Setup & orientation·4 min
The corpus
0 / 1
Before you retrieve anything, meet the data. The corpus is a set of short factual paragraphs, and the QA set pairs each question with a gold answer that is a verbatim span of one paragraph. That single property is the engine of this whole course's determinism.
A QA record (with its source paragraph)
{ "question": "When did construction of the Eiffel Tower begin?",
"goldAnswer": "1887",
"goldParagraphId": "eiffel-tower#p1",
"context": "...Construction of the Eiffel Tower began in 1887 and was completed in 1889..." }
The corpus ships as fixtures/data/corpus.json — an array of {paragraphId, title, text} — and the QA set as fixtures/data/qa.json — an array of {questionId, question, goldAnswer, goldParagraphId}. You will chunk and index the corpus, then retrieve against the questions.