Hacker News new | ask | show | jobs
by verdverm 729 days ago
My strategy has been to implement in / follow along with llamaindex, dig into the details, and then implement that in a less abstracted, easily understandable codebase / workflow.

Was driven to do so because it was not as easy as I'd like to override a prompt. You can see how they construct various prompts for the agents, it's pretty basic text/template kind of stuff