08. OutputFixingParser

OutputFixingParser

OutputFixingParser Provides the ability to automatically fix errors that may occur during the output parsing process. This parser is basically another parser, for example PydanticOutputParser It is designed to fix the error through additional LLM calls when wrapping and returning an output or error in a format that this parser cannot handle.

At the heart of this approach, if the first attempt results in non-compliance with the schema, OutputFixingParser Is that it automatically recognizes the wrong output, and resubmits it to the model with a new command to fix it. In this process, the command for correction should pinpoint the error and include specific instructions to reconstruct the data in the correct format.

For example, PydanticOutputParser I tried using to generate output that complies with a specific data schema, but it may happen that some fields are missing or the data type is wrong. At this time OutputFixingParser As a next step, we submit a new request to LLM, including instructions to correct the error. LLM will generate a new output that fixes the error based on these instructions.

Copy

# API KEY를 환경변수로 관리하기 위한 설정 파일
from dotenv import load_dotenv

# API KEY 정보로드
load_dotenv()

Copy

True

Copy

from langchain_openai import ChatOpenAI
from langchain.output_parsers import PydanticOutputParser
from langchain_core.pydantic_v1 import BaseModel, Field
from typing import List


class Actor(BaseModel):
    name: str = Field(description="name of an actor")
    film_names: List[str] = Field(
        description="list of names of films they starred in")


actor_query = "Generate the filmography for a random actor."

parser = PydanticOutputParser(pydantic_object=Actor)

Copy

# Deliberately enter incorrect format
misformatted = "{'name': 'Tom Hanks', 'film_names': ['Forrest Gump']}"

# Attempting to parse data entered in an incorrect format
parser.parse(misformatted)

# Error Output

OutputFixingParser I will use to correct the wrong format.

Copy

from langchain.output_parsers import OutputFixingParser

new_parser = OutputFixingParser.from_llm(parser=parser, llm=ChatOpenAI())

Copy

# Output in incorrect format
misformatted

Copy

"{'name':'Tom Hanks','film_names': ['Forrest Gump']}" 

Copy

# OutputFixingParser Parsing malformed output using
actor = new_parser.parse(misformatted)

Copy

# Parsed results
actor

Copy

Actor (name='Tom Hanks', film_names=['Forrest Gump'])