Trendigger

Trendigger is a trend-tracking platform that aggregates and visualizes what people are searching on Google. It offers an hour-by-hour breakdown of popular keywords across 125+ countries and worldwide, making it a handy tool for marketers, journalists, and content creators to monitor viral news.

How it works

[GitHub Actions, every 4h]
   ↓ Fetch Google Trends (Trending Now) for every supported geo via google-trends-now
   ↓ Upsert snapshots into Cloudflare D1 (one row per run × geo × keyword)
   ↓ Generate AI summaries for new keywords (cached forever per keyword × geo)
   ↓ Prune old snapshots, then render static HTML
   ↓ Deploy to Cloudflare Pages

[Visitor] → Cloudflare Pages (static HTML, near-zero DB reads)
            ↳ Dynamic per-date / per-keyword pages served by Pages Functions → D1

• Coverage: Worldwide aggregation plus all 125 countries Google Trends supports.
• Cadence: Collected every 4 hours; the homepage shows the latest run down to the minute, with 4 hours ago / 8 hours ago / … / Yesterday sections, then older dates.
• Summaries: Each keyword gets a short AI-written summary; cards without one stay clean and empty.
• Cost: Runs on free tiers (GitHub Actions + Cloudflare Pages + D1).

Supported geos

Worldwide (WW, aggregated at build time) plus 125 countries — see src/geos.js.

Local dev

npm install
npm run db:init                                   # create local D1 tables

# Run the pipeline
npm run collect                                   # fetch all geos into D1
npm run enrich                                    # generate AI summaries for new keywords
npm run prune                                     # drop snapshots older than the retention window
npm run build                                     # render static site into ./public
npm run serve                                     # preview at http://localhost:5173

# Inspect data
npm run query                                     # today's US trends
node src/query.js JP 2026-08-01                   # other geo/date
npm run query:stats                               # summary

Useful flags:

node collect-trends.js --geos US,JP --with-news 10 --delay-ms 1200
node src/enrich.js --top 50 --limit 200
node src/build.js --days 2 --clean                # full rebuild of the last 2 days

AI provider

enrich picks a provider via AI_PROVIDER:

• snippet (default in CI) — builds summaries from article snippets fetched on the fly. No API key needed.
• gemini — uses gemini-2.5-flash. Set GEMINI_API_KEY (or write it to ./gemini-key). Free tier is ~10 RPM / 250 RPD; intros are cached per (keyword, geo) so re-runs are essentially free. Use --limit N to cap spend and --force to regenerate.

Project layout

collect-trends.js        # fetch → D1
src/
  geos.js                # 125 countries + WW (lang, timezone, fetchGeo)
  db.js                  # D1 adapter (wrangler --local / --remote)
  enrich.js              # AI summary generation (snippet / gemini / cloudflare)
  providers.js           # intro providers + language labels
  build.js               # static site generator → ./public
  templates.js           # HTML templates (home, geo, keyword, sitemap)
  prune.js               # retention: drop old snapshots
schema.sql               # D1 schema
functions/               # Cloudflare Pages Functions (dynamic date/keyword pages)
public/                  # generated static site