Skip to content
mcorucu docs
AutoCore1.0.0-rc.1 · Release Candidate
1.0.0-rc.1 · Release Candidate2 min read

Search and Indexing

Learn how AutoCore uses PostgreSQL as truth and Meilisearch as a derived public search projection.

PostgreSQL is the authoritative transactional source. Meilisearch is a derived read projection optimized for public listing search, facets, sorting, and suggestions. Search documents flatten selected listing, catalog, seller, media, and location facts for query efficiency; sensitive fields such as VIN are not included in the document.

Index lifecycle

  1. A listing is created, updated, approved, or its status changes.
  2. The API requests synchronization of the listing document.
  3. The index service builds a denormalized document from PostgreSQL.
  4. Meilisearch upserts the document or removes it when it is no longer eligible.
  5. A full or bulk reindex can reconcile drift from PostgreSQL.

AutoCore search indexing lifecycle from a listing change through the API and PostgreSQL to a derived Meilisearch projection

Text alternative: PostgreSQL remains authoritative; the API and Worker update a derived Meilisearch projection, which may lag or become unavailable without changing transactional truth.

Consistency and degraded mode

Indexing is asynchronous at the service boundary, so a short projection delay is possible. The public search service falls back to PostgreSQL when Meilisearch is unconfigured or fails. The fallback preserves the response contract but does not provide the same Meilisearch facet distribution. This is a degraded mode, not evidence that the derived index is authoritative.

Filters and reindexing

Search supports text, catalog, seller, dealer-verification, price, year, mileage, location, technical, equipment, and media-presence filters. Index settings define searchable, filterable, and sortable attributes. Reindexing is an administrative recovery concept; detailed operational commands belong to the operations phase.

Related articles

Listing SearchabilityUnderstand which listing state and catalog facts can make a record visible in AutoCore search.Data StoresUnderstand what PostgreSQL, Redis, Meilisearch, and media storage own in the AutoCore architecture.Background ProcessingUnderstand AutoCore queue, Worker, indexing, media, notification, and scheduled-processing boundaries.Failure and Degraded ModesSee how AutoCore represents delayed search, media failure, worker gating, provider failure, and partial readiness.