I have a simple assistant, it works well when single message submitted and run. I tried two messages and got:
Is my assumption wrong that assistant should execute both messages and provide a respnse for each?
The AI underneath the API calls and threads and messages looks at incoming tokens, processes them and produces output , when you add two messages one after the other the AI sees those token groups concatenated together, it’s a way for you manage input that would logically contain multiple parts, perhaps it’s a telephone conversation where different people are speaking, maybe something involving several types of information from different sources.
A simple test to do is, what would a person do if you gave them the exact same text? would they do both math problems? In my head, if I see a list of math problems, that’s a test… so I’d do them all. Seems reasonable.
1 Like