FietSQL Prompt & Eval Toolkit

Utilities for building high-quality prompts and validating SQL produced by the cwolff/fietSQL text-to-SQL model (also packaged for Ollama as cowolff/fietSQL). Supply your database schema plus a natural-language question, and this package will both format the chat prompt expected by the model and sanity-check the SQL it returns by running it inside an in-memory SQLite database.

Features

• Prompt builder: creates the <|im_start|> style prompt template used to steer the fietSQL model.
• Schema-aware evaluation: applies your DDL to SQLite and runs the generated SQL to verify executability.
• Lightweight footprint: pure-Python utilities that work in notebooks, batch jobs, or services.

Installation

pip install fietSQL

Or install straight from source:

pip install git+https://github.com/trl-lab/FietSQL-Package.git

Quick start

from fietSQL import build_prompt, evaluate_executability

SCHEMA = '''
CREATE TABLE customers (
    id INTEGER PRIMARY KEY,
    name TEXT,
    city TEXT
);

CREATE TABLE orders (
    id INTEGER PRIMARY KEY,
    customer_id INTEGER,
    total NUMERIC
);
'''

question = "Which city has the highest total sales?"

prompt = build_prompt(SCHEMA, question)

# send `prompt` to cwolff/fietSQL via Hugging Face or Ollama

sql = '''
SELECT c.city, SUM(o.total) AS total_sales
FROM customers c
JOIN orders o ON o.customer_id = c.id
GROUP BY c.city
ORDER BY total_sales DESC
LIMIT 1;
'''

report = evaluate_executability(SCHEMA, sql)
assert report.is_executable, report.details

Related resources

• Hugging Face model card: https://huggingface.co/cwolff/fietSQL
• Pulling the Ollama build: ollama pull cowolff/fietSQL
• Issues and discussions: https://github.com/trl-lab/FietSQL-Package/issues

License

Released under the MIT License. See LICENSE for details.