Developing Packages#

Steamship is the fastest way to build and deploy a full-stack language AI package.

Here’s how to go from zero-to-deployed in about 60 seconds. Then you can customize your new package from there.

First, create a new package with our CLI.

ship create

> Select *Package*
> Select *Empty Package*

Then deploy the package to the cloud:

ship deploy

Now create a new instance of your package. If you used the Empty Package template, you can use the CLI to create one:

ship package:instance:create --default_name="Beautiful"

That keyword argument above is part of the required configuration. You can see where that argument is defined in the src/ file of your new package, and you can see where it is required in the steamship.json file.

The response you get back will contain your Instance Handle.

Let’s invoke a method on that instance. Replacing your Instance Handle below, run:

ship package:instance:invoke --instance="INSTANCE_HANDLE" --method="greet"

Now let’s create an instance and invoke it from Python. After running pip install steamship, run the following code, replacing your package and instance handles:

from steamship import Steamship

# TODO: Replace with your package and instance handle below
instance = Steamship.use("PACKAGE_HANDLE", "INSTANCE_HANDLE", config={
    "default_name": "Beautiful"


That’s it! You’ve just created a new package, deployed it, and invoked it from two different environments.

Now the real fun begins…

Customizing your Package#

Steamship packages run on top of a cloud stack designed for Language AI. You can import files, parse and tag them, query over them, return custom results – and far more.

Package Project Structure

Understand the basic project structure that defines a package and how to develop in it.

Package Cookbook

A cookbook of common operations you may want to perform, such as parsing files, transcribing audio, and querying results.

Developer Reference

Lower-level details on Steamship development, such as environment setup, configuration, testing, and secret management.