Requirement:
I have Factory operational data coming in from different DBs from multiple IT Systems. Now I want to ingest these data into my local LLM (tried with Llama and Phi-3) and then ask questions like how was the Production volume in the last 5 months or perhaps even more detailed questions like tell me about the machine health for the last 1 hour.

The Model should be able to look into the ingested data and provide responses like – “Based on the last available data it seems that the machine was running without any problems. There was 3 spikes recorded by the temperature sensor but the cycle time has remained constant.”

Problem: When I am ingesting the csv data which contains the details of the factory operational data then I am getting responses like – “As an AI model I do not have access to Realtime data.”

However, when I am converting the csv data into descriptive sentences then it is provided some responses (not satisfactory) but at least something.

I was wondering if converting the csv to descriptive text is the way to go about it since in terms of complexity it might increase as an when we got more variety of data or is there an easy way to achieve this ? I have tried downloading Llama 7B Q2 (since I have a small RAM) and Phi-3 and ingested it using Anything LLM with the LLM hosted on LM Studio.