Profound Logic

IBM watsonx.ai Models

Verified

See Model Configuration for a full list of available model properties.

Overview

Use the procedure outlined below to configure and use IBM watsonx.ai models within Profound AI.

Selecting a Model to Use

IBM watsonx.ai provides various models to choose from. The model you select should be based on the capabilities you intend to use within Profound AI:

Granite models are best suited for fast backend / batch processing via the Ask AI Agent plugin.
For an interactive chat experience that includes Data Access, Data Analytics, and custom Agent Routines, use a larger model such as mistralai/mistral-large.

Install IBM watsonx.ai Node.js SDK

From the command line within your Profound AI installation directory, issue the following command:

npm install @ibm-cloud/watsonx-ai

Steps

Login to IBM watsonx.ai.
Create or select a project.
In the project, generate a new API key.
Edit your Profound AI configuration file to add a model with a watsonx provider property.
1. If apiVersion and endpoint properties are not specified in the configuration, default IBM watsonx.ai values will be used for those properties.

Example

The example below shows a sample configuration for the mistralai/mistral-large IBM watsonx model:

models: {
  "IBM watsonx": {
    provider: "watsonx",
    model: "mistralai/mistral-large",
    apiKey: "...",
    projectId: "....",
    stream: true
  }
}