backend.generators module

class backend.generators.PrivacyPolicyGenerator(regulation_query_engine: SubQuestionQueryEngine, syllabus_pipeline: QueryPipeline, generate_pipeline: QueryPipeline, regenerate_pipeline: QueryPipeline, judge_pipeline: QueryPipeline)

Bases: object

classmethod from_defaults(links_df: DataFrame, model_name: str = 'Equall/Saul-7B-Instruct-v1', model_url: str | None = None, verbose: bool = False)

Create a privacy policy generator from defaults

Parameters:

• links_df – a DataFrame with columns ‘regulations’ and ‘links’

• verbose – whether to show verbose output

Params model_url:

the model url; if passed, use TGI, else load locally

Returns:

a privacy policy generator

generate(section_name: str, information: str, threshold: int = 5) → Dict[str, Any]

Generate a section of a privacy policy

Parameters:

• section_name – the name of the section

• information – the information for generating the section

• threshold – the maximum number of attempts to regenerate

Returns:

success or not and the generated section

get_syllabus(regulations: List[str]) → Dict[str, List[str]]

Get a privacy policy syllabus

Parameters:

regulations – the regulations to comply with

Returns:

a privacy policy syllabus

regenerate(section_name: str, section_text: str, suggestions: str, threshold: int = 5)

Regenerate a section of a privacy policy

Parameters:

• section_name – the name of the section

• section_text – the text of the section

• suggestions – the suggestions for improvement

• threshold – the maximum number of attempts to regenerate

Returns:

success or not and the regenerated section