EdgeGPT与bot-on-anything搭建个人new bing web版

EdgeGPT与bot-on-anything搭建个人new bing web版

作者：不会起名字的小五

PS：

本文环境为ubuntu18.04+python3.8

EdgeGPT版本为10.6

bot-on-anything版本为1.1.0

由于bot-on-anything 1.1.0版本不适用于EdgeGPT，所以对bot-on-anything进行了二次开发。二次开发的代码已经上传到csdn了，等审核通过后会发在评论区

二次开发代码包：https://download.csdn.net/download/weixin_43433778/87885491

待改进

1、提问后回复等待时间较长。

2、长回复会导致报错

0x01 账号申请与注册

1、打开edge注册账号

2、连接vpn后打开edge，点击右侧快捷工具栏里的new bing图标

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-OO4rS4eQ-1686287186158)(C:\Users\Lenovo\Desktop\本地搭建new bing\1686191247522.png)]

3、点击立即聊天，即可使用new bing。

0x02 虚拟机环境配置

系统：ubuntu 18.04

1、安装python3.11环境

sudo apt install python3.8

3、安装pip

sudo apt install python3.8-distutils \\安装pip依赖

wget https://bootstrap.pypa.io/get-pip.py \\下载 get-pip.py 脚本

sudo python3.8 get-pip.py \\运行pip安装脚本

3、安装selenium

pip install -U selenium

0x03 安装EdgeGPT

1、使用官网安装命令

python3.11 -m pip install EdgeGPT --upgrade

2、测试是否安装成功

python3.11 -m EdgeGPT.EdgeGPT -h

成功回显帮助列表即为成功

0x04 获取bing cookie

1、edge浏览器安装插件Cookie Editor

2、在new bing页面打开Cookie Editor,并导出cookie为json格式

3、回到系统桌面，创建名为cookie.json的文件，并将导出的cookie粘贴进文件，关闭保存文件。

0x05 测试EdgeGPT是否可以正常使用

1、输入命令 python3.8 -m EdgeGPT.EdgeGPT --cookie-file '/home/edgegpt/Desktop/cookie.json' --locale zh-CN

查看是否可以正常对话。

2、以下是参数列表

options:
  -h, --help            查看帮助
  --enter-once			似乎是开启学习模式？增加该参数后会将后续输入输出进行保存和学习
  --search-result		似乎是开启查询模式？但是好像开启与否对结果影响并不大
  --no-stream			禁用文本流输出。不建议使用该参数，会报错。大概率是因为返回数据超过限制
  --rich				应该是是否使用rich库的参数。对返回数据进行美观修饰，但是没什么太大用处
  --proxy PROXY         设置代理。Proxy URL (e.g. socks5://127.0.0.1:1080)
  --wss-link WSS_LINK   设置wss连接。WSS URL(e.g. wss://sydney.bing.com/sydney/ChatHub)
  --style {creative,balanced,precise}		设置回答风格{创新型，平衡型，精确型}
  --prompt PROMPT       设置一个prompt启动，感觉没什么用。prompt to start with
  --cookie-file COOKIE_FILE		设置cookie文件路径，必须参数。path to cookie file
  --history-file HISTORY_FILE	设置历史记录文件记录历史数据，非必要不添加。path to history file
  --locale LOCALE       设置语言，这个很好理解。your locale (e.g. en-US, zh-CN, en-IE, en-GB)

3、常用的启动命令

python3.8 -m EdgeGPT.EdgeGPT --style creative --cookie-file '/home/edgegpt/Desktop/cookie.json' --locale zh-CN \\以创新模式启动

python3.8 -m EdgeGPT.EdgeGPT --style balanced --cookie-file '/home/edgegpt/Desktop/cookie.json' --locale zh-CN \\以平衡模式启动

python3.8 -m EdgeGPT.EdgeGPT --style precise --cookie-file '/home/edgegpt/Desktop/cookie.json' --locale zh-CN \\以精确模式启动

0x06 创建web页面

1、在github上下载bot-on-anything

git clone https://github.com/zhayujie/bot-on-anything

2、进入bot-on-anything文件夹后，安装依赖，命令如下

pip3.8 install -r requirements.txt

3、复制config-template.json并创建config.json

cp config-template.json config.json

4、根据官方文档给出的模板对config.json进行配置

5、启动。

python3.8 app.py

0x07 牢来！

1、经过上面的测试，发现正常启动是没有问题的，但是提出问题之后ai回答会报出错误。错误如下

[ERROR][2023-06-08 06:38:33][log.py:47] - [MultiChannel] Start up failed on terminal: cannot import name 'Chatbot' from 'EdgeGPT' (/home/edgegpt/.local/lib/python3.8/site-packages/EdgeGPT/__init__.py) [ERROR][2023-06-08 06:38:33][log.py:45] - App startup failed! [ERROR][2023-06-08 06:38:33][log.py:50] - cannot import name 'Chatbot' from 'EdgeGPT' (/home/edgegpt/.local/lib/python3.8/site-packages/EdgeGPT/__init__.py) Traceback (most recent call last): File "app.py", line 42, in main start_process(channel_type[0], args.config) File "app.py", line 23, in start_process raise e File "app.py", line 20, in start_process channel.startup() File "/home/edgegpt/Desktop/bot-on-anything/channel/terminal/terminal_channel.py", line 21, in startup for res in super().build_reply_content(prompt, context): File "/home/edgegpt/Desktop/bot-on-anything/channel/channel.py", line 31, in build_reply_content return Bridge().fetch_reply_content(query, context) File "/home/edgegpt/Desktop/bot-on-anything/bridge/bridge.py", line 18, in fetch_reply_content return model_factory.create_bot(type).reply(query, context) File "/home/edgegpt/Desktop/bot-on-anything/model/model_factory.py", line 29, in create_bot from model.bing.new_bing_model import BingModel File "/home/edgegpt/Desktop/bot-on-anything/model/bing/new_bing_model.py", line 6, in <module> from EdgeGPT import Chatbot, ConversationStyle ImportError: cannot import name 'Chatbot' from 'EdgeGPT' (/home/edgegpt/.local/lib/python3.8/site-packages/EdgeGPT/__init__.py)

2、经过后续研究，发现是由于EdgeGPT中Chatbot类缺失导致的报错。

3、由于EdgeGPT更新速度较快，bot-on-anything更新较慢，所以猜测可能是bot-on-anything无法兼容新版EdgeGPT导致的

4、查看EdgeGPT项目更新日志，发现导致问题的原因为EdgeGPT在10.0版本进行了重构，所以引用Chatbot类的命令不再是

from EdgeGPT import Chatbot而是 from EdgeGPT.EdgeGPT import Chatbot

5、先对model/bing下的两个代码进行修改改动如下：

jailbroken_sydney.py：

将

from EdgeGPT import ChatHubRequest, Chatbot, Conversation, ChatHub

改为

from EdgeGPT.EdgeGPT import ChatHubRequest, Chatbot, Conversation, ChatHub

new_bing_model.py:

将

from EdgeGPT import Chatbot, ConversationStyle

改为

from EdgeGPT.EdgeGPT import Chatbot, ConversationStyle

将

from ImageGen import ImageGen

改为

from EdgeGPT.ImageGen import ImageGen

6、修改之后出现新的报错

[ERROR][2023-06-08 07:12:19][log.py:47] - [MultiChannel] Start up failed on terminal: unsupported operand type(s) for |: 'type' and 'NoneType' [ERROR][2023-06-08 07:12:19][log.py:45] - App startup failed! [ERROR][2023-06-08 07:12:19][log.py:50] - unsupported operand type(s) for |: 'type' and 'NoneType' Traceback (most recent call last): File "app.py", line 42, in main start_process(channel_type[0], args.config) File "app.py", line 23, in start_process raise e File "app.py", line 20, in start_process channel.startup() File "/home/edgegpt/Desktop/bot-on-anything/channel/terminal/terminal_channel.py", line 21, in startup for res in super().build_reply_content(prompt, context): File "/home/edgegpt/Desktop/bot-on-anything/channel/channel.py", line 31, in build_reply_content return Bridge().fetch_reply_content(query, context) File "/home/edgegpt/Desktop/bot-on-anything/bridge/bridge.py", line 18, in fetch_reply_content return model_factory.create_bot(type).reply(query, context) File "/home/edgegpt/Desktop/bot-on-anything/model/model_factory.py", line 29, in create_bot from model.bing.new_bing_model import BingModel File "/home/edgegpt/Desktop/bot-on-anything/model/bing/new_bing_model.py", line 9, in <module> from model.bing.jailbroken_sydney import SydneyBot File "/home/edgegpt/Desktop/bot-on-anything/model/bing/jailbroken_sydney.py", line 13, in <module> class SydneyBot(Chatbot): File "/home/edgegpt/Desktop/bot-on-anything/model/bing/jailbroken_sydney.py", line 17, in SydneyBot cookies: dict | None = None, TypeError: unsupported operand type(s) for |: 'type' and 'NoneType'

这是一处语法错误，是由于python版本过低导致的，需进行如下修改

将

cookies: dict | None = None

改为

cookies: Optional[dict] = None

剩下的同类错误都是这样修改。

最后把jailbroken_sydney.py中的

from typing import Generator

改为

from typing import Optional, Generator

7、报错如下：

[ERROR][2023-06-08 07:32:24][log.py:47] - [MultiChannel] Start up failed on terminal: module 'EdgeGPT' has no attribute 'CONVERSATION_STYLE_TYPE' [ERROR][2023-06-08 07:32:24][log.py:45] - App startup failed! [ERROR][2023-06-08 07:32:24][log.py:50] - module 'EdgeGPT' has no attribute 'CONVERSATION_STYLE_TYPE' Traceback (most recent call last): File "app.py", line 42, in main start_process(channel_type[0], args.config) File "app.py", line 23, in start_process raise e File "app.py", line 20, in start_process channel.startup() File "/home/edgegpt/Desktop/bot-on-anything/channel/terminal/terminal_channel.py", line 21, in startup for res in super().build_reply_content(prompt, context): File "/home/edgegpt/Desktop/bot-on-anything/channel/channel.py", line 31, in build_reply_content return Bridge().fetch_reply_content(query, context) File "/home/edgegpt/Desktop/bot-on-anything/bridge/bridge.py", line 18, in fetch_reply_content return model_factory.create_bot(type).reply(query, context) File "/home/edgegpt/Desktop/bot-on-anything/model/model_factory.py", line 29, in create_bot from model.bing.new_bing_model import BingModel File "/home/edgegpt/Desktop/bot-on-anything/model/bing/new_bing_model.py", line 9, in <module> from model.bing.jailbroken_sydney import SydneyBot File "/home/edgegpt/Desktop/bot-on-anything/model/bing/jailbroken_sydney.py", line 13, in <module> class SydneyBot(Chatbot): File "/home/edgegpt/Desktop/bot-on-anything/model/bing/jailbroken_sydney.py", line 49, in SydneyBot conversation_style: EdgeGPT.CONVERSATION_STYLE_TYPE = None, AttributeError: module 'EdgeGPT' has no attribute 'CONVERSATION_STYLE_TYPE'

原因：没有导入CONVERSATION_STYLE_TYPE，CONVERSATION_STYLE_TYPE存在于EdgeGPT
中的conversation_style

添加一个新的引用：

from EdgeGPT import conversation_style

并将所有的

EdgeGPT.CONVERSATION_STYLE_TYPE

修改为

conversation_style.CONVERSATION_STYLE_TYPE

8、报错

[ERROR][2023-06-08 17:55:38][log.py:47] - [MultiChannel] Start up failed on terminal: 'type' object is not subscriptable [ERROR][2023-06-08 17:55:38][log.py:45] - App startup failed! [ERROR][2023-06-08 17:55:38][log.py:50] - 'type' object is not subscriptable Traceback (most recent call last): File "app.py", line 42, in main start_process(channel_type[0], args.config) File "app.py", line 23, in start_process raise e File "app.py", line 20, in start_process channel.startup() File "/home/edgegpt/Desktop/bot-on-anything/channel/terminal/terminal_channel.py", line 21, in startup for res in super().build_reply_content(prompt, context): File "/home/edgegpt/Desktop/bot-on-anything/channel/channel.py", line 31, in build_reply_content return Bridge().fetch_reply_content(query, context) File "/home/edgegpt/Desktop/bot-on-anything/bridge/bridge.py", line 18, in fetch_reply_content return model_factory.create_bot(type).reply(query, context) File "/home/edgegpt/Desktop/bot-on-anything/model/model_factory.py", line 29, in create_bot from model.bing.new_bing_model import BingModel File "/home/edgegpt/Desktop/bot-on-anything/model/bing/new_bing_model.py", line 16, in <module> class BingModel(Model): File "/home/edgegpt/Desktop/bot-on-anything/model/bing/new_bing_model.py", line 74, in BingModel def reply(self, query: str, context=None) -> tuple[str, dict]: TypeError: 'type' object is not subscriptable

原因：在 Python 3.8 及更早版本中，类型提示中使用的 tuple[str, dict] 语法是不支持的。要解决这个问题，你可以尝试使用 typing.Tuple 来替代类型提示中的 tuple

修改代码为：

from typing import Tuple

...

def reply(self, query: str, context=None) -> Tuple[str, dict]: ...

9、版本有点太老了，加上语言冲突，要改的东西很多，后续会直接把可以使用的环境包直接发出来，就不在把坑点写出来了。如果不嫌弃EdgeGPT版本过低，建议使用官方推荐版本。
Python 3.8 及更早版本中，类型提示中使用的 tuple[str, dict] 语法是不支持的。要解决这个问题，你可以尝试使用 typing.Tuple 来替代类型提示中的 tuple

修改代码为：

from typing import Tuple

...

def reply(self, query: str, context=None) -> Tuple[str, dict]: ...

9、版本有点太老了，加上语言冲突，要改的东西很多，后续会直接把可以使用的环境包直接发出来，就不在把坑点写出来了。如果不嫌弃EdgeGPT版本过低，建议使用官方推荐版本。