A health RAG chatbot splits language understanding from data control
A fully assistant is being designed to answer questions over health-related Excel files. The Excel files are not mainly numbers; they are tables with many text-heavy columns, ranging from single words to long sentences. The data is in both English and French.
The planned flow is: a user asks a question, an LLM parser turns it into a JSON intent, a repair and step checks it, Python and pandas filter the Excel rows, and the answer is shown to the user. The LLM is used mainly to understand the question and put it into structure. The actual row selection is handled by fixed rules.
The main concern is that users may ask with words that do not exactly match the Excel data. The system also cannot allow , because wrong answers in health use could have serious consequences.
Key points
- The assistant is fully local and aimed at health-domain Excel files.
- The Excel files contain mostly text, not mostly numbers.
- The LLM turns a user question into a JSON intent.
- Python and pandas then filter the rows with fixed rules.
- The main tension is flexible user language versus strict control to prevent .