Skip to main content

❓ Frequently Asked Questions

This page covers common issues, deployment questions, and operational concerns related to ChildSafeNet (Web, API, AI, and Extension).

TroubleshootingDeploymentAI Operations

🗺 Roadmap

🔐 Security

🧠 AI Training

1. Extension & Pairing​

1
Edge

Why does Pair fail on Microsoft Edge?

Edge is Chromium-based, but its service worker may stop quickly when idle.

Check the following:

  • Extension is loaded and enabled
  • Service Worker console has no runtime errors
  • host_permissions in manifest.json include the target domain
  • API URL is correct and reachable
  • CORS is properly configured

How to debug

  1. Open edge://extensions
  2. Click your extension
  3. Open Service Worker
  4. Inspect console logs

2
Pair

Why does Pair fail with "port closed" error?

Possible causes:

  • Background service worker stopped
  • API unreachable
  • Invalid or expired pairing token
  • Runtime exception in background script

Fix

  • Reload extension
  • Regenerate pairing token
  • Check API logs
  • Verify JWT authentication

3
Status

Why does extension show NOT READY?

This means the content script was not injected.

Possible causes:

  • Missing host_permissions
  • Page loaded before extension initialized
  • Extension disabled

Fix

  • Reload extension
  • Reload page
  • Check manifest configuration

2. Deployment & Docs​

Can I deploy docs to GitHub Pages?​

Yes.

Use GitHub Actions to:

  1. Build Docusaurus site
  2. Deploy to gh-pages branch
  3. Configure repository settings → Pages → Source = gh-pages

Ensure:

  • baseUrl in docusaurus.config.js matches your repo name
  • url is set to your GitHub Pages domain

Example:

// docusaurus.config.js
url: "https://YOUR_USERNAME.github.io",
baseUrl: "/YOUR_REPO/",

Why does GitHub Pages show 404?​

Check:

  • baseUrl is correct
  • GitHub Pages branch selected properly
  • No hardcoded paths in links
  • Use useBaseUrl() for images and static assets

3. AI & Training​

Why not train after every new URL?​

Immediate retraining can:

  • Introduce noisy labels
  • Cause model instability
  • Slow down system

Use Periodic Training (Option Periodic) instead.

What if the new model performs worse?​

The system keeps the previous model as fallback.

If quality gates fail:

  • Do not activate new model
  • Roll back to previous version

How do I measure model quality?​

Check:

  • Accuracy
  • Precision (especially for Block categories)
  • Recall
  • F1 Score
  • False-positive rate on trusted domains

4. Security​

Is user browsing data stored?​

Only necessary metadata is stored:

  • URL
  • Prediction result
  • Timestamp

No full page content should be stored.

How do I secure the API?​

  • Use HTTPS
  • Use JWT authentication
  • Restrict admin endpoints
  • Enable CORS policy
  • Add rate limiting

5. Development​

How do I run full system locally?​

  1. Start SQL Server
  2. Run API
  3. Run Web app
  4. Install extension (Developer mode)
  5. Pair extension
  6. Trigger scan

How do I reset database?​

  • Drop database
  • Re-run migrations
  • Seed baseline dataset (if required)

6. Miscellaneous​

Can this system scale?​

Yes, with improvements:

  • Move AI engine to separate service
  • Add caching layer
  • Use message queue for training jobs
  • Deploy database with proper indexing

If your issue is not listed here:

  • Check logs (API + Extension)
  • Review Security page
  • Open a GitHub Issue with detailed steps