SQLFluff: Auto-Fix Messy SQL with One Command

Plus evaluate LLM apps with PydanticAI

Grab your coffee. Here are this week’s highlights.

---

📅 Today’s Picks

Evaluate LLM Apps in One Line with PydanticAI

Problem

Testing LLM apps means validating multiple factors at once: is the answer correct, properly structured, fast enough, and natural sounding?

Rewriting this logic for every project is inefficient and error-prone.

Solution

pydantic-ai includes pydantic-evals, which provides these capabilities out of the box. Simply choose the evaluators you need and add them to your evaluation suite.

Built-in evaluators:

• Deterministic: validate that outputs are correct, properly typed, and fast enough

• LLM-as-judge: have another LLM grade qualities like helpfulness or tone

• Report-level: generate classification metrics across all cases automatically

🧪 Run code

---

SQLFluff: Auto-Fix Messy SQL with One Command

Problem

Consistent SQL style matters. It improves readability, speeds up code reviews, and makes bugs easier to identify.

Manual reviews can catch formatting issues, but they’re time-consuming and often inconsistent.

Solution

SQLFluff solves this with automated linting and formatting across 30+ SQL dialects. It identifies violations, applies consistent standards, and auto-corrects many problems.

SQLFluff also supports the following templates:

• Jinja

• SQL placeholders (e.g. SQLAlchemy parameters)

• Python format strings

• dbt (requires plugin)

🧪 Run code

---

🎓 Latest Interactive Course

Python Data Modeling with Dataclasses and Pydantic

Choosing between dict, NamedTuple, dataclass, and Pydantic comes down to how much safety you need. In this free interactive course, you’ll learn when to use each:

• Dictionary: Flexible, but no built-in field checks. Typos and missing keys only show up at runtime.

• NamedTuple: Immutable with fixed fields, helping catch mistakes early.

• dataclass: Mutable data containers with defaults and optional validation logic.

• Pydantic: Strong type validation, automatic coercion, and detailed error reporting.

All exercises run directly in your browser. No installation required.

---

☕️ Weekly Finds

spec-kit [Dev Tools] - Toolkit for Spec-Driven Development that helps define specs, generate plans and tasks, and implement code with AI coding tools

ty [Code Quality] - Extremely fast Python type checker and language server written in Rust, by the creators of uv and Ruff

nbQA [Code Quality] - Run ruff, isort, pyupgrade, mypy, pylint, flake8, and more on Jupyter Notebooks

---

💬 Rate Your Experience

How would you rate your newsletter experience? Share your feedback →

---

🔍 Explore More on CodeCut

• Tool Selector - Discover 70+ Python tools for AI and data science

• Production Ready Data Science - A practical book for taking projects from prototype to production