dex0shubham

Really Cool, the on-device shifts the economics a bit. Usually output tokens are longer and 5x more expensive than input tokens, but here it's more about speed. so shouldn't we count the tokens an agent needs for completing a task rather than tokens per call ?

Does Pensieve reduce token usage by giving more relevant context to the agents, instead of agents searching and looking through lot of files ?

Thanks for the detailed explanation, I will try with apsw.ext.query_info. In my case I was building column-keyed dicts from results but as the result was an empty set, there was nothing to key on. Appreciate the pointer to the execution tracer too.

This is a really interesting project, I wanted to try this myself but at times the agent produces code that looks right but it isn't. Is there a way to automate the the process of finding these kind of mistakes ?

I was experimenting with APSW and I noticed that when a query doesn't return any row, it becomes difficult to get column names from the cursor, is there any recommended way to do this ?