Basic Extraction Patterns

Optional Fields

This example demonstrates how to handle optional or missing data in structured extractions:
1. Using Optional type annotations for fields that might not be present in the source text
2. Using Instructor's Maybe type to explicitly track whether information was present

Optional fields allow your models to gracefully handle incomplete information without
causing extraction failures when certain data isn't mentioned in the text.

Import the necessary libraries

from pydantic import BaseModel, Field
from typing import Optional
import instructor
from openai import OpenAI

Define the Person class with Pydantic

class Person(BaseModel):
    name: str
    age: int
    # Optional fields with default value of None
    email: Optional[str] = None
    phone: Optional[str] = None
    occupation: Optional[str] = None

Patch the client

client = instructor.from_openai(OpenAI())

Extract with optional fields

person = client.chat.completions.create(
    model="gpt-3.5-turbo",
    response_model=Person,
    messages=[
        {"role": "user", "content": "John Smith is 35 years old and works as a software engineer."}
    ]
)

Output:
Name: John Smith
Age: 35
Email: None
Phone: None
Occupation: software engineer

print(f"Name: {person.name}")
print(f"Age: {person.age}")
print(f"Email: {person.email}")  # None
print(f"Phone: {person.phone}")  # None
print(f"Occupation: {person.occupation}")  # "software engineer"

Running the Example

First, install Instructor and any dependencies

$ pip install instructor pydantic

Run the Python script

$ python optional-fields.py

Further Information

Documentation link 1