A validator that checks for system prompt leakage in LLM output.

Guardrails AI: Provenance Embeddings - Validates that LLM-generated text matches some source text based on distance in embedding space

Data structures used in the Guardrails Hub

Shared interfaces defined in JSON Schema.

OpenAPI Specifications and scripts for generating SDKs for the various Guardrails services

A Guardrails AI input validator that detects if the user is trying to jailbreak an LLM using unusual prompting techniques that involve jailbreaking and tricking the LLM

Guardrails AI: Toxic language - Validates that the generated text is toxic

FastAPI server for hosting Guardrails AI in your own environment.

Adding guardrails to large language models.