Defog
Open-source text-to-SQL powered by the SQLCoder LLM family.
About
Defog is a data analysis platform built around SQLCoder, an open-source family of large language models fine-tuned specifically for converting natural language questions into SQL queries. SQLCoder outperforms GPT-3.5 and GPT-4 on text-to-SQL benchmarks for novel schemas, and the weights are openly available for self-hosting.
Defog offers enterprise deployment with on-prem options for sensitive data, multi-step reasoning for complex analytical questions, and an open-source sql-eval framework for reproducible evaluation of generated SQL. The company serves customers in healthcare, finance, and government.
What it can do
- SQLCoder open-source text-to-SQL models
- Self-hostable deployment for sensitive data
- Multi-step reasoning for complex questions
- Open-source sql-eval benchmarking framework
Setup
Use the hosted Defog platform or self-host SQLCoder from Hugging Face with pip install sqlcoder.
Prerequisites
A SQL database; a GPU helps for self-hosted inference.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 131; Trending score 80/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.