Setting Up JFrog ML

JFrog ML Documentation
Products
JFrog ML
Content Type
User Guide

This page describes how to set up JFrog ML, including installing and configuring FrogML CLI and SDK.

Self-Managed

For self-managed, follow the instructions in Activate AI ML.Activate AI ML

JFrog SAAS

To set up JFrog ML in SAAS: 

Important

Only a Platform Admin can perform the initial setup.

  1. Click on the AI/ML link in the left sidebar of your JFrog platform. This will open the JFrog ML setup form.

    JFrog ML Setup Form

  2. During setup, you'll need to select:

    1. A JFrog project where JFrog ML will store generated artifacts, such as Docker images and files.

    2. A cloud provider (currently AWS & GCP; Azure support).

    3. A cloud region for deployment.

  3. Click Get Started. JFrog ML is enabled in your JFrog account.

Install FrogML CLI

After your Admin has setup the JFrog ML on your account (SAAS or self-managed), you are ready to start using the JFrog ML solution. Within JFrog ML, the FrogML SDK & CLI are your gateway to building and deploying models and features on JFrog ML, seamlessly integrating with your existing JFrog platform.

Important

frogml-sdk supports Python versions 3.9 to 3.11. Make sure you install it using a compatible Python version.

To install frogml-cli: 

  1. Optional: Create a new virtual environment using your preferred Python framework, such as virtualenv, poetry or conda.

  2. Install the FrogML CLI in your virtual environment with one of the following commands:

    Using uv (recommended method, as this enables the CLI to be accessed from any environment)

    uv tool install frogml-cli

    Using pip / conda:

    pip install frogml-cli

    Using poetry:

    poetry add frogml-cli

Note

Installing the FrogML CLI also installs the FrogML SDK.

Configuring FrogML SDK and CLI

After installing the frogml-cli, you need to configure it. You have two main options, with several sub-options:

To configure via CLI:

Sub-option 1: Define access token: 

  1. With this option, you must first generate an access token in the Jfrog platform.Access Tokens

    configuring-qwak-sdk-jfrogml.png

  2. Open the terminal and type in the following command, replacing the variables:

    frogml config add --url={server_url} --access-token={access_token_from_artifactory} --server-id={optional_server_id}

    For example:

    frogml config add --url=https://mydemo.jfrog.io --access-token=token123 --server-id=testing-server

Sub-option 2: Only if you already have the JF_CLI installed and configured, use your username and password: 

For this option, you need to make sure you have enabled token generation API as follows:

  1. In the Administration module, select Security > General.

  2. Scroll down and click Enable token generation via api. under Basic Authentication.

  3. Click Save.

Then, type in the following command, replacing the variables:

frogml config add --url={server_url} --username={user_name} --password{password} --server-id={optional_server_id}frogml config add --url={server_url}

When configured correctly, you'll receive the following message:

Logged in successfully to: <PLATFORM_BASE_URL>

Note

There is an additional option, called "interactive". If you have previously configured the frogml-cli, it enables you to load the configuration. Alternatively, it enables your to connect to a new server using the access token or username and password, but in an interactive 'guided' manner.

interactive_config_via_cli.gif

To configure via environment variable:

JFrog ML also allows you to configure your session authentication key via environment variables. Note that this is only relevant for the individual Terminal session or until the next computer restart.

To set the environment variable, use the following command:

export JF_URL=<PLATFORM_BASE_URL>
export JF_ACCESS_TOKEN=<ACCESS_TOKEN>

Here you need to define the platform_base_URL and the access_token.

The same behavior occurs when you're running Python code that calls the FrogML SDK; it will read the environment variable for authentication.

Troubleshooting

Unrecognized frogml Command

Problem

Local system fails to recognize the frogml command.

Solution

Ensure it is added to your system's PATH environment variable.

Read/Write Permission Issues

Problem

Issues related to read/write permissions with the following error message:

Could not read user configuration from <home_path>/.frogml/config.json. Please make sure one has been set using frogml config command.

Solution

Manually create the .frogml directory in your home path and rerun the configure command:

mkdir -p <home_path>/.frogml

Additional Troubleshooting

For any other issues with installing and configuring your FrogML SDK/CLI please refer to Installation Issues or Performance Issues pages.