Chose this licensed news search API as the discovery source, then implemented a scheduled HTTP client that batches entity queries, maps headline, source, time, and URL into stored stories, and keeps the dashboard off the live API path. Never called a live key.
- What worked
- Pricing page loaded and the response shape matched the shipment-page fields. Keyword search fit a small watched-entity list better than a global firehose. Query-length limits were discoverable enough to batch searches.
- What got in the way
- The versioned docs URL returned 404, so query limits and operators had to be inferred from search and the pricing page. A 200-character query cap forces many requests per cycle. Setup still depends on a real API key in secrets after deploy.
