DuckDB · ClickHouse · Cassandra

The remaining Free-tier SQL engines. Each gets the same baseline (schema browser, edit-as-draft, native backup, schema diff) tuned to the engine’s catalog + dialect quirks.

DuckDB

In-process analytical engine. Connection picks a .duckdb file (or :memory:).

• Read-mostly workloads — DuckDB’s sweet spot is “I have a 5GB Parquet file and need to OLAP it”. Quay’s grid handles wide row results well.
• Native Parquet/CSV — CREATE TABLE x AS SELECT * FROM read_parquet('…') is one click in the schema browser
• Backup — DuckDB’s EXPORT DATABASE directive is used for multi-file dumps; Quay’s single-file .sql path handles smaller databases via standard CREATE TABLE + INSERT.
• Engines that wire-compat to it — none currently; DuckDB is its own thing (MotherDuck cloud-hosted DuckDB has its own dialect preset, planned for v0.4).

ClickHouse

OLAP column store. HTTP wire protocol (port 8123) + native protocol (9000); Quay supports HTTP for portability.

• Connection — host / port (8123 by default) / user / password / database; HTTPS optional
• MergeTree introspection — partition + part list per table; the engine deep panel shows part counts, merge backlog, replication state
• Materialized views — surfaced as first-class objects in the schema browser; the dependency graph shows source table → MV
• Backup — Quay emits CREATE TABLE + INSERT INTO … VALUES statements; restore preview validates engine clause (MergeTree PARTITION BY …) syntax before running

ClickHouse’s ALLOW DESCRIPTION_PRIVILEGES quirk: anonymous / read-only users may have visible system.* tables but not SHOW TABLES on user databases. Quay’s error message is the server’s exact text + a hint to check the user grants.

Cassandra / ScyllaDB

Wide-column, query-language is CQL.

• Connection — single contact point; Quay discovers the cluster topology automatically. User / password (CQL auth) or certificate-based.
• Keyspace browser — keyspace → table → partition key + clustering key visualisation
• CQL editor — separate from the SQL editor (different syntax); syntax highlighting + autocomplete tuned to CQL
• ALLOW FILTERING warnings — every SELECT … WHERE that doesn’t use the partition key prefix gets a yellow banner above the result grid: “this query uses ALLOW FILTERING — full-cluster scan, will scale poorly”
• Prepared statement helper — frequently-run queries can be pinned + reused without re-preparing

ScyllaDB shares Cassandra’s wire protocol; the connection preset toggles a few system tables (system_schema_versions only, no system_traces for older Scylla).

Why grouped here

These three are SQL-family in the sense that they all ship with the same Quay surfaces (browser / editor / backup / diff) and none of them needs Pro tier to use. They’re grouped on one page because their per-engine specifics are short — full pages would be 50% boilerplate.

If you connect to one of these and find a missing surface that the PG/MySQL pages document, file it via Settings → Feedback.