Multiple LLM Engines

Natural Language Processing can be based on Pyramid's internal Portable Language Model (PLM) or customers can leverage more powerful third-party Large Language Models (LLMs). When you use the Chatbot to ask Pyramid questions and automatically generate discoveries, calculations, presentations, and publications, the underlying action depends on which language model interprets and completes your requests.

The option to pick one or more language models is the basis for Pyramid’s multi-pronged language model strategy.

Note: You can select to use different LLM engines for different Models. There is a separate option to pick a language model to drive the speech-audio interpreter.

LLM Engine selection

The engines that are available for use when sending NLQs to the Chatbot will always include the Pyramid Internal portable language model (PLM) and may include one or more other third-party LLM engines that your administrator has added to the LLM Manager.

Your administrator selects which LLM engines should be used by the Chatbot in the following locations in the Admin Console:

• In AI Settings, the administrator enables NLQ and sets the System Default Providers for the Chatbot, LLM, and Speech. The list of available third-party LLM engines is retrieved from the LLM Manager plus, in the case of the Chatbot only, Pyramid Internal.
• In Tenant AI Settings, the administrator sets up the Providers for each tenant or chooses to use the System Default.
• Optionally, in Data Model Management under NLQ, the administrator configures which providers to use for particular models.

The availability of multiple LLMs allows different capabilities to be made available for different models or tenants. It is important to note that the engines will also behave slightly differently to each other in response to prompts, since that is the nature of intelligence of the type that is in use.

Pyramid's Portable Language Model (PLM)

The Pyramid Internal engine is a "portable" language model (PLM). The built-in engine is very fast and does not need to be connected to external services to run; that is, it can run 'off the grid'. It also operates without any risk of data or insight sharing with external LLM providers.

The PLM engine enables the Chatbot at Design time in Discover and at Runtime in Discover and Present only.

However, Pyramid's PLM only supports English language "chats", and its syntactical interpretation capabilities are limited compared to its LLM cousins. As such, stronger guidance is provided to access its capabilities.

• Portable Language Model Guidance

Third-Party LLM engines

Third-party LLM engines can be included if they are set up and configured in the LLM Manager.

The third-party LLM engines all enable the Chatbot at Design time in Discover, Formulate, Present, and Publish, and at Runtime in Discover and Present. They include capabilities that are not available with the Pyramid Internal engine, such as multi-language support.

Note: You need a separate license to use the Chatbot in Formulate.