The community members are facing issues with using the Pandas query engine with open-source language models like Zephyr 7b. They are seeking guidance on working with CSV files and building a RAG chatbot. The comments suggest that using open-source language models for specific outputs or outputs that need to be parsed requires a lot of work to be reliable. Additionally, the community members are experiencing slow output generation with the CodeLLaMA instruct model and are looking for ways to speed up the process and enable memory/chat history in the Pandas query engine.
Did you set up the prompts properly for zephyr? In any case though, using open-source LLMs for very specific outputs, or outputs you need to parse, requires a lot of work to be reliable
