Overview

In this topic you find general information that can be useful to know before adding AI models. For example, information about prerequisites, a summary of the initial settings required for each model, and some additional options for already added models.

You can manage AI models in two ways:

• Using the Spotfire server administration pages (Configuration > AI models): This can be convenient when you want to add a single provider endpoint, validate settings interactively, when you need to adjust access permissions, or inspect history.

  See Adding and configuring AI models using the web administration pages for details.

• Using command-line commands: This can be particularly useful when you want to automate creation or updates, import and export model definitions between environments, or need to set several advanced options in one repeatable command.

  See AI model configuration commands, for details about the available commands, and Adding and configuring AI models using commands for some example usages.

Before you begin

• You must have the AI Services Configurator role. See System groups to learn more about roles.
• Verify that the Spotfire Server can reach the provider endpoint over HTTP or HTTPS.
• Collect the provider-specific endpoint URL, model identifier, and authentication details.
• Decide whether access to the model should be open to everyone who can use AI features, or restricted to specific groups or users.
• If you plan to move configurations between environments, decide which settings will be recreated manually and which settings will be scripted.

AI model configuration settings

When you add a new AI model, you must specify the following settings:

• Name: The display name shown on the AI models page in the Server administration pages.
• Model identifier: The provider-side model name or deployment name, for example gpt-4o or claude-3-5-sonnet.
• API: The provider type. Supported values are Azure OpenAI, Azure Anthropic, Azure Mistral, Anthropic, Gemini, Mistral, OpenAI, Amazon Bedrock, and Ollama.
• Mode: The interaction mode. The supported value is Completion.
• Endpoint: The provider base URL. For example, https://api.openai.com/.
• Authentication method: The authentication type used to call the AI provider. Supported values are API Key or None.
  • API key: The API key used for authentication if API key is the selected authentication method.
  • Authorization header mode: Controls how the authorization header is constructed. Supported values are Default, Authorization Bearer, or Custom.
• Path override: An optional client-facing path override. When set, clients use this path instead of the default when constructing gateway requests.

For more details on the settings, and for examples of configurations, see AI model configuration reference.

Additional options

Apart from the initial settings, you can manage existing models in various ways.

For example:

• Enable or disable a model.
• Change model priority.
• Change access rights to a model for specific groups and users.
• Add custom HTTP headers.
• Review change history.

See Enabling, disabling, and prioritizing models, Granting access to groups and users, Reviewing change history, and update-ai-model to learn more.