Example with JSONOutputParser, PydanticOutputParser, Pydantic and Optional Values

jsonoutputparser_pydantic.py

from langchain_openai import ChatOpenAI
from langchain_core.output_parsers import JsonOutputParser
from langchain_core.output_parsers import PydanticOutputParser
from langchain_core.prompts import PromptTemplate
from pydantic import BaseModel, ValidationError
from typing import List, Optional


# Initialize the language model. Replace 'openai' with your specified model.
llm = ChatOpenAI(model="gpt-4o-mini",temperature=0)

class OutputItem(BaseModel):
    title: str
    description: str
    tags: List[str]
    author: Optional[str] = None  # Optional field
    published_date: Optional[str] = None  # Optional fiel

class OutputArray(BaseModel):
    elements: List[OutputItem]


json_output_parser = JsonOutputParser(pydantic_object=OutputArray)
pydantic_output_parser = PydanticOutputParser(pydantic_object=OutputArray)

# Define the prompt template
prompt_template = PromptTemplate(
    input_variables=["keywords"],
    template="Given the keywords: {keywords}, return a JSON array of 10 objects. The elements 'author'  and 'published_date' are optional.\n{format_instructions}"
)


# Define the keywords you want to provide to the model
keywords = "AI, machine learning, LangChain"

# Generate the prompt
prompt = prompt_template.format(keywords=keywords,format_instructions=json_output_parser.get_format_instructions())

# Get the response from the LLM
response = llm.invoke(prompt)

# Parse the response using JsonOutputParser
parsed_json = json_output_parser.parse(response.content)  # Will throw an error if response is not valid JSON
output = pydantic_output_parser.parse(response.content)
# Output the parsed JSON
print(parsed_json)