flowchart TD accTitle: Full archive search accDescr { NutsNews full archive search is backed by PostgreSQL full-text search in Supabase. } A[User enters search query] --> B[Frontend calls /api/search] B --> C[Backend executes Postgres full-text search] C --> D[Search weights apply by title, summary, source, category] D --> E[Return ranked results + nextPage metadata] E --> F[Web home feed renders search mode] F --> G[User taps result] G --> H[Open story detail] G --> I[User taps Clear] I --> J[Return to normal feed and infinite scroll]

Full archive search

NutsNews full archive search is backed by PostgreSQL full-text search in Supabase.

Database pieces

The migration adds:

articles.search_vector, a generated tsvector column
articles_search_vector_idx, a GIN index for fast search
articles_search_published_idx, an index for published story ordering
public.search_articles(search_query, page_size, page_offset), an RPC function used by the web API

Search weights:

Title: A
AI summary: B
Source: C
Category: C

Web API

The API route is:

GET /api/search?q=dogs&page=0&limit=20

Response shape:

{
  "articles": [],
  "nextPage": null,
  "query": "dogs",
  "page": 0,
  "pageSize": 20,
  "languageCode": "en"
}

Home page UX

The search box is built into the existing home feed component:

web/app/components/ArticleFeed.tsx

When a user searches, the home page switches from normal feed mode to archive search mode. The clear button returns to the normal feed and infinite scroll.

Local tests

cd /Users/ramideltoro/WebstormProjects/nutsnews3/web
npm run build
npm run dev

Then:

curl -s "http://localhost:3000/api/search?q=dogs&limit=5" | python3 -m json.tool
curl -I "http://localhost:3000/api/search?q=dogs&limit=5"

Open:

http://localhost:3000

Search for dogs, community, science, and a nonsense query like zzzzzzzzzz.

Primary diagram

Full archive search

Database pieces

Web API

Home page UX

Local tests