HotSQL
The SQL you know, with search built in
Postgres-familiar SQL with vector, full-text, and geospatial search built in. PostgreSQL, DuckDB, Snowflake, and ClickHouse queries run as written.
Example query
-- Join a Postgres connection with an instant database
SELECT o.order_id, o.total, a.segment
FROM postgres.public.orders AS o
JOIN default.public.accounts AS a
ON a.id = o.account_id
WHERE o.status = 'open'
ORDER BY o.total DESC
LIMIT 20;Write in the dialect you know
Set the dialect, and Hotdata translates the query before it runs.
| dialect | Language |
|---|---|
| hotsql | HotSQL |
| postgres | PostgreSQL |
| duckdb | DuckDB |
| snowflake | Snowflake |
| clickhouse | ClickHouse |
A Snowflake query, sent as is
hotdata query --dialect snowflake "
SELECT IFF(amount > 100, 'big', 'small') AS bucket
FROM orders"One query across every source
- Every connection in one namespace
- Every connected database and instant database shares one namespace, so a cross-system join is one query.
- Pull Data →
- Reads and writes
- INSERT, CREATE TABLE, CREATE TABLE AS SELECT, and DROP TABLE, straight from SQL.
- Writing data with SQL →
- A full function library
- Math, string, date, aggregate, window, array, JSON, spatial, and vector functions.
- Functions →
- See how a query runs
- EXPLAIN shows the plan. The information schema lists every table and column.
- EXPLAIN →
Search and geospatial in the same query
Performance
5× faster than Snowflake and 9× faster than ClickHouse, with 10 clients querying at once.
p95 query latency
- 1
Hotdata
4 vCPU · 16 GiB
0.38 s
p50 0.17 s · 51.1 q/s
- 2
Databricks
~8 vCPU · ~64 GiB · Pro
1.18 s
p50 0.76 s · 12.4 q/s
- 3
Snowflake
~8 vCPU · 16 GiB
1.74 s
p50 0.85 s · 10.9 q/s
- 4
ClickHouse
4 vCPU · 16 GiB
3.39 s
p50 1.40 s · 6.4 q/s
Get started
Install the CLI, connect a source, and run your first query in minutes.
Run a query from the CLI
hotdata query "
SELECT order_id, total
FROM postgres.public.orders
WHERE status = 'open'
LIMIT 100"