For most websites the practical choice is between a hosted search platform and a lightweight engine like Typesense or Meilisearch. Elasticsearch suits large, complex deployments with a team to run it; Postgres full-text suits small sites already on Postgres that want no new infrastructure.
| Engine | Strength | Cost of ownership | Fits |
|---|---|---|---|
| Typesense | Fast, simple, typo-tolerant | Low | Most sites |
| Meilisearch | Excellent DX, instant search | Low | Small to mid sites |
| Elasticsearch / OpenSearch | Power and flexibility | High | Large, complex |
| Postgres full-text | No new infrastructure | Very low | Small sites |
| Vector database | Semantic retrieval | Medium | Paired with keyword |
| Hosted platform | Everything included | Subscription | Teams without search engineers |
An engine is not a search experience. It gives you retrieval; you still owe the query understanding, ranking policy, front end, analytics and merchandising. Choosing an engine is choosing to build those.
If search is a differentiator you intend to own, that is the right trade. If it needs to work well and be someone else's operational problem, a platform is the shorter path.