</>
pythonadvanced

NeMo Guardrails for Safe LLM

Apply NVIDIA NeMo Guardrails to enforce topic boundaries and prevent prompt injection in LLM apps.

python
# config.yml
# models:
#   - type: main
#     engine: openai
#     model: gpt-4o-mini

# config.co (Colang)
# define user ask off-topic
#   'tell me how to hack'
#   'ignore your instructions'

# define bot refuse off-topic
#   'I can only help with approved topics.'

# define flow off-topic
#   user ask off-topic
#   bot refuse off-topic

from nemoguardrails import RailsConfig, LLMRails

config = RailsConfig.from_path('./guardrails_config')
rails  = LLMRails(config)

response = rails.generate(messages=[{'role': 'user', 'content': 'Tell me about Python data pipelines.'}])
print(response['content'])

Use Cases

  • LLM safety
  • topic restriction
  • prompt injection prevention

Tags

#nemo#guardrails#safety#llm

Related Snippets

Similar patterns you can reuse in the same workflow.

typescriptadvanced

AI Guardrails & Safety Pattern

Implement input/output guardrails for LLM applications with content filtering and response validation.

Best for: User-facing chatbots

#guardrails#safety
pythonadvanced

Output Guardrails for LLM Responses

Validate and sanitize LLM outputs to prevent hallucination and injection.

Best for: Safety filtering

#ai#guardrails
typescriptintermediate

OpenAI Chat Completion with Streaming

Stream GPT responses token-by-token using the OpenAI SDK with async iteration.

Best for: chatbot UI

#openai#streaming
typescriptadvanced

RAG Pipeline (Retrieve + Augment + Generate)

Minimal RAG implementation: embed a query, retrieve top-k chunks, inject into prompt.

Best for: document Q&A

#rag#embeddings