Classify user input and system outputs across six safety categories.
npx skills add https://github.com/ovachiever/droid-tings --skill llamaguard
Please help me install this Agent Skill. Skill: llamaguard Source: https://github.com/ovachiever/droid-tings/tree/main/skills/llamaguard Command: npx skills add https://github.com/ovachiever/droid-tings --skill llamaguard
This Skill introduces LlamaGuard for safe input/output moderation with high accuracy, including optional deployments via vLLM, HuggingFace, and NeMo Guardrails.
Deploy LlamaGuard with a chosen backend (vLLM/HuggingFace/NeMo Guardrails) and route user input through the moderator before invoking LLMs.
High-intent search queries and answers about installing and using this skill.
Input and output moderation classifies text across six safety categories (violence, sexual content, weapons, substances, self-harm, criminal planning) using LlamaGuard. Deploy it with vLLM, HuggingFace, or NeMo Guardrails to filter prompts before reaching your LLM and responses before returning to users, ensuring production safety compliance.
Yes. LlamaGuard integrates with vLLM for high-throughput inference and HuggingFace transformers for model loading and inference. Both backends support deterministic safety classification, enabling you to route moderation through your existing inference stack without separate infrastructure.
LlamaGuard classifies content into six categories: violence and hate speech (S1), sexual content (S2), weapons (S3), illegal substances (S4), self-harm (S5), and criminal planning (S6). Each category receives a deterministic safety score for filtering decisions in API endpoints and production pipelines.
Deploy LlamaGuard using FastAPI or REST endpoints with vLLM or HuggingFace backends to handle batch moderation requests. Route multiple user inputs and system outputs through the moderator before LLM invocation, enabling scalable safety filtering across chat interactions and production deployments.
Yes. LlamaGuard integrates with NeMo Guardrails for orchestrated safety workflows and supports cloud deployment on SageMaker and edge environments. This enables deterministic safety classification as part of larger guardrail orchestration and managed inference pipelines.
LlamaGuard provides high-throughput moderation with optimized inference using vLLM or HuggingFace transformers and torch, minimizing latency overhead. Pre- and post-LLM filtering adds minimal latency while maintaining deterministic safety classification across concurrent requests in production.
Orchestra-ResearchFilter LLM input and output for harmful content using Meta's LlamaGuard model.
SufficientDaikonDesign a multi-stage content moderation pipeline with duplicate detection and quality scoring.
Supporter09Classify LLM chat messages into six safety categories for moderation.
ovachieverClassify user input and system outputs across six safety categories.
