Structured output

model.with_structured_output(Schema) makes the model return data in a shape you define, not free text, so your code gets fields it can trust instead of a string to parse.

Last updated: 29 Sep, 2026 · LangChain 1.4

A model's reply is text, worded differently every time. When a node needs a decision your code will act on, such as a category and whether a ticket is urgent, you want a fixed shape, not a sentence.

A movie in four typed fields

The LangChain section of the video starts structured output with Pydantic. A Movie class inherits BaseModel and declares four fields: the title (a string), the year the movie was released (an int), the director (a string) and the rating out of 10 (a float). Each Field carries a description that tells the model which value goes where, and Pydantic checks each value against its type.

with_structured_output(Movie) wraps the model so its reply is a Movie object. Asked without the wrapper, "Provide details about the movie Inception" gets paragraphs; through the wrapper it gets four typed fields. The video's model, groq:qwen/qwen3-32b, has been retired, so the example runs on the course model. The video's prompt reads "moview", corrected here, and the video shows the result bare where this prints it:

from langchain.chat_models import init_chat_model

model = init_chat_model("groq:openai/gpt-oss-120b")

from pydantic import BaseModel, Field

class Movie(BaseModel):
    title: str = Field(description="The title of the movie")
    year: int = Field(description="This year the movie was released")
    director: str = Field(description="The director of the movie")
    rating: float = Field(description="The movies rating out of 10")

model_with_structure = model.with_structured_output(Movie)
response = model_with_structure.invoke("Provide details about the movie Inception")
print(repr(response))

Movie(title='Inception', year=2010, director='Christopher Nolan', rating=8.8)

The shop's version below triages a complaint into a category and an urgency flag its code can branch on.

The with_structured_output call

from pydantic import BaseModel

class Triage(BaseModel):
    category: str
    urgent: bool

structured = model.with_structured_output(Triage)
result = structured.invoke("I was charged twice and I am furious")
print(result.category, result.urgent)   # -> billing True

The output schema

First describe the shape you want back. A Pydantic model lists the field names and their types.

from pydantic import BaseModel

class Triage(BaseModel):   # the shape: two fields with types
    category: str
    urgent: bool

Asking the model for that shape

Wrap the model with with_structured_output. The wrapped model reads the message and hands back a Triage object instead of text.

structured = model.with_structured_output(Triage)
result = structured.invoke("I was charged twice and I am furious")

Reading the fields

Read the fields straight off the object. There is no string to parse and no wording to guess at.

print(result.category, result.urgent)   # read fields directly

Reading a structured result end to end

The same pieces in one file, ready to run.

from pydantic import BaseModel
from langchain.chat_models import init_chat_model

class Triage(BaseModel):
    category: str
    urgent: bool

model = init_chat_model("groq:openai/gpt-oss-120b", temperature=0)  # uses your GROQ_API_KEY
structured = model.with_structured_output(Triage)

result = structured.invoke("I was charged twice and I am furious")
print(type(result).__name__)
print(result.category, result.urgent)

Triage
billing True

The model read the complaint and filled both fields itself.

Why the fields are safe to read

• You describe the shape with a Pydantic model: field names and types.
• with_structured_output returns a model that fills that shape and hands you an object, not text.
• Your node then reads result.category directly, with no parsing and no guessing at the wording.

Free text vs structured output

Where structured output fits

• Classifying a ticket into a category your code branches on.
• Extracting fields (an order id, a date, a yes or no) the next node needs.

Related

• Previous: ToolNode and tools_condition
• Next: Agent loop
• Reference: Structured output