LawVu Draft uses Microsoft Azure GPT-4 by default.
However, administrators can add additional Large Language Models (LLMs).
In this article:
Add a new LLM
Azure GPT-4 configuration
Setting the Default LLM (Priority Order)
Configuring LLM availability
Add a new LLM
Go to Administration > LLMs tile
Click + Add new LLM.
Complete the form with the required details.
Use the following provider identifiers
Anthropicfor Claude modelsOpenAifor OpenAI GPT modelsazurefor Azure GPT models
When these identifiers are used, LawVu Draft automatically applies some configuration settings.
Azure GPT-4 configuration
If the Provider is set to azure:
The endpoint you enter will automatically be extended with:
/openai/deployments/[model]/chat/completionsIf you enter a Query param string, it will be added to the endpoint URL.
If no query parameter is provided, LawVu Draft will automatically add a recent Microsoft API version (for example:
?api-version=2024-03-01-preview).Your API key is securely stored in encrypted form on LawVu Draft servers. For security reasons, it will not appear again in the editing field after saving.
Setting the Default LLM (Priority Order)
If multiple LLMs are available, administrators can define which model is selected by default.
Using Edit priority order, you can arrange LLMs in a preferred sequence. For example, the order might be:
Claude 3 Opus (Large)
Azure GPT-4
In this configuration, Claude 3 Opus will be pre-selected in AI modules that support both models. If a user clicks Draft without changing the selection, Claude will automatically be used.
Configuring LLM availability
Each LLM (including the default LawVu Draft models, shown in italics) can be configured so that it is only available in specific AI modules, such as Compose Document or Doc Chat.
LawVu Draft determines whether a user can use a particular LLM through two checks: user access and activity compatibility.
First, the system checks whether the LLM is available to the user. When a new LLM is added, it is available to all users by default. Administrators can limit access by enabling the Restricted option in the LLM settings and assigning the model to one or more user profiles.
Next, LawVu Draft checks what the user is currently doing. The platform only displays LLMs that are enabled for that specific activity, such as drafting, reviewing, or summarizing a document.
For example, imagine the system includes three LLMs: GPT-4o, Claude, and Noxtua. Claude might be configured so that it is only available to the Senior Lawyer and Partner profiles, and only for reviewing tasks. In this case, a Senior Lawyer would see Claude when reviewing a document, but not when using the Summary module.