IBM watsonx.ai Flows Engine

Connect to watsonx.ai

  • Overview
  • Installation
  • Authentication
  • Getting Started
  • Learn
    • Using the CLI
    • Building flows
    • Connect Vector Databases
    • Consume flows
    • Connect to watsonx.ai

Connect to watsonx.ai

To explore with watsonx.ai Flows Engine you don't have to set up your own account for watsonx.ai (or other LLM providers) . Instead, watsonx.ai Flows Engine instance can connect to models from watsonx.ai out-of-the-box.

However, when you want to make sure your data stays confidential or bring your flows into production you will need to link your own watsonx.ai instance.

  • Connect your watsonx.ai instance from the Flows Engine dashboard (Python CLI only):

    1. Open the watsonx.ai Flows Engine dashboard and scroll to the bottom of the page.

    2. Select "Connect" under "watsonx.ai instance" and follow the steps. You will be prompted to connect to your IBM Cloud account to retrieve the watsonx.ai credentials.

    3. Select your IBM Cloud account and watsonx.ai project (make sure the project has access to

      ).

    4. Run the generated

      command with your own unique code in the terminal, your watsonx.ai and Flows Engine accounts are now linked. You don't have to include any other values in the
      file of you project before running
      .

  • Manually set the credentials of your own watsonx.ai instance:

    1. Open the Developer Access page in the watsonx.ai dashboard.

    2. Select the project you want to use, and copy the

      and
      (from the "watsonx.ai URL" without
      ).

    3. Retrieve your IBM Cloud API Key using the link at the bottom of the page.

    4. Add the following credentials to the

      file in your project:

      STEPZEN_WATSONX_HOST=us-south.ml.cloud.ibm.com
STEPZEN_WATSONX_PROJECTID=<YOUR WATONSX.AI PROJECT ID>
STEPZEN_WATSONX_AI_APIKEY=<YOUR IBM CLOUD API TOKEN>

    5. After adding these values, make sure to redeploy your project by running:

      wxflows deploy # Node.js
wxflows-py deploy # Python

That's it, the requests will now be routes against your own watsonx.ai instance. This includes both requests to completion and embedding models.

Need help?

Please go to our Discord if you need help or have any other questions about using watsonx.ai Flows Engine.