Targets

A target represents an AI system you want to test. Scanner supports API-based models and browser-based chat UIs.

Creating a Target

1. Navigate to Targets in the sidebar
2. Click New Target
3. Fill in the target configuration (see below)
4. Click Create Target

Target Configuration Fields

Field	Required	Description
Name	Yes	Display name for this target
Model Type	Yes	The garak generator class (e.g., rest.RestGenerator)
Model	Varies	Model endpoint URL or model identifier
JSON Config	No	Additional generator options as JSON

Common Target Templates

OpenAI (GPT models)

Field	Value
Model Type	openai.OpenAIGenerator
Model	gpt-4o (or any OpenAI model ID)

Set OPENAI_API_KEY as a target-specific environment variable (see Environment Variables).

OpenRouter

Field	Value
Model Type	openai.OpenAIGenerator
Model	openai/gpt-4o
JSON Config	{"api_base": "https://openrouter.ai/api/v1"}

Set OPENROUTER_API_KEY.

Anthropic (Claude via LiteLLM)

Field	Value
Model Type	litellm.LiteLLMGenerator
Model	claude-3-5-sonnet-20241022

Set ANTHROPIC_API_KEY.

Azure OpenAI

Field	Value
Model Type	azure.AzureOpenAIGenerator
Model	Your deployment name
JSON Config	{"api_base": "https://YOUR_RESOURCE.openai.azure.com/"}

Set AZURE_API_KEY.

Ollama (Local Models)

Field	Value
Model Type	ollama.OllamaGenerator
Model	llama3.2 (or any Ollama model name)

No API key needed. Ollama must be accessible from within the Docker network. If running Ollama on your host machine, use http://host.docker.internal:11434 as the base URL in your JSON config.

Hugging Face Inference API

Field	Value
Model Type	huggingface.HFInferenceAPIGenerator
Model	mistralai/Mixtral-8x7B-Instruct-v0.1

Set HF_TOKEN.

Generic REST API

For any OpenAI-compatible HTTP endpoint:

Field	Value
Model Type	rest.RestGenerator
Model	http://your-api-host/v1/chat/completions

Mock LLM (Testing)

Field	Value
Model Type	rest.RestGenerator
Model	http://mock-llm:9292/api/v1/mock_llm/chat

The Mock LLM ships with the Docker Compose setup and requires no API keys. See Mock LLM for details on available modes.

Per-Target API Keys

Each target can have its own set of environment variables (API keys) that override any global variables with the same name. This lets you:

• Use different API keys per target
• Isolate credentials for different environments
• Test the same model with different authentication contexts

To set a per-target API key:

1. Open the target's detail page
2. Scroll to Environment Variables
3. Click Add Variable
4. Set the variable name (e.g., OPENAI_API_KEY) and value

See Environment Variables for the full list of supported variable names.

Webchat Targets

Webchat targets use Playwright browser automation to interact with chat UIs that don't expose a direct API.

info

Webchat scanning requires additional Playwright setup. This feature is marked experimental in the current release.

Managing Targets

• Edit: Click a target name to view and edit its configuration
• Delete: Remove a target from the target detail page
• Test connection: Not yet available — run a scan against the Mock LLM first to validate your setup

note

Target configurations (including JSON config) are encrypted at rest using per-tenant encryption keys derived from SECRET_KEY_BASE.