Jupyter AI installation - Amazon SageMaker AI

Jupyter AI installation

To use Jupyter AI, you must first install the Jupyter AI package. For Amazon SageMaker AI Distribution users, we recommend selecting the SageMaker Distribution image version 1.2 or later. No further installation is necessary. Users of JupyterLab in Studio can choose the version of their Amazon SageMaker Distribution when creating a space.

For users of other IPython environments, the version of the recommended Jupyter AI package depends on the version of JupyterLab they are using.

The Jupyter AI distribution consists of two packages.

  • jupyter_ai: This package provides a JupyterLab extension and a native chat user interface (UI). It acts as a conversational assistant using the large language model of your choice.

  • jupyter_ai_magics: This package provides the IPython %%ai and %ai magic commands with which you can invoke a large language model (LLM) from your notebook cells.

Note

Installing jupyter_ai also installs jupyter_ai_magics. However, you can install jupyter_ai_magics independently without JupyterLab or jupyter_ai. The magic commands %%ai and %ai work in any IPython kernel environment. If you only install jupyter_ai_magics, you can't use the chat UI.

For users of JupyterLab 3, in particular Studio Classic users, we recommend installing jupyter-ai version 1.5.x or any later 1.x version. However, we highly recommend using Jupyter AI with JupyterLab 4. The jupyter-ai version compatible with JupyterLab 3 may not allow users to set additional model parameters such as temperature, top-k and top-p sampling, max tokens or max length, or user acceptance license agreements.

For users of JupyterLab 4 environments that do not use SageMaker Distribution, we recommend installing jupyter-ai version 2.5.x or any later 2.x version.

See the installation instructions in the Installation section of Jupyter AI documentation.