To start with, I'm not coordinating across multiple systems or chaining LLM calls. I write all data to a central data store, run a state machine on it, and each LLM call then operates independently.
At that point, you can measure and optimize each call as needed and it is fairly straightforward.
To start with, I'm not coordinating across multiple systems or chaining LLM calls. I write all data to a central data store, run a state machine on it, and each LLM call then operates independently.
At that point, you can measure and optimize each call as needed and it is fairly straightforward.