Hello HN!<p>We have been working on an open source routing mechanism that has built in guardrails (i.e. preventing jailbreak, injection, PII leakage). It's really similar to OpenRouter but we've added a guardrail mechanism to it.<p>These guardrails are highly extensible but have focused around natural language based interactions (e.g. you have a chatbot that you don't want leaking PII).<p>Some our guardrails are other open source models (like Metas prompt guard they released with 3.1 or Microsoft Presidio).<p>Would love feedback to see what folks think!<p>Thanks!