node-bridge-build

name

node-bridge-build

description

Build an "environment-patching" fallback bridge for anti-bot SDKs (e.g. PerimeterX / Akamai / similar vendors) using jsdom + 11 env-patching modules + a Python curl_cffi coordinator, without relying on a real browser. Trigger scenarios: (1) a pure-algorithm generator temporarily breaks because the SDK was upgraded and you need a Plan B firefight; (2) a new-site spike — get something working in 30 minutes, then decide whether a pure-algorithm approach is worth it; (3) high-bar endpoints (such as the iFood feed) where the pure-algorithm score is not good enough; (4) adapting the existing ifood/ template to a new site (Grubhub / Walmart / Doordash, etc.). Trigger keywords: node bridge env patching / jsdom fake browser / PX SDK fallback / SDK upgrade emergency / build a bridge for <site> / env-patching template adaptation.

Skill: Build a Node Bridge Environment Patch

🔝 Fallback / upgrade path: sdenv — https://github.com/pysunday/sdenv (718⭐, jsdom fork + plugin, the public ceiling for environment patching) When to switch to sdenv: see methodology.md §7.

User trigger phrases

• "Build a node bridge fallback for <site>"
• "The pure-algorithm approach for <site> stopped working, put together a jsdom env patch"
• "Adapt the ifood bridge to <new_site>"
• "Set up a Plan B for <site>"
• "The PX SDK was upgraded, the pure-algorithm approach is dead for now, build me a temporary bridge"

---

What you (the AI) must deliver

node_bridge/<site>/
├── perimeterx/
│   └── <sdk_file>.js              ← the locked target SDK
├── px-node-env/
│   ├── env/                       ← 11 JSDOM env-patching modules (site-specialized)
│   └── ...
├── px_node_bridge.js              ← IPC bridge (adapted from the ifood template SDK path)
├── px_cookie_generator.py         ← Python coordinator (curl_cffi chrome131)
├── package.json
└── README.md                      ← the actual working commands for this site + journal

Plus: a verified live run output (target cookie obtained + business API 200 + a complete journal committed to stample/live_validation/journal/).

---

Prerequisite skill dependencies (must read) ⭐

This skill is not self-contained. The node bridge = an organic synthesis of 3 upstream skills. Read these 3 first:

Skill	Role in this workflow	Repo-relative path
cdp-browser	Upstream — dump fingerprint values from a real Chrome	~/projects/Sourcing-AI-Skills/cdp-browser/
jni-env-patching	Methodology — the 4-step "inspect the real environment first, then supply reasonable values" method	~/projects/Sourcing-AI-Skills/jni-env-patching/
curl_cffi_integrate_scrapy_performance	Network layer — chrome131 TLS impersonation	~/projects/Sourcing-AI-Skills/curl_cffi_integrate_scrapy_performance/

For the full synthesis diagram, see methodology.md §0.

---

Workflow overview

Phase	Content	Effort	Details
Analyze	Research the target site / lock the SDK / analyze the PX endpoints	30 min - 1 h	methodology.md §1-3
Implement	Copy the ifood template → change path constants → dump real Chrome fingerprints → write env/	4 - 8 h	new_site_guide.md 9 steps
Validate	Get it working → call the real business API → write the journal	30 min - 1 h	new_site_guide.md step 9 + stample/live_validation/

---

Existing reference material

Resource	Path	Purpose
The complete working iFood bridge	node_bridge/ifood/	Copy directly as a template
iFood working log + full request/response	stample/live_validation/journal/2026-05-21.md	See what real output looks like
Methodology	node_bridge/skill/methodology.md	Required reading
Hands-on tutorial	node_bridge/skill/new_site_guide.md	9-step walkthrough
sdenv (upgrade fallback)	https://github.com/pysunday/sdenv	Use to upgrade when this bridge is not enough

---

The three-layer fallback decision (which layer this skill sits at)

Layer 1: Pure algorithm (revers/ + stample/<site>/px_cookie/) — primary recommendation
   ↓ when it stops working
Layer 2: Our node_bridge (what this skill teaches)  ← you are at this layer
   ↓ when it still does not work (hitting typeof document.all / V8-level Proxy detection)
Layer 3: sdenv (jsdom fork + plugin) — fallback

For details, see methodology.md §"When to upgrade to sdenv".

---

Things not to do

• ❌ Do not use generate_px.js directly — it is a standalone debug version left over from sourcing-cracked, and Node's default TLS will be detected by PX. You must go through the Python coordinator (px_cookie_generator.py).
• ❌ Do not hardcode credentials in demo code — proxy user/pass and account credentials all go through env vars (see stample/grub/px_cookie/business_api_demo.js).
• ❌ Do not add node_modules/ to git — use npm install (package.json already records it).
• ❌ Do not try to mock every browser API — it is impossible and unnecessary. Following the jni-env-patching 4-step method, only patch what the SDK actually reads.

---

Quick sanity check (30 seconds before starting)

# 1. Confirm the target site has a locked SDK
ls stample/<site>/source/

# 2. Confirm the ifood template runs (prerequisite)
cd node_bridge/ifood
ls perimeterx/main.min.js
python -c "import curl_cffi; print('OK')"

# 3. Confirm the cdp-browser skill is available
which python && python -c "import websockets; print('OK')"

# All OK → start on new_site_guide.md

---

Skill date 2026-05-22 · iFood _px3 verified working