How I can use LLM transformer to apply LLM for all data source output and shcema type?

[YuriyGavrilov]

Just in case: I have one table with some 10 fields or it will be all tables in data source like for while each table in datasource.
Next step it will be LLM to identify sensitive data in the source output. field names and 10 lines of real data limit should be taken in considiration. The output of LLM should be a json string with field names, identified data category and flag true or false about sensitive. So what do you think is it possible? Can't imagine how to collect all the table schema and rows and send to LLM transformer in one request.

[Hisoka-X]

I haven't actually tried it, but it's theoretically possible. For each row of data, we convert it to json and send it to LLM.
eg:

name	age	loc
Jack	24	USA
Tom	18	ENG

{"name":"Jack", "age": 24, "loc": "USA"}      --> LLM (prompt: "check...., the output of LLM should be a json string with field names, identified data category and flag true or false about sensitive") 
{"name":"Tom", "age": 18, "loc": "ENG"}    --> LLM (prompt: "check...., the output of LLM should be a json string with field names, identified data category and flag true or false about sensitive") 

The output by LLM transform:

name	age	loc	llm_output
Jack	24	USA	{"name":false, "age": true, "loc": true}
Tom	18	ENG	{"name":false, "age": true, "loc": true}

Then you can use JsonPath transform to process the json data.