The prompt is sent to GPT. It looks at the data frame, it does a bunch of summarization stats on it. And then it pulls those summary stats together for all the columns and throws them together with my the rest of the prompt I have. So you actually look tries to go find that variable name in your call stack so that it can when it writes code,. puts all that together, sends it off to open AI, gets code back, uses Python AST to parse it, check that it's valid.