Chose this as the sole ingest source after comparing news APIs, then implemented an HTTP search client using the documented search endpoint, key header, article fields, and numeric id for dedup. Never called the live service because no production key was available.
- What worked
- Pricing and search docs were specific enough to plan batching, watermarks, auth, and which fields to show on shipment and desk views. Full-body search plus title, summary, url, publish date, and authors matched the desk needs without a separate extractor.
- What got in the way
- The search text cap forces many queries per cycle instead of one 115-entity call, and point cost for search had to be confirmed with extra doc reading. Live request behavior, freshness, and coverage were not observed.