Fast full-text search for a website with filters

We have a large website with a big catalog and the built-in search powered by our CMS database has become a bottleneck. Queries are slow, results rank poorly and a single typo returns an empty page. We want to replace it with a separate search engine in the spirit of Meilisearch or Typesense and connect it to the frontend through a clean API. The goal is simple: a visitor starts typing a query and immediately sees relevant suggestions, the engine forgives typos, plurals and word order, and the most important products rise to the top of the results.

The task covers the entire function, not just plugging in a library. We need to design an index schema for our catalog, write an indexer that does the initial bulk load and then maintains the index as up-to-date when records are created, modified and deleted, and also provide autocomplete and full search endpoints. On top of this we need faceted filtering by category, price range, tags and availability so we can narrow a broad query without reloading the whole page. What matters most to us is relevance: query words should have weight by field, popular and in-stock products should rank higher, and we should be able to tweak ranking rules and synonyms without deploying new code.

Backend is in PHP, frontend is simple, we are ready to run the search service in a container next to the site. Please send a short plan: indexing strategy, how you will ensure data consistency and how you will measure relevance quality before launch.

— Index schema design and robust indexer for bulk and incremental catalog sync
— Autocomplete with typo tolerance and prefix matching plus a full search endpoint
— Faceted filters: category, price, tags, availability with fast counting
— Relevance tuning: field weights, custom ranking rules, synonyms and stop words
— Clean API for the frontend, container deployment and brief handover documentation