LangChain · Lesson 11 of 15
Structured Output and Tool Calling
Force JSON that matches a Pydantic model and bind tools the model can call.
- Intermediate
- 16 min read
- 3 objectives
Before this lessonLesson 10: Streaming and Callbacks
What you will learn
- with_structured_output
- Bind a tool
- Parse a tool call
Your Progress
0 of 15 lessons 0%
- Lessons0 / 15
- Completed0
- Est. time left~ 4 hours
Create a free account to keep your progress on every device.
Free-form text is a bad API. Ask the model to fill a schema, or to call a tool whose arguments are that schema.
Pydantic first
from pydantic import BaseModel, Field
class Ticket(BaseModel):
title: str = Field(description="Short summary")
priority: str = Field(description="low, medium, or high")
steps: list[str]
extractor = model.with_structured_output(Ticket)
ticket = extractor.invoke("The checkout 500s on Safari when the cart has a coupon.")
print(ticket.priority, ticket.steps)Tools
from langchain_core.tools import tool
@tool
def get_weather(city: str) -> str:
"""Return a one-line forecast for a city."""
return f"Sunny, 24C in {city}"
llm = model.bind_tools([get_weather])
msg = llm.invoke("Weather in Lisbon?")
print(msg.tool_calls)If tool_calls is set, run the function, append a tool result message, and call the model again. That loop is an agent; bound it with a max-steps counter.