📘 Overview

PromptedRefiner is an operator that rewrites, refines, or normalizes text within a specified column of a dataframe. It operates by concatenating a user-defined system_prompt with the content of each row in the target column (input_key). The combined text is then sent to a Large Language Model (LLM) to generate a refined version. The key feature of this operator is that it performs an in-place update, overwriting the original content in the input_key column with the newly generated text.

__init__

def __init__(self, llm_serving: LLMServingABC, system_prompt: str = "You are a helpful agent."):

Parameter	Type	Default	Description
llm_serving	LLMServingABC	Required	An instance of the Large Language Model serving class, responsible for inference.
system_prompt	str	"You are a helpful agent."	A system-level instruction that guides the LLM on how to refine the text.

Prompt Template Descriptions

Prompt Template Name	Primary Use	Applicable Scenarios	Feature Description

run

def run(self, storage: DataFlowStorage, input_key: str = "raw_content"):

Parameter	Type	Default	Description
storage	DataFlowStorage	Required	The dataflow storage instance for reading the input dataframe and writing the result.
input_key	str	"raw_content"	The name of the column containing the text to be refined. This column will be overwritten with the refined text.

🧠 Example Usage

🧾 Default Output Format (Output Format)

The operator modifies the input dataframe in-place by overwriting the column specified by input_key. The output dataframe will have the same schema as the input, but the content of the target column will be replaced by the LLM's generated text.

Field	Type	Description
other_columns	any	Other columns from the input are preserved.
input_key	str	The original text in this column is replaced by the refined text generated by the model.

Example Input:

{
    "id": 1,
    "raw_content": "this text has several errrors and bad formatting"
}

Example Output (with input_key="raw_content"):

{
    "id": 1,
    "raw_content": "This text has several errors and poor formatting."
}