The AI Web API is an interface to access configured AI providers, skills, request execution, resource upload, and an OpenAI-compatible chat completions proxy.

Requests have to be sent as HTTP requests with the scheme https://[HOST]:[PORT]/api/ai/[Command]. To access the AI Web API a user has to have the required permissions - which have to be set in the configuration manager. All clients have to be authenticated in order to use the API.

Note: The AI Web API can be used only by users who have both the Web API and ai.api permissions set. All clients must be authenticated.

The AI Web API exposes the following features:

• Listing configured AI providers and resolving the default provider
• Running synchronous AI requests against a provider
• Uploading resources to providers that support uploads
• Listing configured AI skills
• Proxying OpenAI-compatible chat completions to the default AI provider

In this API, providerId is the configured provider display name. The display name must be unique.

• If providerId is omitted or empty, the default provider is used. The default provider is the first configured provider.
• Path and query tokens are URL-decoded. Matching is case-insensitive.
• Tokens also accept a hyphenized form where spaces are replaced by -.
• For GET /api/ai/providers/default, the literal path token default resolves the default provider.

The following commands are available for the AI Web API: