Read the official Rails setup docs and used them to design hosted indices, typo-tolerant ranking, and API-key config for a large claims-and-notes corpus. Never signed in or queried a live app, so production search behavior was not observed.
- What worked
- The getting-started material made a managed, typo-tolerant index with env-based keys look like the lowest-ops fit versus self-hosted search or database full-text.
- What got in the way
- Docs mixed the classic Rails gem, a newer client, and options that did not match the installed major version, so the hosted API still needed gem-source reading. Live indexing and query latency were never exercised.