# NVIDIA-NeMo/Guardrails

NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems.

Repository: https://github.com/NVIDIA-NeMo/Guardrails
Canonical: https://ross.abutalabs.com/products/nvidia-nemo-guardrails
Homepage: https://docs.nvidia.com/nemo/guardrails/latest/index.html
Language: Python
License: NOASSERTION
License Family: other
Topics: agents, guardrails, python, safety, generative-ai, llms, nvidia, llm-safety, llm-security
Last push: 2026-08-26T21:38:00+00:00

## Health v2 (maintenance only)
Score: 93/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 87, longevity 88
- inputs: {"age_days": 1233, "days_push": 7, "days_rel": 7, "gap_med": 39, "n_releases_24m": 18}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 7017, forks 812 (observed 2026-08-28T04:09:53.356998+00:00)

## What it is
NVIDIA NeMo Guardrails is an open-source Python toolkit for adding programmable guardrails to LLM-based conversational applications. It sits between your application and the LLM, validating inputs, filtering retrieved context, steering dialog flows, and blocking or transforming unsafe or policy-violating outputs using YAML configs, Colang flows, and custom actions.

## Use cases
- add guardrails to my llm chatbot
- block prompt injection and jailbreak attempts
- keep the llm from talking about off-topic subjects
- detect and mask PII in user inputs and model outputs
- enforce a predefined conversation flow for a support bot
- validate retrieved documents before passing them to the llm
- scan llm outputs for safety and policy violations
- secure tool-calling in my ai agent workflow

## When to choose
- you need input, output, retrieval, and dialog rails around an LLM application
- you want declarative YAML/Colang configuration plus custom Python actions
- you need defense against prompt injection, jailbreaks, and PII leakage
- you want observability, evaluation, and vulnerability scanning for your rails
- you use LangChain or multiple LLM providers and need a provider-agnostic safety layer

## When to avoid
- you need a fully managed commercial content-moderation service rather than a self-hosted library
- your application does not involve LLMs or conversational AI
- you want zero-latency filtering, since rails add LLM calls and processing overhead
- you need non-Python runtimes, as the library is Python-only

## Facets
- artifact type: library
- maturity: active
- function: security, middleware, llm-inference, agent-framework, chatbot, prompt-engineering, cli, api-framework
- domain: large-language-models, security, chatbots, developer-tools
- platform: python, windows, self-hosted, cli
- tags: guardrails, llm-safety, colang, prompt-injection, jailbreak-protection, pii-detection, content-moderation, conversation-flow, langchain-integration, nvidia, ai-agents, natural-language-processing, linux, macos, docker

## Member repositories
- NVIDIA-NeMo/Guardrails (main) score 93

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:53.356998+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:40:38.744818+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA-NeMo/Guardrails (fetched 2026-08-28T04:09:53.356998+00:00, sha 7ae5c38135bf)
  - homepage: https://docs.nvidia.com/nemo/guardrails/latest/index.html (fetched 2026-08-29T08:36:43.153874+00:00, sha 6962834b84f7)
  - site_page: https://docs.nvidia.com/nemo/guardrails/about-nemo-guardrails-library/how-it-works (fetched 2026-08-29T08:36:43.163376+00:00, sha 17bfd42bc26e)
  - site_page: https://docs.nvidia.com/nemo/guardrails/about-nemo-guardrails-library/rail-types (fetched 2026-08-29T08:36:43.165246+00:00, sha 9f48a34a4822)
  - site_page: https://docs.nvidia.com/nemo/guardrails/about-nemo-guardrails-library/supported-llms (fetched 2026-08-29T08:36:43.166972+00:00, sha d7f7451c1e7b)
  - site_page: https://docs.nvidia.com/nemo/guardrails/about-nemo-guardrails-library/release-notes (fetched 2026-08-29T08:36:43.168764+00:00, sha 011853d1eaf3)
  - site_page: https://docs.nvidia.com/nemo/guardrails/get-started/installation-guide (fetched 2026-08-29T08:36:43.171137+00:00, sha e7c0e02307b1)
  - site_page: https://docs.nvidia.com/nemo/guardrails/configure-guardrails/configure-rails (fetched 2026-08-29T08:36:43.172940+00:00, sha 09cdc3c68cc4)
  - site_page: https://docs.nvidia.com/nemo/guardrails/run-guardrailed-inference/run-rails (fetched 2026-08-29T08:36:43.174539+00:00, sha 814bb0d657c9)
  - site_page: https://docs.nvidia.com/nemo/guardrails/resources/runtime-security-faq (fetched 2026-08-29T08:36:43.176033+00:00, sha dbac9a09f01f)
- Data as of 2026-08-30T08:39:29.467469+00:00.