LLM Configs

To access LLM Configs, navigate to Administration > Environment Settings > LLM Settings > LLM Configs.

LLM Configs define how an AI agent behaves: its system prompt, AI model, and response settings. Each config is stored on the server and referenced by name in Smart Components, like the AI Chat Agent. Administrators create and update configs here; Creators reference them through the Build Agent during Chat Agent setup.

Managing LLM Configs requires administrative permissions. Contact an environment administrator if you need access.

[docs-mockup:DocsAdminListCanvas] Administration panel overview

Config List

The config list displays all LLM configurations registered in the environment.

Column	Description
Service	The config name.
Description	A description of the config, if provided.
Auth Type	Identifies the record type as an LLM config.

Page Actions

Action	Description
Create Config	Opens a new page to create a new LLM config.

Config Actions

Select the (ellipsis) on any row to access the following actions.

Action	Description
Edit	Opens a new page to update the config's system prompt, model, or response settings. The config name cannot be changed after creation.
Delete	Permanently removes the config. This action cannot be undone.

Deleting a config that is referenced by a Smart Component, like the AI Chat Agent, causes that component to stop responding. Update the component before deleting the config.

Adding a Config

Navigate to Administration > Environment Settings > LLM Settings > LLM Configs and select Create Config.

Field	Description
Config Title	The display name for this config. Required.
Config Name	A unique identifier for this config. Only letters, numbers, and hyphens are accepted. The name cannot be changed after the config is created. Required.
System Prompt	The instructions that define the agent's role, tone, and behavior. The system prompt is sent with every request and is not visible to end-users. Accepts a maximum of 32,768 characters. Required.
Model	The AI model this config uses. When left blank, this config uses the default model for the selected LLM credential. Optional.
Thinking Level	Controls the depth of reasoning the model applies before generating a response. Options include: Model's default (no override), Off, Low, Medium, or High. Optional.
Max Tokens	The maximum length of each AI response, measured in tokens. Accepts values of 1 or above. Optional.
Thinking Budget Tokens	The reasoning token allowance for models that accept a budget instead of a level. Accepts values of 1 or above. Optional.
Temperature	The randomness of the AI's responses. Accepts values between 0 and 2. Lower values produce more consistent responses; higher values produce more varied ones. Optional.

Thinking Level and Thinking Budget Tokens apply only to AI models that support extended reasoning. Use one or the other based on the model's requirements, not both.

System Prompt

The system prompt defines the agent's role, knowledge boundaries, and tone. It is sent before every conversation and is not visible to end-users. The system prompt takes instructions in plain language that describe what the agent should do and what it should avoid.

Example:

You are a client intake assistant for a financial services firm. Help clients complete intake forms accurately. Do not provide legal or financial advice. Keep responses brief and professional.

The system prompt is stored on the server and is not exposed in the browser.

Changelog

Date	Change
2026-08-18	Initial publication.