AI City Popy 🏙️

District 09 · Teacher Academy

🏙️ Journey

0/13

📋

✅

🗂️

District 9 · Teacher Academy

Welcome to Pydantic Academy

Where messy data gets a strict makeover — every field clean, typed, and validated.

Enter The Academy

Before Pydantic vs After

❌ Messy

age: "twenty"
email: null
budget: "lots"

✅ Pydantic

age: 20
email: "a@b.com"
budget: 1500

Messy Data Breaks Everything

AI outputs and API inputs can arrive in any shape. One missing field or wrong type and everything crashes.

# Raw API response — anything goes 😱

data = {
  "name": "Alice",
  "age": "twenty-five", # should be int!
  "budget": None, # should be required!
  "email": "not-valid", # not an email!
}

👩‍🏫

Pydantic reads your class definition and automatically checks every incoming value. No more silent errors!

Watch Pydantic Catch Errors

Try entering wrong types — Pydantic instantly spots the problem.

name : str

age : int

email : EmailStr

budget : float

Pydantic + AI Structured Outputs

Modern LLMs can return validated JSON matching your Pydantic model — no prompt hacking needed.

Nesting Models Inside Models

Pydantic models can contain other models — they all validate recursively.

street: str

city: str

zip: str

👩‍🏫

User validates Company which validates Address — Pydantic checks the whole tree in one call!

Mission: Fix The Broken Model

Mrs. Pydantic's student model has wrong types. Click each broken field to fix it.

class Student(BaseModel):

Tap the red broken fields to fix them.

Get started

Master Pydantic in 4 steps 👩‍🏫

From messy dict to fully validated, typed data model.

1
Install Pydantic
Pydantic is already included in FastAPI. Or install standalone.
```
pip install pydantic[email]
```

Define your model

Subclass BaseModel and declare fields with Python type hints.

from pydantic import BaseModel, EmailStr

class User(BaseModel):
    name: str
    age: int
    email: EmailStr
    budget: float = 0.0  # with default

Validate incoming data

Pass a dict and Pydantic validates, coerces, and raises clear errors.

# Valid
user = User(name="Alice", age="25",
           email="a@b.com", budget=1500)
print(user.age)  # → 25 (int, not "25")

# Invalid — raises ValidationError
User(name="Alice", age="oops",
     email="bad", budget="lots")

Use with OpenAI

Force the model to return a validated Pydantic object.

from pydantic import BaseModel
from openai import OpenAI

client = OpenAI()

class Recommendation(BaseModel):
    title: str
    reason: str
    score: float

response = client.beta.chat.completions.parse(
    model="gpt-4o-mini",
    messages=[{"role":"user","content":"Recommend a film"}],
    response_format=Recommendation,
)
rec = response.choices[0].message.parsed
print(rec.score)  # → always a float ✅

Ask Popy

Chat with the Teacher ✨

Questions about validation errors, field constraints, or complex schemas? Ask away!

Hi! I'm Popy 👩‍🏫 Ask me anything about Pydantic, data validation, type hints, or structured AI outputs!

You're a Data Guardian now!

Your AI outputs are clean, typed, and validated. Next: learn how to measure whether your AI is actually doing a good job — Evals!

Continue Journey →🧪 Testing Lab →

Mini Project

Build Quest

Strict Output Contract

Deliverable: Define one schema and validate model output into typed fields.

Stretch: Add one custom validator for a business rule.

Complete the deliverable first, then unlock the stretch goal.

🏫 Open-Book School

🧪 Testing Lab