Germaine Brown
98073dc636
Publish Docs Site / build (push) Failing after 5s
- mkdocs.yml with dark slate theme, nav for 12 ITPP projects - build-docs.sh aggregates docs from all project repos - .gitea/workflows/docs-publish.yml for nightly rebuild+deploy - README and CHANGELOG for the itpp-docs repo itself - docs-source/ populated from all 12 repos - site/ ready for deployment to docs.itpropartner.com
2.0 KiB
2.0 KiB
Super Search — Cloudflare Bypass
Added: July 21, 2026 Version: 2.1.0
Extraction Chain
| Tier | Provider | What it handles | Fallback Trigger |
|---|---|---|---|
| 1 | Trafilatura | Normal sites | Error OR CF challenge detected |
| 2 | Browserless Chrome | CF-protected sites | Chrome render + Trafilatura |
| 3 | Firecrawl | Everything else | API-based extraction |
CF Detection
Nine detection markers for caught challenge pages (from Hound's code + additional):
- cf-turnstile, challenges.cloudflare.com/turnstile
- cf_chl_opt, __cf_chl
- cf-browser-verification, challenge-platform, cf-mitigated
- "Checking your browser", "Just a moment"
Infrastructure
- Browserless Chrome on app1 (152.53.36.131), port 3005
- Caddy proxy: app1:3006 → localhost:3005
- Firewall: only Core (152.53.192.33) can reach port 3006
- Super Search:
/root/docker/super-search/server.py
Verify
# Test CF bypass
cd /root/docker/super-search && source venv/bin/activate
python3 -c "
from server import _extract_one
import asyncio
r = asyncio.run(_extract_one('https://nowsecure.nl'))
print(r['provider']) # Should be 'trafilatura' or 'browserless'
"
v2.1 Features (Added July 21, 2026)
13 Tools
| # | Tool | Description |
|---|---|---|
| 1-10 | Original 10 | Search, extract, lookup, suggest, images |
| 11 | web_search_fast |
Parallel racing: all providers fire simultaneously |
| 12 | screenshot |
Browserless Chrome → base64 PNG |
| 13 | circuit_status |
Provider health states (closed/open/half-open) |
Circuit Breaker
- 8 provider circuits: 3 failures → open for 60s
- Prevents hammering dead providers
- Auto-recovers when provider comes back
Parallel Racing
web_search_fast: SearXNG ∥ Exa ∥ DuckDuckGo ∥ Wikipedia- First successful result wins — others cancelled
- Typically 2-3x faster than sequential fallback
Screenshot
- Browserless Chrome on port 3006 (Caddy HTTP proxy)
- Full-page or viewport capture
- Base64-encoded PNG in JSON response