import json
import datasets

class MNLP_M2_rag_documents(datasets.GeneratorBasedBuilder):
    def _info(self):
        return datasets.DatasetInfo(
            features=datasets.Features({
                "title": datasets.Value("string"),
                "text": datasets.Value("string"),
                "source": datasets.Value("string"),
            }),
            description="RAG documents for MNLP",
        )

    def _split_generators(self, dl_manager):
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepath": "wiki_rag.jsonl"},
            )
        ]

    def _generate_examples(self, filepath):
        with open(filepath, encoding="utf-8") as f:
            for i, line in enumerate(f):
                data = json.loads(line)
                yield i, {
                    "title": data.get("title", ""),
                    "text": data.get("text", ""),
                    "source": data.get("source", "wikipedia"),
                }