AI Schema Detection: Automate Workflows with LLM APIs

What This App Does

Connect Excel directly to LLM APIs for AI-powered analysis:

Gemini (free tier available) or any model on OpenRouter: type the model name in a cell
Schema detection with structured JSON output: the LLM's answer follows a JSON schema set in the code
Identify categorical and numerical variables
Generate PostgreSQL data types and a description per column
Python EDA driven by the detected schema: statistics, value counts, correlations
Charts: histogram and box plot per numeric column, bar chart per categorical column

Keys are read from xlwings Lite environment variables with Local scope. They are kept in your browser and are not saved in the file.

Download & open: get the workbook and open xlwings Lite (Home > Add-ins). The HOW_IT_WORKS sheet explains every step; the buttons are on the MASTER sheet.
Add an API key: in xlwings Lite, Settings > Environment Variables, add GEMINI_API_KEY (free at aistudio.google.com) or OPENROUTER_API_KEY (openrouter.ai), scope Local, then restart xlwings Lite
Choose the table: RBICC (RBI bank-wise ATM, POS and card statistics) or RESP (bank marketing campaign), or any Excel table you add
Detect the schema: click "1a. Detect Schema (Gemini)" or "1b. Detect Schema (OpenRouter)"; results go to the SCHEMA sheet
Review: edit the categorical / numeric lists on SCHEMA if you disagree with the LLM
Run the EDA: click "2. Run EDA"; output goes to DISTROS_<table> and PLOT_<table>

Sends the header and first 10 rows of your table to the chosen LLM; the EDA runs on the full table inside Excel
Two calls in parallel: one for the categorical / numeric split, one for the PostgreSQL schema
Structured output: both answers follow a JSON schema, so the script reads them directly
Busy or temporary API errors are retried automatically