Quail
Open-source AI-SQL engine that writes LLM-powered conditions directly inside SQL queries.
About
Quail (the QUery-Aware Inference Layer) is an open-source AI-SQL engine released in September 2026 by Shreya Shankar's team at Full Stack Data Lab, developed in collaboration with Modal. AI-SQL integrates natural-language conditions into SQL itself: users register data tables in Apache Arrow format in a Python environment and write custom AI functions directly inside queries, so LLM-powered filtering and joining runs in batch on local or cloud GPUs.
The problem it solves is the old iterative loop of filtering text rows by sending each row to an external LLM API through client libraries. Quail jointly plans the query and the model inference; across 29 QUAIL-B benchmark queries it is 1.84x faster on average than well-tuned vLLM baselines, up to 14x (team-reported). The source is MIT licensed on GitHub at fsdatalab/quail with a live Modal-hosted demo.
The honest picture: Quail is an engine for data engineers and researchers, not a no-code text-to-SQL product. It assumes comfort with Python, Arrow tables, and GPU or Modal compute. Benchmark figures are team-reported and should be validated against your own workload.
What it can do
- AI functions embedded directly in SQL queries
- Batch LLM inference over Arrow tables
- Joint query planning and inference optimization
- Open-weight model support instead of closed LLM APIs
Setup
Clone the GitHub repo and install in a Python environment; GPU or Modal compute for inference.
Prerequisites
Python environment, data in Apache Arrow format, access to GPU compute or a Modal account.
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 114; Trending score 76/100 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.