NL2SQL-Viz (Snowflake TPCH): NL → SQL → Smart Visuals + Insights

A Streamlit app that converts natural language questions into guardrailed Snowflake SQL against SNOWFLAKE_SAMPLE_DATA.TPCH_SF1000, executes it, and renders smart visualizations with automatic insights. Includes optional audit logging to a writable database.

Features

• Guardrailed NL → SQL (whitelist TPCH tables, SELECT‑only, forced LIMIT, statement timeout)
• Smart visuals (Trend, Ranking + Pareto, Scatter + Correlation heatmap, Histogram fallback)
• Insights (trend delta %, Pareto 80/20, strongest correlation, outlier count)
• Advanced panel to show generated SQL + raw results
• Audit log to configurable AUDIT_DB.AUDIT_SCHEMA.AUDIT_TABLE using server-side CURRENT_TIMESTAMP()

Quickstart

python -m venv .venv && source .venv/bin/activate # Windows: .venv\Scripts\activate pip install -r requirements.txt cp .env.example .env # fill Snowflake creds and AUDIT_* values streamlit run app/app.py

Docker

docker build -t nl2sql-viz:local . docker run --rm -it -p 8501:8501 --env-file .env nl2sql-viz:local

Audit notes

Use a writable DB (not SNOWFLAKE_SAMPLE_DATA). Grant USAGE on DB/SCHEMA and CREATE TABLE/INSERT on the schema to your role, or run sql/create_audit_table.sql.