[cyanydeez]

was reading through open llama, looks like way to get pertinent results is via different ranking algorithm and score based on convergence. then shove that back into the LLM