protectai/llm-guard

Provides modular input and output scanners for LLM pipelines—including prompt injection detection, secret redaction, toxicity analysis, and factual consistency checking—deployable as a Python library or standalone API. Uses a composable scanner architecture enabling fine-grained control over which security checks run on user inputs and model outputs. Integrates with OpenAI's API and other LLM providers through straightforward configuration.

2,660 stars and 329,796 monthly downloads. Used by 1 other package. Available on PyPI.