# InfoReader — guide for AI agents

InfoReader is a free AI news-briefing site. All public content is also
published as static, unauthenticated JSON feeds on this domain. There are
no accounts and no write API.

## What you can do

- Read the full current briefing feed: GET /data/latest.json
  (generated_at, stats, briefing[12], by_topic{World, Business, AI/Tech,
  Science, Health, General}).
- Read only the top briefing: GET /data/briefing.json
- Read one topic desk: GET /data/topics/world.json, /data/topics/business.json,
  /data/topics/ai.json, /data/topics/science.json, /data/topics/health.json,
  /data/topics/general.json
- Read one story record: GET /data/stories/<story-id>.json
  (ids appear as `id` in the feeds; story pages live at /story/<slug>-<id>/).
- Read this documentation as JSON: GET /openapi.json (OpenAPI 3 description
  of the read-only feeds above).

## Data notes

- Every story carries: headline, summary, key_points, summary_method,
  category, source_count, sources[{name, url, published}], updated, score.
- `summary_method` tells you how a summary was produced: "extractive"
  (composed from the sources' own text), "groq" / "gemini" / "workers_ai"
  (that AI tier). Trust notes: summaries are derived from the linked
  sources only; cite the original publishers, not InfoReader, for the
  underlying reporting. Source URLs are in each story's `sources` array.
- Feeds are regenerated whenever the news pipeline runs (roughly hourly
  in production). `generated_at` is the freshness timestamp. There is no
  versioning or history API in this edition.

## Honest limits

- Read-only: no posting, commenting, or personalization API.
- Email signup is via the web form on the site (Brevo-backed from Phase 3);
  there is no subscription API yet.
- Rate limits: none published — the feeds are static files on a CDN.
  Please cache and poll politely (the data changes at most hourly).