How can we pass memory or a planned trajectory to the model during inference?

Hi,

I am trying to understand the best way to provide the model with memory or a planned trajectory during inference.
I sometimes already have a structured plan for the model, what I want to know is: What is the recommended way to pass this kind of memory to the model? Should it simply be appended as extra text in the prompt, or is there a more effective format?

Thank you

The issue names no file, test, or inference entry point. Start by locating the model's inference interface and existing prompt or input handling, then determine how memory or a planned trajectory is represented; done means a documented or implemented recommendation with a clear validation path.

Written by the indexing model from the issue text.