NeMo Guardrails Deployment Guide#

Deploy the NeMo Guardrails microservice as a standalone component of the NeMo platform with minimal dependencies, without requiring other functional microservices.


Deployment Options#

Choose one of the following deployment options based on your use case.

Helm Chart

Deploy the microservice using the Helm chart for production environments.

NeMo Guardrails Basic Installation Using Helm
Docker Compose

Quickstart using Docker Compose for local experimentation, development, testing, or lightweight environments.

Deploy NeMo Guardrails with Docker
Deploy Guardrails on GKE

Deploy Guardrails extension for GKE Inference Gateway.

Deploy NeMo Guardrails on GKE with Inference Gateway Integration

Deployment with External LLM Endpoints#

Review deployment options for the microservice to connect with external LLM endpoints.

Configure with LLM NIM Microservice in Cluster

Deploy the microservice using the LLM NIM microservices you have deployed in your cluster separately outside of the NeMo platform.

Integrate NeMo Guardrails with LLM NIM Outside of NeMo Platform
Configure with LLM NIM from build.nvidia.com

Deploy the microservice using the LLM NIM endpoints from build.nvidia.com.

Integrate NeMo Guardrails with NIM from build.nvidia.com

Advanced Configurations#

Review advanced deployment and configuration options for the microservice.

High Availability

Deploy the microservice with high availability by adding configurations for multiple replicas and affinity rules.

High Availability Configuration
Configuration Store

Use an external database for NeMo Guardrails configuration store.

Guardrails Configuration Store
Custom Dependencies

Extend the NeMo Guardrails microservice image with additional dependencies and use it in the Helm installation.

Install Custom Dependencies in NeMo Guardrails
Observability

Integrate with NeMo Guardrails observability.

Enable Observability for NeMo Guardrails