Reference

Use Case Diagrams

View as Markdown

The following diagrams illustrate the NeMo Guardrails architecture for each use case.

1. Content Safety

NeMo Guardrails Content Safety Options Application LLM Input Content Check Output Content Check LLM Self-Check Llama 3.1 NemoGuardContent Safety NIM LlamaGuard Fiddler ActiveFence AI Defense

Content safety guardrails check both user inputs and LLM outputs for harmful content.

2. Jailbreak Protection

NeMo Guardrails Jailbreak Detection Options Application LLM Input Jailbreak Detection Self-Check Detection Heuristic Detection NemoGuard JailbreakDetection NIM Prompt Security Pangea AI Guard

Jailbreak protection prevents adversarial attempts from bypassing safety measures.

3. Topic Control

Topic control ensures conversations stay within predefined subject boundaries.

4. PII Detection and Masking

NeMo Guardrails PII Detection Options Application LLM Input PII Detection Output PII Masking Microsoft Presidio Private AI AutoAlign PII GuardrailsAI Validators

PII detection protects user privacy by detecting and masking sensitive data.

5. Agentic Security

NeMo Guardrails LLM Agent External Tools & APIs Application Pre-Tool Validation Post-Tool Validation Action Monitoring LLM Tool Calls Database File System External APIs

Agentic security provides guardrails for LLM agents using tools and external systems.

6. Custom and Third-Party Guardrails

NeMo Guardrails Custom Guardrail Options Application LLM Custom Input Rails Custom Output Rails Python Actions LangChain Tools Third-Party APIs Custom Scripts

Build custom guardrails using Python actions, LangChain tools, or third-party APIs.

7. Integration Options

Python SDK Integration Server API Integration OpenAI-Compatible Endpoint HTTP API Python App LLMRails LLM Any Application NeMo GuardrailsServer :8000 LLM Existing OpenAIIntegration /v1/chat/completions

Multiple integration options: Python SDK, HTTP Server API, or OpenAI-compatible endpoint.

Combined Architecture Overview

NVIDIA NeMo Guardrails Library Input Rails Output Rails NemoGuard NIMs Third-Party Integrations Application Code Large Language Model Content Safety Jailbreak Detection PII Detection Topic Validation Content Check PII Masking Response Validation Content Safety NIM Jailbreak Detection NIM Topic Control NIM Presidio / Private AI ActiveFence / AI Defense LlamaGuard / Fiddler

Complete architecture showing all guardrail types and integration options.