NeMo-Guardrails is an open-source toolkit that allows developers to add programmable guardrails semantically matched on utterances to LLM-based conversational applications. NeMo-Guardrails can be easily integrated with watsonx.ai models using LangChain’s WatsonxLLM Integration.
Feb 27, 2024
Effective governance of enterprise services and applications that utilise generative models requires a multi-layered approach of different classifiers that guardrail the inputs to and outputs from generative models. These models, which are called synchronously by the application and drive application logic and consumed via an API, abstracted away through an SDK or inferenced directly, must themselves be governed. These models too, must be explainable, monitored for drift (if neural) and for fairness.
Feb 10, 2024