AI Configuration

Turbot AI Configuration

In this guide, you will learn how to:

• Configure various parameters in Turbot > AI > Configuration policy

This guide is the starting point for all AI-powered features, including Intelligent Assessment Control, Intelligent Fixes, and Policy Pack Summary. Before using any AI capabilities in Guardrails, ensure you have completed the steps outlined in this guide.

Prerequisites

• Turbot/Admin permissions at the Turbot resource level.
• Familiarity with the Guardrails console.
• Access credentials for the AI model.

Step 1: Find Turbot > AI Policy

Log in to the Guardrails console using your local credentials or via a SAML-based login. From the Policies tab, navigate to Turbot > AI in the Guardrails console.

This section provides AI-related policies that allow Guardrails to govern the use of AI for various configurations and settings. Select Configurations.

Step 2: Configure AI Provider

Select Turbot > AI > Configuration > Provider [Default] to configure the AI service provider. The default value is set to openai.

In this example, anthropic is selected.

Step 3: Configure API Key

Select Turbot > AI > Configuration > API Key [Default] policy under the Turbot > AI > Configuration main policy.

Enter the API key for the provider chosen in Step 2. Guardrails will use this key to authenticate with your AI service provider.

[!IMPORTANT]

• The API key is a sensitive credential used for authentication with your chosen AI service provider. Ensure it is stored and handled securely.
• For OpenAI:
• Obtain the API key from the OpenAI platform
• The key should start with "sk-"
• For Anthropic:
• Get the API key from the Anthropic console
• The key should start with "sk-ant-"
• Store the API key securely and rotate it periodically according to your organization's security policies.
• Never share or expose the API key in logs, code repositories, or public forums.

Step 4: Configure AI Model

Select Turbot > AI > Configuration > Model [Default] policy under the Turbot > AI > Configuration main policy.

Choose the model to use with the selected AI provider for processing requests.

[!NOTE] Supported and tested models include:

• OpenAI models: gpt-5, gpt-4.1, gpt-4.1-mini, gpt-4o
• Anthropic models: claude-opus-4-1-20250805, claude-sonnet-4-20250514, claude-3-7-sonnet-20250219, claude-3-5-haiku-20241022, claude-3-5-sonnet-20241022

For optimal results:

• Use the latest model versions when possible.
• Higher-tier models (like GPT-5 series and Claude Sonnet/Haiku series) provide more accurate and detailed assessments.
• Lower-tier models may produce different or less consistent results.
• Model availability may vary based on your AI provider subscription.
• For more information about the models, see the documentation for OpenAI and Anthropic.

Step 5: Configure Max Tokens

Select Turbot > AI > Configuration > Max Tokens [Default] under the Turbot > AI > Configuration main policy. The default value is set to 1000 as a best practice.

[!TIP]

• Configure the maximum number of tokens to use for the AI API calls. This setting controls the length of the AI's response by limiting the number of tokens (words, subwords, or characters) that can be generated.
• A higher value allows for longer, more detailed responses but may increase API costs and response time.
• A lower value produces more concise responses.
• The default value of 1000 tokens provides a good balance for most use cases, but you can adjust this based on your specific needs and the complexity of the tasks you want the AI to perform.

You can adjust this based on your specific needs and the complexity of the tasks you want the AI to perform.

Step 6: Configure Temperature

The policy Turbot > AI > Configuration > Temperature [Default] allows you to configure the temperature parameter for AI API calls. Temperature controls the randomness or creativity of the AI's responses.

A lower value (closer to 0) makes responses more focused, deterministic, and consistent, while a higher value (closer to 1) makes responses more diverse and creative. The default value of 0.2 provides a good balance between consistency and creativity, suitable for most business and technical use cases.

The default value is set to 0.2. You may choose to update it based on your needs.

Note: The temperature parameter is no longer supported for OpenAI's GPT-5 series models; any value set for this parameter will be ignored.

Step 7: Enable Configuration

The Turbot > AI > Configuration > Enabled [Default] policy under Turbot > AI > Configuration plays an important role if you wish to apply configurations to all features under Turbot > AI. You may choose to continue with the default value Disabled and enable it in respective features separately.

To enable it, select New Policy Settings, set Resource to Turbot, and choose Enabled.

Review

• [ ] Validate that all the required settings are in place.

Next Steps

After completing this configuration, continue with the following guides to leverage AI features in Guardrails:

• Enable Intelligent Assessment
• Enable Intelligent Fixes
• Enable Policy Pack Summary
• Developing guardrails for Guardrails

Troubleshooting