Guardrail Tutorials#

Use the following tutorials to learn how to accomplish common guardrail tasks using the NeMo Guardrails API.

Tip

The tutorials reference NMP_BASE_URL, whose value depends on the ingress in your cluster. Before starting a tutorial, complete the ModelProvider setup below.

Prerequisite: Configure a ModelProvider#

The tutorials use NIMs hosted on build.nvidia.com.

Note

The platform pre-configures a system/nvidia-build model provider during startup. This provider routes inference requests to models hosted on build.nvidia.com using the API base URL https://integrate.api.nvidia.com and the NGC API key with Public API Endpoints permissions provided during deployment (automatically saved as the built-in system/ngc-api-key secret).

You can verify this provider exists by running nmp inference providers list --workspace system.

The tutorials in these docs use this provider for inference, but you can alternatively create your own and use it instead.

Once the ModelProvider is configured, use Model Entity references (workspace/model_name format) as the model in your guardrail configurations. Internally, the Inference Gateway service routes requests to the Model Provider. Refer to Model Routing for more details.

guardrails_config = {
    "models": [
        {
            "type": "content_safety",
            "engine": "nim",
            "model": "system/nvidia-llama-3-1-nemotron-safety-guard-8b-v3",
        }
    ],
    # ... rest of config
}
Deploy NemoGuard NIMs

Deploy NemoGuard NIMs in your environment

Deploy NemoGuard NIMs
Improving Content Safety with NemoGuard NIM

Use Content Safety checks to detect and block harmful content

Improving Content Safety with NemoGuard NIMs
Running Rails in Parallel

Configure parallel rails for input and output guardrails.

Executing Input and Output Rails in Parallel
Adding Safety Checks to Multimodal Data

Safety checks for multimodal data with NeMo Guardrails API.

Adding Safety Checks to Multimodal Data
Detecting Injection Attacks

Configure checks for SQL, XSS, template, and code injection.

Detecting Injection Attacks with Guardrails