AACSearch / guide
Choose an integration
Choose a practical path from your data to a search experience.
Choose your starting point
| You have | Start here | What you own |
|---|---|---|
| A CMS shop | CMS connectors, 1C-Bitrix | Catalogue quality and the storefront template |
| A product feed | YML feeds | Feed availability and update schedule |
| An application backend | Quick start, incremental indexing | Schema, stable IDs and update events |
| A custom frontend | Storefront UI, widget | Presentation and a restricted search key |
| An AI client | MCP setup, SDKs | Credentials and approval of write operations |
Every path converges on the same three things, in the same order:
What a document looks like — see indexing
Your catalogue, in one collection
Keywords, facets, ranking
The path you choose only decides how documents arrive.
Add capabilities on top
Whichever path you start from, the platform's capabilities switch on independently — each guide below works on top of any of the starting points above:
| When you want to… | Read |
|---|---|
| Suggest queries while the visitor types — and before they type | Query intelligence |
| Find documents by meaning in any language, by photo or by voice | Semantic, image and voice search |
| Let visitors ask in plain language ("red shoes under $50") | Natural-language search |
| Merchandise results — pins, boosts, redirects, synonyms | Search rules and synonyms |
| Give each facet value its own crawlable page | SEO landing pages |
| Learn from visitor behaviour — similar items, trending, personal | Events and recommendations |
| See what search earns — funnel, zero results, revenue attribution | Analytics |
| Test a relevance change before shipping it to everyone | A/B experiments |
| Answer questions that span the whole catalogue | Graph search or an assistant |
| React to events in your own systems | Webhooks |
Product catalogues
Index one document per stable product ID. Keep price, stock, brand and category as typed fields. Choose facetable fields deliberately, then expose a small set of named sorts through a search profile. Verify out-of-stock handling and category filters with real queries before publishing.
- Stable id
- your catalogue's own primary key
- Title
- the field
query_bysearches - Facets
- brand, category — marked facet: true
- Price
- float, for sorting and display
Content and knowledge bases
Keep a stable article ID, title, text and destination URL. Test terminology and common misspellings. Add an assistant when you need answers grounded in retrieved content; verify its citations and behavior when no useful source is found.
An existing search integration
Import a representative sample first. Compare expected results, filters and sorts before moving traffic. Keep full keys on your backend, create a restricted browser key, and follow the launch checklist. A profile release changes configuration; it does not back up documents or external policy definitions.
Review a dataset profiles it and proposes the field roles, so you are not guessing at the schema.
Run the queries your visitors actually use through both, and compare the top results, filters and sorts.
Full keys stay on the backend; the browser gets a restricted search key.
Keep the previous engine reachable until the launch checklist passes.