JsonOutputParser for structured data in LangChain - Mini Project: Build & Apply

Using JsonOutputParser in Langchain for Structured Data

📖 Scenario: You are building a chatbot that needs to return answers in a clear, structured JSON format. This helps other programs easily understand the chatbot's replies.

🎯 Goal: Create a Langchain setup that uses JsonOutputParser to parse the chatbot's JSON-formatted output into structured data with specific fields.

📋 What You'll Learn

Create a response_schema list with two fields: answer and source

Create a JsonOutputParser instance using the response_schema

Use the parser to parse a sample JSON text into structured data

Print the final parsed structured data

💡 Why This Matters

🌍 Real World

Many chatbots and AI tools need to return answers in a structured format so other programs can easily read and use the data.

💼 Career

Understanding how to parse structured outputs with Langchain's JsonOutputParser is useful for building reliable AI applications and integrations.

1

Create the response schema list

Create a list called response_schema with two dictionaries: one with name as answer and description as The chatbot's answer, and another with name as source and description as Where the answer came from.

LangChain

# Your code here

Need a hint?

Think of response_schema as a list of fields your JSON output will have.

2

Create the JsonOutputParser instance

Create a variable called parser and assign it to JsonOutputParser initialized with the response_schema list.

LangChain

response_schema = [
    {'name': 'answer', 'description': "The chatbot's answer"},
    {'name': 'source', 'description': 'Where the answer came from'}
]
# Your code here

Need a hint?

Import JsonOutputParser from langchain.output_parsers and create parser with response_schema.

3

Parse a sample JSON using the parser

Create a variable called sample_text with the JSON string '{"answer": "42", "source": "Encyclopedia"}'. Then create a variable called formatted_output by calling parser.parse(sample_text).

LangChain

from langchain.output_parsers import JsonOutputParser

response_schema = [
    {'name': 'answer', 'description': "The chatbot's answer"},
    {'name': 'source', 'description': 'Where the answer came from'}
]

parser = JsonOutputParser(response_schema=response_schema)

# Your code here

Need a hint?

Use the parse method of parser to parse sample_text into structured data.

4

Print the formatted JSON output

Add a line to print the variable formatted_output.

LangChain

from langchain.output_parsers import JsonOutputParser

response_schema = [
    {'name': 'answer', 'description': "The chatbot's answer"},
    {'name': 'source', 'description': 'Where the answer came from'}
]

parser = JsonOutputParser(response_schema=response_schema)

sample_text = '{"answer": "42", "source": "Encyclopedia"}'
formatted_output = parser.parse(sample_text)

# Your code here

Need a hint?

Use print(formatted_output) to see the structured output.