WINDOW_SIZE = 2
sentence_rows = []

for document in documents:
    sentences = split_sentences(
        document["text"]
    )

    for i, sentence in enumerate(sentences):
        start = max(
            0,
            i - WINDOW_SIZE,
        )

        end = min(
            len(sentences),
            i + WINDOW_SIZE + 1,
        )

        sentence_rows.append({
            "sentence_id": (
                f"{document['source']}"
                f"::sentence::{i}"
            ),
            "source": document["source"],
            "sentence_text": sentence,
            "window_text": " ".join(
                sentences[start:end]
            ),
            "window_start": start,
            "window_end": end - 1,
        })

sentence_df = pd.DataFrame(
    sentence_rows
)

sentence_df["embedding"] = get_embeddings(
    sentence_df[
        "sentence_text"
    ].tolist(),
    "sentences",
)

        __
        __