📘 Overview

ExtractSmilesFromText is an operator designed to extract or parse chemical SMILES expressions from OCR text. It uses a given prompt_template to construct model inputs, combining text content and optional abbreviation/monomer information, then calls a Large Language Model (LLM) to produce structured outputs which are parsed into JSON format.

__init__

def __init__(self, llm_serving: LLMServingABC, prompt_template=None)

Parameter	Type	Default	Description
llm_serving	LLMServingABC	Required	The Large Language Model serving instance for executing inference.
prompt_template	Any	None	The prompt template object used to construct the model input.

Prompt Template Descriptions

Prompt Template Name	Primary Use	Applicable Scenarios	Feature Description

run

def run(self, storage: DataFlowStorage, content_key: str = "text", abbreviation_key: str = "abbreviations", output_key: str = "synth_smiles")

Parameter	Type	Default	Description
storage	DataFlowStorage	Required	The data flow storage instance responsible for reading and writing data.
content_key	str	"text"	The name of the input column containing the OCR text.
abbreviation_key	str	"abbreviations"	The name of the input column containing abbreviation/monomer information.
output_key	str	"synth_smiles"	The name of the output column where the extracted results will be stored.

🧠 Example Usage

🧾 Default Output Format (Output Format)

Field	Type	Description
...	...	Input columns from the source dataframe.
synth_smiles (default)	list/dict	The JSON-parsed SMILES structure extracted by the model. Returns [] on failure.