Integrating LLMs with Pyramid

You can integrate Large Language Models (LLMs) with Pyramid to enhance Pyramid's data and business analysis capabilities and to automate your analysis and publication processes. Integrated LLMs can be used to build visuals using natural language prompts, generate explanations of visualized data that can be shown on your presentations or in your publications, generate forecasts and other analytic data, and many other time-saving and analytic functions.

This help topic guides you through the process of deciding which LLM you might want to integrate with Pyramid to best satisfy your requirements; and the steps that you need to perform to complete the integration process.

  • Click here for information about Pyramid's LLM features

Choosing an LLM

When selecting your LLM you should consider which provides the best functionality for you. You might want to think about your familiarity with the toolsets, the accuracy associated with the LLM, any considerations around language support, or privacy. Once you have selected an LLM you need to set it up ready for use.

Set up your LLM

Before you begin the set up process, you should ensure that you have internet access and that https is enabled.

Watsonx

If you want Pyramid to use Watsonx as its underlying LLM, you will need the following:

  • You need Internet access to allow interaction with the service.
  • You will be sent an LLM license from Pyramid that enables IBM Watsonx for use with Pyramid.
  • The minimum requirement for your model is Mistral-large.

WatsonX needs to be licensed separately. Click here for more or contact sales.

OpenAI

If you want Pyramid to use OpenAI as its underlying LLM, you will need the following:

  • You need Internet access.
  • You need an OpenAI account.
  • You need to generate a Secret Key for creating a connection in the LLM Manager.
  • The minimum requirement for your model is GPT 4 (gpt-4-1106-preview).

Azure AI

If you want Pyramid to use Azure AI as its underlying LLM, you will need the following:

  • You need Internet access.
  • From the Azure OpenAI studio, you need to create a deployment.
  • You need an OpenAI account.
  • You need to capture the Deployment Name, Endpoint, and generate an API key for creating your connection in the LLM Manager.
  • The minimum requirement for your model is GPT 4 or higher.

Mistral

If you want Pyramid to use Mistral as its underlying LLM, you will need the following:

  • You need Internet access.
  • You need to be able to access the Mistral Console.
  • You need to generate an API Key for creating a connection in the LLM Manager.
  • The minimum requirement for your model is mistral-large-latest.

Google Gemini

If you want Pyramid to use Google Gemini as its underlying LLM, you will need the following:

  • You need Internet access.
  • You need a Google account.
  • From the Cloud console, you need to set up authentication and an API client.
  • You need to generate a Client ID and Client Secret and, from the datasource details, copy the Project ID. You will use these values when creating a connection in the LLM Manager.
  • The minimum requirement for your model is gemini-1.5-pro-001 (this value is pre-filled by default and cannot be edited).

Adding LLM providers

Once you have created a deployment or accessed a service, you need to set up the LLM in the LLM Manager. This provides Pyramid with the details to connect to the LLM.

  • Click here to learn how to set up an LLM in Pyramid

Global settings (Redirect URL)

You must always set the Redirect URL in the Global Settings, to allow Pyramid to know where any responses will be returned from.

  • Click here for more details about global settings