Architecture
Visual reference for the AACsearch internals — write path, read path, security model, connector lifecycle, and the analytics feedback loop.
This section is the visual reference for how AACsearch is built. Five inline Mermaid diagrams cover the request flows that matter most for integrators, operators, and security reviewers.
If you are looking for the prose explanation of the monorepo layout, service boundaries, and technology choices, start with Overview → Architecture. This section zooms in on the runtime flows.
Diagrams
- Write path — how a document write enters AACsearch,
through
SearchIngestBuffer/SearchSyncOutbox, into the Typesense alias. - Read path — how a search query flows from SDK to
Typesense
multi_searchand back. - Key types & security model — the four
key categories, their
ss_*prefixes, hash-only storage, and scoped-token HMAC + TTL. - Connector lifecycle — the six connector operations: handshake, heartbeat, full-sync, delta-sync, delete, diagnostics.
- Analytics feedback loop — how search → click → conversion events aggregate and feed back into relevance tuning.
Plans & Limits
The canonical AACsearch plan matrix, Search-Unit definition, quota catalog, soft/hard semantics, and the code paths that enforce them.
Write path
How a document write enters AACsearch — from POST /v1/indexes/.../documents through SearchIngestBuffer / SearchSyncOutbox and the worker, into the Typesense alias.