"""Summarizer — map-reduce summarization for long documents."""
from __future__ import annotations

from chunker import chunk_text


def summarize_long_document(text: str, chunk_size: int = 20) -> dict:
  """Produce a hierarchical summary using map-reduce over chunks.

  Each chunk is summarized to its first two sentences, then the first
  three chunk summaries are joined into a final summary.

  Args:
      text: Raw document text (can be arbitrarily long).
      chunk_size: Words per intermediate chunk (default 20).

  Returns:
      Dict with keys: chunk_summaries (list[str]), final_summary (str).
  """
  chunk_summaries: list[str] = []
  for chunk in chunk_text("longdoc", text, chunk_size=chunk_size):
    summary = ". ".join(
        sentence.strip()
        for sentence in chunk.text.split(".")[:2]
        if sentence.strip()
    )
    chunk_summaries.append(summary)

  final_summary = ". ".join(chunk_summaries[:3])
  return {
    "chunk_summaries": chunk_summaries,
    "final_summary": final_summary,
  }