ClickHouse vs Alopex DB

What ClickHouse is

ClickHouse is a columnar database for analytics. It can also search vectors with SQL.

Vector storage: vectors go in Array(Float32/Float64/BFloat16) columns.
Search: you search with SQL ORDER BY distance_function LIMIT N.
Approximate search: uses the vector_similarity index (25.8 and later). The method is HNSW.

Columnar analytical SQL combines with approximate vector search.

The vector index is loaded into memory in full for searches, and it is built in memory. The default cache is 5 GB.
The documentation says HNSW indexes take long to build and can slow INSERTs and merges significantly, so they are ideal only when data is immutable or rarely changes.
Parts that contain deleted rows skip the index and scan every row.
With pre-filtering, where WHERE conditions are evaluated first, search falls back to brute force without the index. With post-filtering, fewer rows than LIMIT can come back.
A LIMIT above max_limit_for_vector_search_queries (default 100) raises an error.
Existing data does not get the index automatically; you run MATERIALIZE INDEX.

Both handle columnar data and vector search with SQL.

Alopex DB handles columnar tables, row tables updated row by row, and vectors in the same core. It works from a single embedded file.

Alopex DB handles aggregation, app reads and writes, and vector search in one place. Give it a try with the quickstart.