It’s actually kinda easy. Neural networks are just weirder than usual logic gate circuits. You can program them just the same and insert explicit controlled logic and deterministic behavior. To somebody who don’t know the details of LLM training, they wouldn’t be able to tell much of a difference. It will be packaged as a bundle of node weights and work with the same interfaces and all.
The reason that doesn’t work well if you try to insert strict logic into a traditional LLM despite the node properties being well known is because of how intricately interwoven and mutually dependent all the different parts of the network is (that’s why it’s a LARGE language model). You can’t just arbitrarily edit anything or insert more nodes or replace logic, you don’t know what you might break. It’s easier to place inserted logic outside of the LLM network and train the model to interact with it (“tool use”).
Real-time in computing usually either means a real-time OS with guaranteed low latency response (typically for stuff like microcontrollers regulating machines) or streaming live data (low latency delivery of the most recent value)
This sounds like the latter, and a typical SQL database don’t guarantee real-time updates (you can have “atomic writes” to prevent inconsistency but usually this would make it slower) but some databases like this one are designed to ensure you can read out updated correct and consistent values much faster. Also with standard databases you usually make scheduled individual requests, but a real-time database could often send a stream of updated values to a “subscribing” program