Skip to Content
SettingsAI Providers

AI Providers

AI Providers store the API credentials and model configuration used by the AI assistant and AI-powered custom detectors. Each provider entry specifies a model, an API key, and optional settings depending on the provider type.


Supported providers

ProviderDefault modelBase URL
OpenAI-Compatiblegpt-4oConfigurable — use for OpenAI, Azure OpenAI, Together AI, OpenRouter, Groq, etc.
Claude (Anthropic)claude-sonnet-4-5Fixed (Anthropic API)
Gemini (Google)gemini-2.0-flashFixed (Google AI API)

Configuration fields

FieldDescription
NameA human-readable label (e.g. “Production Claude”, “Dev OpenAI”). Max 120 characters.
ProviderOne of OpenAI-Compatible, Claude (Anthropic), or Gemini (Google). Determines which API endpoint is used.
ModelThe model identifier (e.g. gpt-4o, claude-sonnet-4-5, gemini-2.0-flash). Max 200 characters.
API keyThe API key for the provider. Stored encrypted at rest. Only a masked preview is shown after saving (first 4 + last 4 characters).
Base URLOnly for OpenAI-Compatible providers. The full URL of the API endpoint. Ignored for Claude and Gemini.
Context sizeOptional. The context window size in tokens (e.g. 200000).
Vision supportOptional. Whether the model accepts image and PDF input.

Base URL behaviour

  • For OpenAI-Compatible providers, the base URL is a required configuration field (e.g. https://api.openai.com/v1 for OpenAI, or a custom endpoint for self-hosted or third-party providers).
  • For Claude and Gemini, the base URL is fixed to the provider’s API and cannot be overridden.

Managing providers

Create

  1. Click New Provider on the AI Providers tab.
  2. Enter a name, select the provider type, and fill in the configuration fields.
  3. Optionally click Test connection to validate the credentials.
  4. Save — the API key is encrypted before storage.

Update

Edit any field. To clear the API key, submit an empty value. Changes take effect immediately.

Delete

A provider can be deleted unless it is:

  • Selected as the instance-wide AI assistant model
  • Referenced by one or more custom detectors

In either case, reassign or delete the dependent resources first.


Testing a connection

The Test connection button sends a small structured JSON round-trip to the provider:

system: "Reply with valid JSON only."
user:   "Return JSON with fields: status, square of 7, language used."

A successful test returns the provider name and model. A failure shows the error message returned by the provider.

The configuration is saved before the test runs, so a passing test confirms that the stored credentials are valid.


API keys

  • API keys are encrypted at rest using the application’s encryption service.
  • Only the first 4 and last 4 characters are ever displayed (e.g. sk-p...xyz4).
  • Keys shorter than 8 characters are fully masked.
  • The raw key is never returned by the API after creation.

AI assistant default

One provider is selected as the default AI assistant model on the General settings tab. This provider is used for all AI assistant interactions across the instance. When a new provider is created from the General tab, it is automatically set as the default.

Last updated on