rebase

[EthanThePhoenix38]

---

Continue Tasks: ▶️ 1 queued — View all

[Copilot]

Pull request overview

This PR appears to be a “rebase”-style update that mixes French copy fixes with operational hardening across the Node aggregator, GitHub issue-based automation, and the in-browser reader UI.

Changes:

• Improved robustness in src/aggregator.js (config load error handling, email sending validation/rate limiting, language validation scaffolding).
• Hardened GitHub Actions workflows that parse issue forms (added validation/sanitization; updated labels with accents).
• Added localStorage availability checks and some UI text fixes in the reader UI.

Reviewed changes

Copilot reviewed 13 out of 14 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
templates/email-digest.html	Fixes French accents in email template copy.
src/aggregator.js	Adds config load guards + email digest improvements; introduces language validation and refactors dedupe.
readme-viewer.html	Adds localStorage availability checks; updates French copy; introduces scroll-spy code.
package.json	Bumps openai dependency version.
package-lock.json	Updates lockfile for dependency bumps.
js/tracker.js	Adds localStorage availability checks and better error handling around tracking.
config.json	Fixes French accents in category/tag labels.
CHANGELOG.md	Appends automated security update entries across multiple versions.
.github/workflows/manage-subscriber.yml	Adds stricter email validation/abuse heuristic; updates labels with accents.
.github/workflows/label.yml	Adds issues: write permission for labeler workflow.
.github/workflows/add-source.yml	Adds validation/sanitization for source proposals; updates category label text.
.github/ISSUE_TEMPLATE/unsubscribe.yml	Adds a new unsubscribe issue form.
.github/ISSUE_TEMPLATE/subscribe.yml	Updates labels/options with accents.
.github/ISSUE_TEMPLATE/new-source.yml	Updates labels/options with accents.

Comments suppressed due to low confidence (2)

.github/workflows/manage-subscriber.yml:96

• Category mapping mismatch: the subscribe issue form uses the option label "Tech Generale" (no accent) but the workflow expects "Tech Générale". If a user selects only that category, it won’t map and they’ll be subscribed to all categories due to the fallback. Align the option label in .github/ISSUE_TEMPLATE/subscribe.yml with the keys in catMap, or accept both spellings in the mapping.

            const catMap = {
              'AI / IA': 'ai',
              'Cybersecurity': 'cybersecurity',
              'IoT': 'iot',
              'Windows': 'windows',
              'Mac / Apple': 'mac',
              'Linux': 'linux',
              'Tech Générale': 'tech',
              'Entrepreneuriat': 'entrepreneurship',
              'Bourse & Finance': 'finance',
              'Crypto & Blockchain': 'crypto',
              'Open Source & GitHub': 'opensource',

src/aggregator.js:477

• deduplicateArticles() now computes threshold/normalizedTitles but the loop still references normalizedData, titleThreshold, and contentThreshold, which are no longer defined. This will throw at runtime and break aggregation. Either restore the original titleThreshold/contentThreshold + normalizedData variables, or refactor the comparison logic to consistently use the new threshold and the new normalized cache array.

  const threshold = (typeof SETTINGS.deduplication.similarityThreshold === 'number' && 
                     SETTINGS.deduplication.similarityThreshold > 0 && 
                     SETTINGS.deduplication.similarityThreshold <= 1) 
                    ? SETTINGS.deduplication.similarityThreshold 
                    : 0.7;
  const kept = [];
  const normalizedTitles = [];

  for (const article of articles) {
    // Normaliser le titre et le résumé de l'article courant
    const titleWords = normalizeText(article.title);
    const contentWords = normalizeText(article.summary || '');
    let isDuplicate = false;

    // Comparer avec chaque article déjà gardé
    for (let i = 0; i < normalizedData.length; i++) {
      const titleSim = textSimilarity(titleWords, normalizedData[i].title);
      const contentSim = textSimilarity(contentWords, normalizedData[i].content);

      // Conditions pour être considéré comme doublon:
      // 1. Titre très similaire (70%+)
      // 2. OU contenu très similaire (50%+)
      // 3. OU les deux modérément similaires (40%+ chacun)
      if (titleSim >= titleThreshold ||
          contentSim >= contentThreshold ||
          (titleSim >= 0.4 && contentSim >= 0.4)) {

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Language mapping will always fall back to 'en' because it checks languageRaw.includes('Français'), but the issue form options are "Francais / French", "Anglais / English", etc. Update the check to handle the actual option text (e.g., match both Francais and Français, or map explicit option values).

Suggested change

	const lang = languageRaw.includes('Français') ? 'fr' : 'en';
	const normalizedLanguage = (languageRaw || '').toLowerCase();
	const isFrench = normalizedLanguage.includes('francais') || normalizedLanguage.includes('français');
	const lang = isFrench ? 'fr' : 'en';

[Copilot]

Changelog entries appear to be duplicated across multiple released versions and include timestamps after the stated release date (e.g., entries dated 2026-02-14 under [1.1.0] - 2026-01-31). This makes the changelog misleading. Consider moving these automated security-update lines to an [Unreleased] section or the correct version, and avoid repeating the same items under multiple releases.

[Copilot]

Language validation is currently ineffective/inconsistent: lang is computed once (after an initial validation), but detectedLang may be reassigned later from full-article text without re-validation, and the returned article uses detectedLang || feedLang instead of the validated lang. Recompute/validate the final language after all detection steps and use that same value both in the generated HTML (<html lang=...>) and the returned article.lang.

[Copilot]

The preferences UI removed the prefHideRead checkbox, but the script still does document.getElementById('prefHideRead')... in setupPrefsPanel(). This will throw (null dereference) and break the preferences panel. Either re-add the checkbox markup, or remove the prefHideRead event wiring and the hideRead preference if it’s no longer supported.

Suggested change

	<div class="prefs-section">
	<h4>Articles lus</h4>
	<label>
	<input type="checkbox" id="prefHideRead" name="prefHideRead">
	Masquer les articles déjà lus
	</label>
	</div>

[Copilot]

applyFilters() is missing its closing brace before the new setupScrollSpy() declaration, which nests setupScrollSpy, setupPrefsPanel, and loadReadme inside applyFilters and leaves braces unbalanced. This will prevent the script from running correctly (and may cause a syntax error). Close applyFilters() before declaring the other functions, and call setupScrollSpy() after rendering (e.g., in the post-render setup of loadReadme()).