Question: How to get the context used when calling RA.answer_question?

[younes-io]

How to get the context used when calling RA.answer_question?

(this should help when testing/debugging/evaluating, etc)

[3CE8D2BAC65BDD6AA9]

Look for def answer_question in the class GPT3TurboQAModel in QAModels.py, then , you can print the context out.

[parthsarthi03]

You can use

context, __ = RA.retrieve(question)
print(context)

[daniyal214]

But why this does not retrieve the lists of chunks of the original document. Seems returning the summarized version.
Does the QA class answers the question from the summarized text? Instead of the actual content?

What I want is QA model to ask the question from the context which should be the chunks of the actual document not the summarized one. Is this possible?

Because I've been wanting to return the URLs in response, which are receiving broken. When I dig up found out that actual doc is not going to QA model.

Thanks.