Valkey Search 1.2 adds text filters and server-side aggregations

Valkey announced Search 1.2 on March 17, extending its search module beyond vectors with full-text, exact-tag and numeric-range queries, plus server-side aggregation. The announcement shows those query types combined in one request.

Search becomes a richer application primitive

The module can group matching results and calculate counts, sums and averages, then filter or sort the output. For an application team, that creates a concrete opportunity to move a bounded search-and-summarize operation closer to indexed data.

Consider a parts catalog where users select a component category, specify a price ceiling and search a description. Our analysis is that the useful comparison is the complete user request: the rows returned, facet counts, update visibility and latency under ingestion. Counting fewer network requests alone would miss incorrect filters or an expensive index-maintenance path.

Consistency depends on where the query runs

Valkey says an indexed write completes after its index update is applied. Subsequent searches sent to that primary see the change. Replica searches are eventually consistent, and a query spanning shards does not gain a single global transaction snapshot.

Those boundaries should shape acceptance tests. Update a catalog item, search immediately through the actual client routing configuration, and check both the item and its aggregated category count. Repeat while a replica lags and while a query touches multiple shards.

Before adopting the new query surface, inventory which fields need exact matching and which need text matching. Build a fixture with punctuation, missing numeric values and recently changed records, then compare expected results before tuning throughput.