Has anyone made an Ai system that can respond using different models?

[naive summit]

I didn't know what tag to use so I just put no code. The question is above.

[obtuse tide]

It happens all the time and can be approached in a couple of ways

1. You sort the inputs into the models
2. You extract features from the inputs and select the appropriate model from those features for the input
3. You pass the input through all the models and select the best output or evaluate the outputs for the best response

[naive summit]

I made a chat bot app for my computer and I made it use a different pretrained model based on the content in the prompt. I just wondered if anyone else made one so I might be able to improve mine using different ideas.

[fickle tundra]

It's not uncommon to use different models for different inputs, or even chain models, but if it's just by differnet prompts, it's a better practice to just have a well trained model that can generalize over all prompts

[naive summit]

@fickle tundra The reason I'm wanting to do it this way is because the really large models use too much RAM and I thought I might be able to first check if it can do the task with a smaller model and fall back on a much larger one if it can't.