🛡️ PromptGuard

LLM prompt safety evaluator — powered by RAG + Llama
Built on a 181k prompt safety dataset.

Demo Notice
This live demo uses a 50K sample of the dataset (out of 181K) due to storage limitations on HuggingFace Spaces.
For full results, clone the project locally from GitHub, no changes needed, it runs on the complete dataset automatically.