Securing Generative AI Deployments with NVIDIA NIM and NVIDIA NeMo Guardrails

Defining the use case

This example showcases how to intercept user questions related to personal data using topical rails.

Setting up a guardrailing system with NIM

1. Ensure NeMo Guardrails library is up to date.
2. Add dialog rails in the flows.co file.
3. Configure guardrails with Colang and YAML files in the config folder.

Testing the integration

• Greet the LLM NIM through guardrails.
• Verify that the guardrails intercept queries related to personal data.
• The guardrails successfully block the LLM NIM from responding to the query.

By integrating NIM microservices with NeMo Guardrails, developers can deploy AI models securely and efficiently. Check out the full tutorial notebook on the NVIDIA generative AI examples GitHub repository.