parallel browsers, each with unique fingerprints
  • Context rotation — new identity every N pages without restarting the browser
  • Resource blocking — blocks images/fonts/CSS for 3-5x faster loads
  • Adaptive throttling — auto backs off on 429s, speeds up on success
  • Cookie persistence — sessions survive fingerprint rotation
  • Checkpoint/resume — crash mid-job? Pick up where you left off
  • Shared block intelligence — if one worker gets blocked, all workers skip that domain+proxy

Performance: 33.7 pages/min across DataDome, Akamai, PerimeterX, Cloudflare protected sites with zero blocks.

Ghost Mode Architecture

  • Patchright — Playwright fork that never calls Runtime.enable (defeats CDP detection)
  • Full Chromium via xvfb — real window.chrome, plugins, codecs, WebGL (not headless shell)
  • Fingerprint profiles — seed-based deterministic generation with OS-coherent viewport/UA/DPR/locale
  • Human behavior — Bezier mouse curves, variable typing speed, natural scroll patterns
  • Geo-matching — proxy country auto-maps to timezone + locale + languages
  • WebRTC prevention — local IP never leaked

API Endpoints

# Bulk scrape
POST /bulk              — start a bulk job
GET  /bulk/{job_id}     — check progress
POST /bulk/{job_id}/resume — resume from checkpoint
DELETE /bulk/{job_id}   — cancel job

Stats

  • 236 tests passing
  • 11/14 Tier S/A anti-bot sites loaded without blocks
  • 105/105 Pixelscan, 29/29 Sannysoft
  • Zero CDP detection, zero WebRTC leaks
Assets 2
Loading