Yes, I would like to know too.

Right now, once I run a nlp.process(), any attempt to retrain or load a model creates a situation where any subsequent nlp.process() results with outputs from the original model rather than what I would expect from the new model, unless the utterance was an exact match to new phrases added.

Therefore, I'm finding that I also have to end the app and restart with the new model.

I've got a workaround.

Instead of const nlp = new NlpManager( options );, I use var nlp = new NlpManager( options );.
so that when I want to reload a different model or retrain a new model, I would first nlp = new NlpManager( options ); and immediately nlp.load('model.nlp'); BEFORE performing any nlp.process( input );

There may be a better way, but just flat out rebuilding nlp seemed the safest.

@ScienceLion - been digging for this comment for awhile (knew I read this somewhere). Just wanted to say how much I appreciate your advice / workaround (hoping this works in browser). 🤞

Thank you big time!