import re

def chunk_text_paragraphs(text, max_length=500, overlap=1):
    """
    Zerteilt den Text absatzweise in Chunks bis `max_length` Zeichen.
    Optional wird `paragraph_overlap` Absatz(e) vom vorherigen Chunk übernommen.
    """

    paragraphs = [p.strip() for p in re.split(r'\n{2,}', text) if p.strip()]
    chunks = []
    current_chunk = []
    current_len = 0

    i = 0
    while i < len(paragraphs):
        para = paragraphs[i]
        para_len = len(para)

        # Wenn dieser Absatz den max Chunk sprengt → neuen Chunk
        if current_len + para_len + 2 > max_length:  # +2 für Leerzeile
            if current_chunk:
                chunks.append("\n\n".join(current_chunk))
                # Optional: letzte N Absätze behalten
                if overlap > 0:
                    current_chunk = current_chunk[-overlap:]
                    current_len = sum(len(p) for p in current_chunk) + 2 * len(current_chunk)
                else:
                    current_chunk = []
                    current_len = 0
            else:
                # Einzelner Absatz ist zu groß → hart splitten
                chunks.append(para[:max_length])
                para = para[max_length:]
                paragraphs.insert(i + 1, para)  # Rest zurück in Liste
                i += 1
                continue
        else:
            current_chunk.append(para)
            current_len += para_len + 2  # +2 für Trennung

        i += 1

    if current_chunk:
        chunks.append("\n\n".join(current_chunk))

    return chunks