Just for anyone who tracks down this thread later, almost all of the modern blaze implementations (now decision modeler) use a compiled sequential Java code bas

Just for anyone who tracks down this thread later, almost all of the modern blaze implementations (now decision modeler) use a compiled sequential Java code base rather than a RETE algorithm. This is because the performance impact of maintaining the working knowledge is pretty significant, surprisingly hard to predict and there are a lot of technologies inside of modern rules engines that can make dependencies much easier to reason about. RETE is cool and common context applications with LLMs is going to be a very interesting space that may revive this as a more general purpose technology.