Show HN: Voice-to-SQL – ask a database in plain English (LLM → SQL)

A developer built and open-sourced Voice-to-SQL, a tool that uses Llama 3.3 70B via Groq to convert plain English business questions into read-only SELECT SQL queries against a sample SaaS database, displaying the generated SQL. The project's code is available on GitHub.

Background

- This is a demo app that turns spoken English questions into SQL queries using a large language model (Llama 3.3 70B, a Meta AI model, running on Groq's fast inference hardware). It works against a read-only sample database of SaaS metrics (e.g. subscriptions, revenue).<br>- "Voice-to-SQL" is a well-known pattern in the LLM space: using natural language as a front-end for databases, reducing the need to know SQL syntax. Many companies (e.g. Databricks, Microsoft) offer similar tools.<br>- Llama 3.3 70B is a relatively large open-weight model; Groq is a company that provides extremely low-latency LLM inference via specialized chips (LPUs). The combination aims to make the SQL generation feel near-instant.<br>- The project is open-source (on GitHub). The author, Adrian Moreno, built it as a personal project and posted it on Hacker News ("Show HN") for feedback.