Open source · Apache 2.0

Download Sail

The multimodal Rust-native engine for analytics and AI workloads. Bring your Spark SQL and DataFrame code.

pip install pysail

Run your first query

Prebuilt for Linux, macOS, and Windows on PyPI.

Start the Sail server

sail spark server
INFO Starting the Spark Connect server on 127.0.0.1:50051...

It listens on localhost:50051. Leave it running and open a second terminal in the same environment.
02
Connect from Python

pip install pyspark-client==4.2.0
```
from pyspark.sql import SparkSession

spark = SparkSession.builder.remote("sc://localhost:50051").getOrCreate()
spark.sql("SELECT 1 + 1").show()
```
+-------+|(1 + 1)|+-------+| 2|+-------+
Already using Spark Connect? Skip the install and point your session at sc://localhost:50051.

What is Sail?

Spark-compatible

Your Spark SQL and DataFrame code connects to Sail with the standard PySpark client, over Spark Connect.

Built in Rust

Sail is a multimodal Rust-native engine. Delta Lake and Iceberg are built in, with no JVM to install or tune.

Laptop to cluster

Run it as one process on your machine, or deploy it on Kubernetes as a distributed cluster.

Multimodal data for AI

Turn images, video, audio, PDFs, and robot data into training data and embeddings, with SQL and Python.

Go further

Getting started guide → Spark 3.5, the PySpark shell, notebooks, and running Sail inside Python.Deployment guide → Run Sail in a Docker container or on Kubernetes.Release notes → What changed in each Sail release.GitHub → Source code, issues, and contributions.

Rather not run it yourself? LakeSail runs Sail for you in your AWS account. Try LakeSail or talk to an engineer.