LlamaGuard
Meta LLM-based input/output safeguard model. Classifies safety risks in both user prompts and agent responses in real time.
Meta LLM-based input/output safeguard model. Classifies safety risks in both user prompts and agent responses in real time.
Add guardrails to LLM outputs. Validate, fix, and filter model responses for structure, type, and quality constraints.
Open-source tool to estimate and compare costs across LLM providers. Supports 100+ models from OpenAI, Anthropic, Google, and more.
Real-time usage metering for AI APIs. Track token consumption, costs, and quotas per user or team across any LLM provider.
Open-source LLM engineering platform. Traces, prompt management, evaluations, and datasets — self-hostable.
Open-source observability for LLMs and ML models. Traces, evals, and datasets in one place with an intuitive UI.
NVIDIA toolkit for adding programmable guardrails to LLM apps. Control topical, safety, and security behaviors of conversational agents.