skit_pipelines.components.identify_compliance_breaches_llm package

Submodules

skit_pipelines.components.identify_compliance_breaches_llm.utils module

class Call(id: 'int', uuid: 'str', audio_url: 'str', call_url: 'str', flow_uuid: 'str', client_uuid: 'str', reftime: 'str', turns: 'List[Turn]')[source]

Bases: object

audio_url: str
call_url: str
client_uuid: str
flow_uuid: str
id: int
reftime: str
turns: List[skit_pipelines.components.identify_compliance_breaches_llm.utils.Turn]
uuid: str
class Turn(id: 'int', uuid: 'str', reftime: 'str', is_bot: 'bool', utterance: 'str')[source]

Bases: object

id: int
is_bot: bool
reftime: str
utterance: str
uuid: str
format_call(input_call: skit_pipelines.components.identify_compliance_breaches_llm.utils.Call) → str[source]

Convert the turns of a call representing the entire conversation into a single string that would be used as input to the LLM model

get_prompt_text()[source]
parse_calls(input_df) → list[skit_pipelines.components.identify_compliance_breaches_llm.utils.Call][source]

Convert the turn-level dataframe obtained from fetch_calls into a list of calls

slice_json(text: str) → dict[source]

Slice the JSON part from given text and return a parsed dictionary or list.

Module contents

identify_compliance_breaches_llm(s3_file_path: str) → str[source]

Groups turns into calls and pushes them to an LLM (uses openai chatComplete functionality) to identify compliance breaches. The result value for each call is written in an output csv file

param s3_file_path: Csv file containing turns for calls obtained from fsm Db type s3_file_path: str

output: path of csv file containing complaince breach results for each call in the input

identify_compliance_breaches_llm_op(s3_file_path: str)

Identify compliance breaches llm Groups turns into calls and pushes them to an LLM (uses openai chatComplete functionality) to identify