Structured Data Not Working for ChatAnthropic()

[sheitin]

I think it doesn't like that the pydantic model has an array.

here is a reprex and the error message:

from typing import Literal
from pydantic import BaseModel, Field
from chatlas import ChatAnthropic, content_pdf_file
import pandas as pd
import dotenv
import os
import json

dotenv.load_dotenv()

text = "The new quantum computing breakthrough could revolutionize the tech industry."


class Classification(BaseModel):

    name: Literal[
        "Politics", "Sports", "Technology", "Entertainment", "Business", "Other"
    ] = Field(description="The category name")

    score: float = Field(description="The classification score for the category, ranging from 0.0 to 1.0.")


class Classifications(BaseModel):
    """Array of classification results. The scores should sum to 1."""

    classifications: list[Classification]


chat = ChatAnthropic(
    system_prompt = "you are a friendly but terse assistant"
)

data = chat.extract_data(text, data_model=Classifications)
pd.DataFrame(data["classifications"])

gives

BadRequestError: Error code: 400 - {'type': 'error', 'error': {'type': 'invalid_request_error', 'message': 'tools.0.custom.input_schema: JSON schema is invalid. It must match JSON Schema draft 2020-12 ([https://json-schema.org/draft/2020-12](vscode-file://vscode-app/Applications/Positron.app/Contents/Resources/app/out/vs/code/electron-sandbox/workbench/workbench.html#)). Learn more about tool use at [https://docs.anthropic.com/en/docs/tool-use.](vscode-file://vscode-app/Applications/Positron.app/Contents/Resources/app/out/vs/code/electron-sandbox/workbench/workbench.html#)'}}
File [~/DsSandbox/price-parser-python/test-example.py:33](vscode-file://vscode-app/Applications/Positron.app/Contents/Resources/app/out/vs/code/electron-sandbox/workbench/workbench.html#)
     26     classifications: list[Classification]
     29 chat = ChatAnthropic(
     30     system_prompt = "you are a friendly but terse assistant"
     31 )
---> 33 data = chat.extract_data(text, data_model=Classifications)
     34 pd.DataFrame(data["classifications"])

BUT this works (notice that we used Classification for the data_model instead of the array Classifications

data = chat.extract_data(text, data_model=Classification)

print(data)

gives

{'name': 'Technology', 'score': 0.95}

[cpsievert]

transform_schema() seems useful for this problem

https://docs.claude.com/en/docs/build-with-claude/structured-outputs#working-with-json-outputs-in-sdks