You talk about "semantic tree memory" but what does the system use as a storage? Just the text that is passed into and generated by the LLM? Doesn't that strategy risk context window overflows soon, and also content…
You talk about "semantic tree memory" but what does the system use as a storage? Just the text that is passed into and generated by the LLM? Doesn't that strategy risk context window overflows soon, and also content…