app.py

import gradio as gr
import requests
import re
from bs4 import BeautifulSoup
from urllib.parse import urljoin
import json

# --- 核心函式 ---
BASE_URL = "https://ivod.ly.gov.tw"
HEADERS = {
    'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/5.37.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/5.37.36'
}

def get_all_legislators_map():
    """
    (爬蟲 A1) 抓取頁面上的 select 選項，建立 姓名->ID 的對照表。
    """
    target_url = f"{BASE_URL}/Demand?targetSession=current"
    print(f"--- 爬蟲 A1 啟動 ---")
    
    legislator_map = {}
    try:
        response = requests.get(target_url, headers=HEADERS, timeout=10)
        response.encoding = 'utf-8'
        response.raise_for_status()
        soup = BeautifulSoup(response.text, 'html.parser')
        
        selects = soup.find_all('select')
        
        for select in selects:
            options = select.find_all('option')
            for option in options:
                code = option.get('value')
                name = option.text.strip()
                
                # 過濾邏輯
                if code and code.isdigit() and name:
                    legislator_map[name] = code
        
        print(f"爬取完成，共找到 {len(legislator_map)} 位委員。")
        
        if not legislator_map:
             return {}, gr.Dropdown(choices=[], value=None, label="錯誤：未找到資料")
        
        names_list = sorted(list(legislator_map.keys()))
        
        return (
            legislator_map, 
            gr.Dropdown(
                choices=names_list, 
                value=names_list[0], 
                interactive=True,
                label="1. 選擇委員 (已載入)"
            )
        )
    except Exception as e:
        msg = f"錯誤：爬取委員列表失敗。 {e}"
        print(msg)
        return {}, gr.Dropdown(choices=[], label=msg)

def get_videos_for_legislator(legislator_name: str, legislator_map: dict):
    """
    (爬蟲 A2) 爬取指定委員的影片列表。
    """
    
    # --- [修正重點 2] API 呼叫防護 ---
    # 如果是透過 API Client 呼叫，legislator_map 可能是 None 或空字典
    # 這時我們需要現場重新抓一次名單，確保能查到 ID
    if not legislator_map or legislator_name not in legislator_map:
        print(f"⚠️ 檢測到 Map 為空或找不到 '{legislator_name}'，正在重新獲取名單...")
        new_map, _ = get_all_legislators_map()
        if isinstance(new_map, dict) and new_map:
            legislator_map = new_map
        else:
            return None, "錯誤：無法獲取委員名單，請稍後再試。"

    # 1. 查詢 ID
    legislator_id = legislator_map.get(legislator_name)
    
    if not legislator_id:
        return None, f"錯誤：找不到委員 '{legislator_name}' 的 ID。"

    legislator_page_url = f"{BASE_URL}/Demand/ListLgltVideo/{legislator_id}"
    print(f"--- 爬蟲 A2 啟動: {legislator_name} (ID: {legislator_id}) ---")
    
    all_clips = []
    page_num = 1
    
    try:
        while True:
            page_url = f"{legislator_page_url}?page={page_num}"
            # print(f"正在爬取第 {page_num} 頁...")
            
            response = requests.get(page_url, headers=HEADERS, timeout=10)
            response.raise_for_status()
            soup = BeautifulSoup(response.text, 'lxml') # 建議安裝 lxml: pip install lxml
            
            clip_lists = soup.find_all('ul', id='clipUl')
            if not clip_lists:
                break
            
            found_clips_on_page = 0
            for ul in clip_lists:
                list_items = ul.find_all('li')
                for item in list_items:
                    clip_info = {"legislator_name": legislator_name}
                    
                    meet_name_tag = item.select_one("span.metdec")
                    clip_info['meeting_name'] = meet_name_tag.string.strip() if meet_name_tag else "會議名稱未知"
                    
                    play_link_tag = item.find('a', title=re.compile(r"委員寬頻影片"))
                    if play_link_tag and play_link_tag.get('href'):
                        absolute_url = urljoin(BASE_URL, play_link_tag.get('href'))
                        clip_info['playback_page_url'] = absolute_url
                        all_clips.append(clip_info)
                        found_clips_on_page += 1
            
            if found_clips_on_page == 0:
                 break
            
            # 簡易分頁檢查
            pagination_script = soup.find('script', string=re.compile(r"var total ="))
            if pagination_script:
                total_match = re.search(r'var total = "(\d+)"', pagination_script.string)
                size_match = re.search(r'var pageSize = "(\d+)"', pagination_script.string)
                if total_match and size_match and (page_num * int(size_match.group(1)) >= int(total_match.group(1))):
                    break
            else:
                break
                
            page_num += 1
            
        msg = f"爬取完成！共抓取 {legislator_name} 委員 {len(all_clips)} 筆發言片段。"
        print(msg)
        return all_clips, msg

    except Exception as e:
        msg = f"錯誤：爬取影片失敗。 {e}"
        print(msg)
        return None, msg

# --- 建立 Gradio 介面 ---
with gr.Blocks() as iface:
    gr.Markdown("# IVOD 委員發言爬蟲 API (修正版)")
    
    legislator_map_state = gr.State(value={}) # 初始給空字典
    
    with gr.Row():
        # --- [修正重點 1] allow_custom_value=True ---
        # 這允許 API Client 傳入任何字串，不會因為初始 choices 為空而被擋下
        legislator_input = gr.Dropdown(
            label="1. 選擇委員",
            choices=[],
            interactive=True,
            allow_custom_value=True 
        )
        analyze_button = gr.Button("開始爬取", variant="primary")
    
    status_output = gr.Textbox(label="處理狀態", interactive=False)
    json_output = gr.JSON(label="發言片段 URL 列表")

    # 頁面載入時自動抓名單
    iface.load(
        fn=get_all_legislators_map,
        inputs=[],
        outputs=[legislator_map_state, legislator_input],
        api_name="get_all_legislators_map"
    )
    
    analyze_button.click(
        fn=get_videos_for_legislator,
        inputs=[legislator_input, legislator_map_state],
        outputs=[json_output, status_output],
        api_name="get_videos_for_legislator"
    )

if __name__ == "__main__":
    iface.launch()