import PyPDF2
import spacy
import subprocess
from collections import Counter
import heapq
import io

# 自動檢查、下載 spaCy 語言模型（en_core_web_sm），避免 Space 缺模型報錯
try:
    nlp = spacy.load("en_core_web_sm")
except OSError:
    subprocess.run(["python", "-m", "spacy", "download", "en_core_web_sm"])
    nlp = spacy.load("en_core_web_sm")

def read_pdf(file_stream):
    """讀取 PDF 文字內容"""
    text = ''
    reader = PyPDF2.PdfReader(file_stream)
    for page in reader.pages:
        text += page.extract_text() + ' '
    return text.strip()

def extract_key_phrases(text):
    """擷取文章中的關鍵詞與專有名詞"""
    doc = nlp(text)
    key_phrases = [chunk.text for chunk in doc.noun_chunks] + [ent.text for ent in doc.ents]
    return key_phrases

def score_sentences(text, key_phrases):
    """根據關鍵詞分數給每個句子計分"""
    sentence_scores = {}
    doc = nlp(text)
    for sent in doc.sents:
        for phrase in key_phrases:
            if phrase in sent.text:
                if sent in sentence_scores:
                    sentence_scores[sent] += 1
                else:
                    sentence_scores[sent] = 1
    return sentence_scores

def summarize_text(sentence_scores, num_points=5):
    """依據分數挑出重要句子並條列化輸出"""
    summary_sentences = heapq.nlargest(num_points, sentence_scores, key=sentence_scores.get)
    summary = '\n'.join([f"- {sent.text}" for sent in summary_sentences])
    return summary