s3shift

Python client for s3shift — a truly serverless Redshift: Redshift SQL on DuckDB in a Lambda, over your own S3 and RDS data, ≈ $0 when idle. The redshift_connector API, standard library only. Part of Redefine AWS Serverless.

pip install s3shift

from s3shift import redshift_connector   # later: `import redshift_connector` against Redshift, nothing else changes

conn = redshift_connector.connect(host="xxx.lambda-url.us-east-2.on.aws", database="dev",
                                  user="admin", password="...")
cur = conn.cursor()
cur.execute("CREATE EXTERNAL TABLE clicks (user_id INT, day DATE) STORED AS PARQUET LOCATION 's3://my-lake/clicks/'")
cur.execute("SELECT day, count(*) FROM clicks WHERE day >= %s GROUP BY 1 ORDER BY 1", ["2026-01-01"])
cur.fetchall()            # ([datetime.date(2026, 1, 1), 1520], ...)
cur.fetch_dataframe()     # pandas, if installed

Every statement is committed when execute() returns; large results are fetched page by page inside execute(). Not supported (raises NotSupportedError): stored procedures, UDFs, users / GRANT, Glue Data Catalog schemas, materialized views.

The endpoint comes from deploying the main repository. Compatibility: REPORT.md.