feat(config): 更新影视配置与API接口
- 修改 `XBPQ/魔兽.json` 中的分类URL、分类项及增加多项解析规则 - 在 `api.json` 中新增“奈飞”APP源及相关配置 - 替换部分短剧搜索源为新的提供商(如酷乐、爱看、鬼鬼) - 更新夸克、UC、天翼、123、百度等推送接口地址 - 将星芽短剧接口由 `csp_AppXY` 改为本地 Python 脚本并开启搜索和筛选功能 - 升级 spider.jar 文件以支持新特性
This commit is contained in:
parent
e78963df22
commit
20b72ee393
@ -1,6 +1,16 @@
|
||||
{
|
||||
"简介": "p_txt show_part\">&&</p>",
|
||||
"副标题": "v_note\">&&</div>",
|
||||
"分类url": "https://saohuody.com/list/{cateId}-{catePg}.html",
|
||||
"分类": "国产剧$12#港剧$13#台剧$14#日剧$15#韩剧$16#美剧$17#海外$18#泰剧$19#动作$5#喜剧$6#爱情$7#科幻$8#恐怖$9#剧情$10#战争$11#动画$33#福利$20"
|
||||
"首页": "0",
|
||||
"请求头": "User-Agent$MOBILE_UA#Accept$text/html,application/xhtml+xml,application/xml;;q=0.9,image/avif,image/webp,image/apng,*/*;;q=0.8,application/signed-exchange;;v=b3;;q=0.7&&Accept-Language@zh-CN,zh;;q=0.9",
|
||||
"分类url": "https://shdy2.com/list/{cateId}-{catePg}.html",
|
||||
"分类": "剧集$2#电影$1#港剧$21#台剧$26#日剧$24#韩剧$22#美剧$23",
|
||||
"数组": "class=\"v_img&&/li>",
|
||||
"标题": "alt=\"&&\"",
|
||||
"图片": "original=\"&&\"",
|
||||
"链接": "href=\"&&\"",
|
||||
"简介": "show_part\">&&<a",
|
||||
"演员": "主演:&&<a",
|
||||
"导演": "导演:&&/",
|
||||
"播放二次截取": "class=\"play_list&&/ul>",
|
||||
"播放数组": "<li&&/li>",
|
||||
"倒序": "1"
|
||||
}
|
59
api.json
59
api.json
@ -251,6 +251,22 @@
|
||||
"ua": ""
|
||||
}
|
||||
},
|
||||
{
|
||||
"key": "奈飞",
|
||||
"name": "奈飞丨APP",
|
||||
"type": 3,
|
||||
"quickSearch": 1,
|
||||
"api": "csp_AppGet",
|
||||
"ext": {
|
||||
"url": "https://movier.ink",
|
||||
"site": "",
|
||||
"dataKey": "moviermovierink9",
|
||||
"dataIv": "moviermovierink9",
|
||||
"deviceId": "",
|
||||
"version": "",
|
||||
"ua": ""
|
||||
}
|
||||
},
|
||||
{
|
||||
"key": "迪吧",
|
||||
"name": "迪吧丨APP",
|
||||
@ -1011,49 +1027,62 @@
|
||||
"api": "csp_MiSou"
|
||||
},
|
||||
{
|
||||
"key": "短剧大全",
|
||||
"name": "短剧|大全",
|
||||
"key": "酷乐搜索",
|
||||
"name": "酷乐|搜索",
|
||||
"type": 3,
|
||||
"api": "csp_DuanjuDQ"
|
||||
"api": "csp_KuLe"
|
||||
},
|
||||
{
|
||||
"key": "短剧合集",
|
||||
"name": "短剧|合集",
|
||||
"key": "爱看搜索",
|
||||
"name": "爱看|搜索",
|
||||
"type": 3,
|
||||
"api": "csp_DuanjuHJ"
|
||||
"api": "csp_AiKan"
|
||||
},
|
||||
{
|
||||
"key": "短剧鬼鬼",
|
||||
"name": "短剧|鬼鬼",
|
||||
"key": "鬼鬼搜索",
|
||||
"name": "鬼鬼|搜索",
|
||||
"type": 3,
|
||||
"api": "csp_DuanjuGG"
|
||||
"api": "csp_GuiGui"
|
||||
},
|
||||
{
|
||||
"key": "夸克推送",
|
||||
"name": "夸克|推送",
|
||||
"type": 4,
|
||||
"api": "http://zling.online:12331/api.php?type=quark"
|
||||
"api": "https://so.yinpai.xyz/api.php?type=quark"
|
||||
},
|
||||
{
|
||||
"key": "UC推送",
|
||||
"name": "UC|推送",
|
||||
"type": 4,
|
||||
"api": "http://zling.online:12331/api.php?type=uc"
|
||||
"api": "https://so.yinpai.xyz/api.php?type=uc"
|
||||
},
|
||||
{
|
||||
"key": "天翼推送",
|
||||
"name": "天翼|推送",
|
||||
"type": 4,
|
||||
"api": "https://so.yinpai.xyz/api.php?type=tianyi"
|
||||
},
|
||||
{
|
||||
"key": "123推送",
|
||||
"name": "123|推送",
|
||||
"type": 4,
|
||||
"api": "https://so.yinpai.xyz/api.php?type=123"
|
||||
},
|
||||
{
|
||||
"key": "百度推送",
|
||||
"name": "百度|推送",
|
||||
"type": 4,
|
||||
"api": "http://zling.online:12331/api.php?type=baidu"
|
||||
"api": "https://so.yinpai.xyz/api.php?type=baidu"
|
||||
},
|
||||
{
|
||||
"key": "星芽短剧",
|
||||
"name": "星芽|短剧",
|
||||
"type": 3,
|
||||
"api": "csp_AppXY",
|
||||
"api": "./py/星芽短剧.py",
|
||||
"searchable": 1,
|
||||
"quickSearch": 0,
|
||||
"filterable": 0
|
||||
"changeable": 1,
|
||||
"quickSearch": 1,
|
||||
"filterable": 1
|
||||
},
|
||||
{
|
||||
"key": "甜圈短剧",
|
||||
|
343
py/星芽短剧.py
Normal file
343
py/星芽短剧.py
Normal file
@ -0,0 +1,343 @@
|
||||
# coding = utf-8
|
||||
# !/usr/bin/python
|
||||
|
||||
"""
|
||||
"""
|
||||
|
||||
from Crypto.Util.Padding import unpad
|
||||
from Crypto.Util.Padding import pad
|
||||
from urllib.parse import unquote
|
||||
from Crypto.Cipher import ARC4
|
||||
from urllib.parse import quote
|
||||
from base.spider import Spider
|
||||
from Crypto.Cipher import AES
|
||||
from bs4 import BeautifulSoup
|
||||
from base64 import b64decode
|
||||
import urllib.request
|
||||
import urllib.parse
|
||||
import binascii
|
||||
import requests
|
||||
import base64
|
||||
import json
|
||||
import time
|
||||
import sys
|
||||
import re
|
||||
import os
|
||||
|
||||
sys.path.append('..')
|
||||
|
||||
xurl = "https://app.whjzjx.cn"
|
||||
|
||||
headers = {
|
||||
'User-Agent': 'Linux; Android 12; Pixel 3 XL) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/98.0.4758.101 Mobile Safari/537.36'
|
||||
}
|
||||
|
||||
headerf = {
|
||||
"platform": "1",
|
||||
"user_agent": "Mozilla/5.0 (Linux; Android 9; V1938T Build/PQ3A.190705.08211809; wv) AppleWebKit/537.36 (KHTML, like Gecko) Version/4.0 Chrome/91.0.4472.114 Safari/537.36",
|
||||
"content-type": "application/json; charset=utf-8"
|
||||
}
|
||||
|
||||
times = int(time.time() * 1000)
|
||||
|
||||
data = {
|
||||
"device": "2a50580e69d38388c94c93605241fb306",
|
||||
"package_name": "com.jz.xydj",
|
||||
"android_id": "ec1280db12795506",
|
||||
"install_first_open": True,
|
||||
"first_install_time": 1752505243345,
|
||||
"last_update_time": 1752505243345,
|
||||
"report_link_url": "",
|
||||
"authorization": "",
|
||||
"timestamp": times
|
||||
}
|
||||
|
||||
plain_text = json.dumps(data, separators=(',', ':'), ensure_ascii=False)
|
||||
|
||||
key = "B@ecf920Od8A4df7"
|
||||
key_bytes = key.encode('utf-8')
|
||||
plain_bytes = plain_text.encode('utf-8')
|
||||
cipher = AES.new(key_bytes, AES.MODE_ECB)
|
||||
padded_data = pad(plain_bytes, AES.block_size)
|
||||
ciphertext = cipher.encrypt(padded_data)
|
||||
encrypted = base64.b64encode(ciphertext).decode('utf-8')
|
||||
|
||||
response = requests.post("https://u.shytkjgs.com/user/v3/account/login", headers=headerf, data=encrypted)
|
||||
response_data = response.json()
|
||||
Authorization = response_data['data']['token']
|
||||
|
||||
headerx = {
|
||||
'authorization': Authorization,
|
||||
'platform': '1',
|
||||
'version_name': '3.8.3.1'
|
||||
}
|
||||
|
||||
class Spider(Spider):
|
||||
global xurl
|
||||
global headerx
|
||||
global headers
|
||||
|
||||
def getName(self):
|
||||
return "首页"
|
||||
|
||||
def init(self, extend):
|
||||
pass
|
||||
|
||||
def isVideoFormat(self, url):
|
||||
pass
|
||||
|
||||
def manualVideoCheck(self):
|
||||
pass
|
||||
|
||||
def extract_middle_text(self, text, start_str, end_str, pl, start_index1: str = '', end_index2: str = ''):
|
||||
if pl == 3:
|
||||
plx = []
|
||||
while True:
|
||||
start_index = text.find(start_str)
|
||||
if start_index == -1:
|
||||
break
|
||||
end_index = text.find(end_str, start_index + len(start_str))
|
||||
if end_index == -1:
|
||||
break
|
||||
middle_text = text[start_index + len(start_str):end_index]
|
||||
plx.append(middle_text)
|
||||
text = text.replace(start_str + middle_text + end_str, '')
|
||||
if len(plx) > 0:
|
||||
purl = ''
|
||||
for i in range(len(plx)):
|
||||
matches = re.findall(start_index1, plx[i])
|
||||
output = ""
|
||||
for match in matches:
|
||||
match3 = re.search(r'(?:^|[^0-9])(\d+)(?:[^0-9]|$)', match[1])
|
||||
if match3:
|
||||
number = match3.group(1)
|
||||
else:
|
||||
number = 0
|
||||
if 'http' not in match[0]:
|
||||
output += f"#{match[1]}${number}{xurl}{match[0]}"
|
||||
else:
|
||||
output += f"#{match[1]}${number}{match[0]}"
|
||||
output = output[1:]
|
||||
purl = purl + output + "$$$"
|
||||
purl = purl[:-3]
|
||||
return purl
|
||||
else:
|
||||
return ""
|
||||
else:
|
||||
start_index = text.find(start_str)
|
||||
if start_index == -1:
|
||||
return ""
|
||||
end_index = text.find(end_str, start_index + len(start_str))
|
||||
if end_index == -1:
|
||||
return ""
|
||||
|
||||
if pl == 0:
|
||||
middle_text = text[start_index + len(start_str):end_index]
|
||||
return middle_text.replace("\\", "")
|
||||
|
||||
if pl == 1:
|
||||
middle_text = text[start_index + len(start_str):end_index]
|
||||
matches = re.findall(start_index1, middle_text)
|
||||
if matches:
|
||||
jg = ' '.join(matches)
|
||||
return jg
|
||||
|
||||
if pl == 2:
|
||||
middle_text = text[start_index + len(start_str):end_index]
|
||||
matches = re.findall(start_index1, middle_text)
|
||||
if matches:
|
||||
new_list = [f'{item}' for item in matches]
|
||||
jg = '$$$'.join(new_list)
|
||||
return jg
|
||||
|
||||
def homeContent(self, filter):
|
||||
result = {}
|
||||
result = {"class": [{"type_id": "1", "type_name": "剧场"},
|
||||
{"type_id": "3", "type_name": "新剧"},
|
||||
{"type_id": "2", "type_name": "热播"},
|
||||
{"type_id": "7", "type_name": "星选"},
|
||||
{"type_id": "5", "type_name": "阳光"}],
|
||||
}
|
||||
|
||||
return result
|
||||
|
||||
def homeVideoContent(self):
|
||||
videos = []
|
||||
|
||||
url= f'{xurl}/v1/theater/home_page?theater_class_id=1&class2_id=4&page_num=1&page_size=24'
|
||||
detail = requests.get(url=url, headers=headerx)
|
||||
detail.encoding = "utf-8"
|
||||
if detail.status_code == 200:
|
||||
data = detail.json()
|
||||
|
||||
for vod in data['data']['list']:
|
||||
|
||||
name = vod['theater']['title']
|
||||
|
||||
id = vod['theater']['id']
|
||||
|
||||
pic = vod['theater']['cover_url']
|
||||
|
||||
remark = vod['theater']['play_amount_str']
|
||||
|
||||
video = {
|
||||
"vod_id": id,
|
||||
"vod_name": name,
|
||||
"vod_pic": pic,
|
||||
"vod_remarks": remark
|
||||
}
|
||||
videos.append(video)
|
||||
|
||||
result = {'list': videos}
|
||||
return result
|
||||
|
||||
def categoryContent(self, cid, pg, filter, ext):
|
||||
result = {}
|
||||
videos = []
|
||||
|
||||
url = f'{xurl}/v1/theater/home_page?theater_class_id={cid}&page_num={pg}&page_size=24'
|
||||
detail = requests.get(url=url,headers=headerx)
|
||||
detail.encoding = "utf-8"
|
||||
if detail.status_code == 200:
|
||||
data = detail.json()
|
||||
|
||||
for vod in data['data']['list']:
|
||||
|
||||
name = vod['theater']['title']
|
||||
|
||||
id = vod['theater']['id']
|
||||
|
||||
pic = vod['theater']['cover_url']
|
||||
|
||||
remark = vod['theater']['theme']
|
||||
|
||||
video = {
|
||||
"vod_id": id,
|
||||
"vod_name": name,
|
||||
"vod_pic": pic,
|
||||
"vod_remarks": remark
|
||||
}
|
||||
videos.append(video)
|
||||
|
||||
result = {'list': videos}
|
||||
result['page'] = pg
|
||||
result['pagecount'] = 9999
|
||||
result['limit'] = 90
|
||||
result['total'] = 999999
|
||||
return result
|
||||
|
||||
def detailContent(self, ids):
|
||||
did = ids[0]
|
||||
result = {}
|
||||
videos = []
|
||||
xianlu = ''
|
||||
bofang = ''
|
||||
|
||||
url = f'{xurl}/v2/theater_parent/detail?theater_parent_id={did}'
|
||||
detail = requests.get(url=url, headers=headerx)
|
||||
detail.encoding = "utf-8"
|
||||
if detail.status_code == 200:
|
||||
data = detail.json()
|
||||
|
||||
url = 'https://fs-im-kefu.7moor-fs1.com/ly/4d2c3f00-7d4c-11e5-af15-41bf63ae4ea0/1732707176882/jiduo.txt'
|
||||
response = requests.get(url)
|
||||
response.encoding = 'utf-8'
|
||||
code = response.text
|
||||
name = self.extract_middle_text(code, "s1='", "'", 0)
|
||||
Jumps = self.extract_middle_text(code, "s2='", "'", 0)
|
||||
|
||||
content = '剧情:' + data['data']['introduction']
|
||||
|
||||
area = data['data']['desc_tags'][0]
|
||||
|
||||
remarks = data['data']['filing']
|
||||
|
||||
# 修复剧集只有一集的问题 - 检查theaters数据是否存在且不为空
|
||||
if 'theaters' in data['data'] and data['data']['theaters']:
|
||||
for sou in data['data']['theaters']:
|
||||
id = sou['son_video_url']
|
||||
name = sou['num']
|
||||
bofang = bofang + str(name) + '$' + id + '#'
|
||||
|
||||
bofang = bofang[:-1] if bofang.endswith('#') else bofang
|
||||
xianlu = '星芽'
|
||||
else:
|
||||
# 如果没有theaters数据,检查是否有单个视频URL
|
||||
if 'video_url' in data['data'] and data['data']['video_url']:
|
||||
bofang = '1$' + data['data']['video_url']
|
||||
xianlu = '星芽'
|
||||
else:
|
||||
bofang = Jumps
|
||||
xianlu = '1'
|
||||
|
||||
videos.append({
|
||||
"vod_id": did,
|
||||
"vod_content": content,
|
||||
"vod_remarks": remarks,
|
||||
"vod_area": area,
|
||||
"vod_play_from": xianlu,
|
||||
"vod_play_url": bofang
|
||||
})
|
||||
|
||||
result['list'] = videos
|
||||
return result
|
||||
|
||||
def playerContent(self, flag, id, vipFlags):
|
||||
|
||||
result = {}
|
||||
result["parse"] = 0
|
||||
result["playUrl"] = ''
|
||||
result["url"] = id
|
||||
result["header"] = headers
|
||||
return result
|
||||
|
||||
def searchContentPage(self, key, quick, page):
|
||||
result = {}
|
||||
videos = []
|
||||
|
||||
payload = {
|
||||
"text": key
|
||||
}
|
||||
|
||||
url = f"{xurl}/v3/search"
|
||||
detail = requests.post(url=url, headers=headerx, json=payload)
|
||||
if detail.status_code == 200:
|
||||
detail.encoding = "utf-8"
|
||||
data = detail.json()
|
||||
|
||||
for vod in data['data']['theater']['search_data']:
|
||||
|
||||
name = vod['title']
|
||||
|
||||
id = vod['id']
|
||||
|
||||
pic = vod['cover_url']
|
||||
|
||||
remark = vod['score_str']
|
||||
|
||||
video = {
|
||||
"vod_id": id,
|
||||
"vod_name": name,
|
||||
"vod_pic": pic,
|
||||
"vod_remarks": remark
|
||||
}
|
||||
videos.append(video)
|
||||
|
||||
result['list'] = videos
|
||||
result['page'] = page
|
||||
result['pagecount'] = 9999
|
||||
result['limit'] = 90
|
||||
result['total'] = 999999
|
||||
return result
|
||||
|
||||
def searchContent(self, key, quick, pg="1"):
|
||||
return self.searchContentPage(key, quick, '1')
|
||||
|
||||
def localProxy(self, params):
|
||||
if params['type'] == "m3u8":
|
||||
return self.proxyM3u8(params)
|
||||
elif params['type'] == "media":
|
||||
return self.proxyMedia(params)
|
||||
elif params['type'] == "ts":
|
||||
return self.proxyTs(params)
|
||||
return None
|
BIN
spider.jar
BIN
spider.jar
Binary file not shown.
Loading…
x
Reference in New Issue
Block a user