Search Indexing Service · View 01 of 21 · Context and scope
Decisions
- The platform holds no authoritative copy of any business entity. Sources stay with their owners; the index is a projection.
- Four source systems, two with native change streams and two pushing changes through a signed API — the push path is the weakest supported case and is labelled as such.
- Enrichment is a dependency the platform is allowed to lose: the edge is drawn as a failure-tolerant call, not a requirement.
- Consuming products reach the index only through the query API, and only ever through an alias.
Deliberately out of scope
- The product's own search UI and ranking surfaces.
- Semantic retrieval and the embedding pipeline — a separate use case with its own index contract.
- The systems of record, and the analytics warehouse that also reads them.
Assumptions
- 40 million monthly active consumers, 180,000 active merchants, 12 product tenants, 80 million live documents.
- Sources can emit a daily full snapshot to object storage; two of the four cannot emit a change stream at all.