Skip to content

🛠️ Task Configuration

Create a JSON task file in the tasks/ folder using the following naming format:

TaskXXX_taskname.json

📌 Required Fields

Your JSON file must include the following keys:

  • Description: Description of the task
  • Data_Path: Filename of the dataset
  • Input_Field: Column containing the text data
  • Parser_Format: The filename of the file containing the parser format in the tasks/parsers/ folder

🔍 About Parser_Format

This field should reference a Python file located in tasks/parsers/ which defines a Pydantic model called OutputParser.

This file is usually created using the build-parser tool. It must contain a class named OutputParser, which will be used by the LLM to structure its response into a valid JSON object.

You may define other nested models to keep the schema organized, but OutputParser is the root that the system will rely on.

Example:

"Parser_Format": "product_parser.py"

The file tasks/parsers/product_parser.py might look like:

from pydantic import BaseModel

class Product(BaseModel):
    name: str
    price: float

class OutputParser(BaseModel):
    products: list[Product]

🔄 Optional (for using examples)

  • Example_Path: Path to the file with examples to include in the prompt

Note: If you're not using examples, simply omit the Example_Path field.
Do not set it to an empty string!