AI·rete·RAG separates rule-based decisions from LLM explanations for auditable AI

A developer has released AI·rete·RAG, a hybrid system designed to make automated decisions in regulated domains — such as lending, fraud detection, and clinical triage — fully auditable. The tool runs a pure-Python Rete rule engine to produce deterministic verdicts based on YAML-defined rules, ensuring the same inputs always yield the same outcome. A separate RAG layer then uses an LLM to generate plain-English explanations of the decision by retrieving relevant passages from policy documents, without being able to alter the verdict. An audit mode logs every rule evaluated, including those that did not fire, and captures a snapshot of the rule set for replay during regulatory review. The platform is hosted with a free tier, includes a no-signup live demo across eight domains, and offers an open-source MCP client that allows AI agents like Claude to call the decision engine as a tool.