How to get a list as an output?

[HFulcher]

Hi, just want to get some clarification.

I want to prompt ChatGPT to generate a list and return it. I'm looking at the Kor docs but I'm not sure how to use the schema to specify to return a list? The list isn't of a fixed length so I can't create an Object with x Text values.

Thanks!

[eyurtsev]

Documentation is a bit hidden: https://eyurtsev.github.io/kor/validation.html#validating-collections

There are recipes for extracting collections using a CSV encoder and also for using a JSON encoder.

Feel free to drop an example of the schema / task you're trying to do as it would make it easier to suggest a solution.

[HFulcher]

Hey, thanks for the help!

An example would be:

Prompt - "Create 10 baby names as a list"

Potential Schema:

Object(
id='names',
description='Names generated by ChatGPT',
attributes = [
Text(id='name')
],
many=True
)
Will the above specify the output should be a list of Text?

[eyurtsev]

Would this work?

from kor.extraction import create_extraction_chain
from kor.nodes import Object, Text, Number
from langchain.chat_models import ChatOpenAI

schema = Object(
    id="person",
    description="Personal information",
    examples=[
        ("Alice and Bob are friends", [{"first_name": "Alice"}, {"first_name": "Bob"}])
    ],
    attributes=[
        Text(
            id="first_name",
            description="The first name of a person.",
        )
    ],
    many=True,
)

llm = ChatOpenAI(
    model_name="gpt-3.5-turbo",
    temperature=0,
    max_tokens=2000,
)

chain = create_extraction_chain(llm, schema)

text_with_names = llm.predict(text="Generate 10 baby names as a list.")
chain.run(text_with_names)

{'data': {'person': [{'first_name': 'Olivia'},
   {'first_name': 'Liam'},
   {'first_name': 'Ava'},
   {'first_name': 'Noah'},
   {'first_name': 'Sophia'},
   {'first_name': 'Jackson'},
   {'first_name': 'Isabella'},
   {'first_name': 'Aiden'},
   {'first_name': 'Mia'},
   {'first_name': 'Lucas'}]},
 'raw': 'first_name\nOlivia\nLiam\nAva\nNoah\nSophia\nJackson\nIsabella\nAiden\nMia\nLucas',
 'errors': [],
 'validated_data': {}}