Release v2.2.4

[Perf] Tokens相关处理逻辑优化

.devcontainer/devcontainer.json

@@ -0,0 +1,34 @@
+// For format details, see https://aka.ms/devcontainer.json. For config options, see the
+// README at: https://github.com/devcontainers/templates/tree/main/src/python
+{
+	"name": "QChatGPT 3.10",
+	// Or use a Dockerfile or Docker Compose file. More info: https://containers.dev/guide/dockerfile
+	"image": "mcr.microsoft.com/devcontainers/python:0-3.10",
+
+	// Features to add to the dev container. More info: https://containers.dev/features.
+	// "features": {},
+
+	// Use 'forwardPorts' to make a list of ports inside the container available locally.
+	// "forwardPorts": [],
+
+	// Use 'postCreateCommand' to run commands after the container is created.
+	// "postCreateCommand": "pip3 install --user -r requirements.txt",
+
+	// Configure tool-specific properties.
+	// "customizations": {},
+	"customizations": {
+		"codespaces": {
+		  "repositories": {
+			"RockChinQ/QChatGPT": {
+			  "permissions": "write-all"
+			},
+			"RockChinQ/revLibs": {
+			  "permissions": "write-all"
+			}
+		  }
+		}
+	  }
+
+	// Uncomment to connect as root instead. More info: https://aka.ms/dev-containers-non-root.
+	// "remoteUser": "root"
+}

.github/pull_request_template.md

@@ -0,0 +1,22 @@
+### 概述
+
+实现/解决/优化的内容: 
+
+### 事务
+
+- [ ] 已阅读仓库[贡献指引](../CONTRIBUTING.md)
+- [ ] 已与维护者在issues或其他平台沟通此PR大致内容
+
+### 功能
+
+- [ ] 已编写完善的配置文件字段说明（若有新增）
+- [ ] 已测试新功能
+
+### 兼容性
+
+- [ ] 已处理版本兼容性
+- [ ] 已处理插件兼容问题
+
+### 风险
+
+可能导致或已知的问题: 

.gitignore

@@ -13,4 +13,7 @@ sensitive.json
 temp/
 current_tag
 scenario/
-!scenario/default-template.json
+!scenario/default-template.json
+override.json
+cookies.json
+res/announcement_saved

README.md

@@ -1,8 +1,9 @@
 # QChatGPT🤖
-> 2023/3/3  官方接口疑似被墙，可考虑使用网络代理 [#198](https://github.com/RockChinQ/QChatGPT/issues/198)  
+
+> 2023/3/18 现已支持GPT-4 API（内测），请查看`config-template.py`中的`completion_api_params`  
+> 2023/3/15 逆向库已支持New Bing，使用方法查看[插件文档](https://github.com/RockChinQ/revLibs)  
+> 2023/3/15 逆向库已支持GPT-4模型，使用方法查看[插件](https://github.com/RockChinQ/revLibs)   
 > 2023/3/3  现已在主线支持官方ChatGPT接口，使用方法查看[#195](https://github.com/RockChinQ/QChatGPT/issues/195)  
-> 2023/3/2  OpenAI已发布ChatGPT官方接口，我们正在全力接入，预计明日前完成，请查看[此PR](https://github.com/RockChinQ/QChatGPT/pull/194)  
-> 2023/2/16 现已支持接入ChatGPT网页版，详情请完成部署并查看底部**插件**小节或[此仓库](https://github.com/RockChinQ/revLibs)
 
 - 到[项目Wiki](https://github.com/RockChinQ/QChatGPT/wiki)可了解项目详细信息
 - 由bilibili TheLazy制作的[视频教程](https://www.bilibili.com/video/BV15v4y1X7aP)
@@ -17,8 +18,11 @@
 ### 文字对话
 
 - OpenAI GPT-3.5模型(ChatGPT API), 本项目原生支持, 默认使用
-- OpenAI GPT-3模型, 本项目原生支持, 部署完成后前往config.py切换
-- ChatGPT网页版逆向API, 由[插件](https://github.com/RockChinQ/revLibs)接入
+- OpenAI GPT-3模型, 本项目原生支持, 部署完成后前往`config.py`切换
+- OpenAI GPT-4模型, 本项目原生支持, 目前需要您的账户通过OpenAI的内测申请, 请前往`config.py`切换
+- ChatGPT网页版GPT-3.5模型, 由[插件](https://github.com/RockChinQ/revLibs)接入
+- ChatGPT网页版GPT-4模型, 目前需要ChatGPT Plus订阅, 由[插件](https://github.com/RockChinQ/revLibs)接入
+- New Bing逆向库, 由[插件](https://github.com/RockChinQ/revLibs)接入
 
 ### 故事续写
 
@@ -32,6 +36,7 @@
 ### 语音生成
 
 - TTS+VITS, 由[插件](https://github.com/dominoar/QChatPlugins)接入
+- Plachta/VITS-Umamusume-voice-synthesizer, 由[插件](https://github.com/oliverkirk-sudo/chat_voice)接入
 
 ## ✅功能
 
@@ -106,6 +111,12 @@
     - “丢弃”策略：此分钟内对话次数达到限制时，丢弃之后的对话
   - 详细请查看config.py中的相关配置
 </details>
+<details>
+<summary>✅支持使用网络代理</summary>
+
+  - 目前已支持正向代理访问接口
+  - 详细请查看config.py中的`openai_config`的说明
+</details>
 
 详情请查看[Wiki功能使用页](https://github.com/RockChinQ/QChatGPT/wiki/%E5%8A%9F%E8%83%BD%E4%BD%BF%E7%94%A8#%E5%8A%9F%E8%83%BD%E7%82%B9%E5%88%97%E4%B8%BE)
 
@@ -117,7 +128,7 @@
 
 参考以下文章自行注册
 
-> [国内注册ChatGPT的方法(100%可用)](https://www.pythonthree.com/register-openai-chatgpt/)
+> [国内注册ChatGPT的方法(100%可用)](https://www.pythonthree.com/register-openai-chatgpt/)  
 > [手把手教你如何注册ChatGPT，超级详细](https://guxiaobei.com/51461)
 
 注册成功后请前往[个人中心查看](https://beta.openai.com/account/api-keys)api_key  
@@ -162,8 +173,7 @@ cd QChatGPT
 2. 安装依赖
 
 ```bash
-pip3 install yiri-mirai openai colorlog func_timeout
-pip3 install dulwich
+pip3 install yiri-mirai openai colorlog func_timeout dulwich Pillow
 ```
 
 3. 运行一次主程序，生成配置文件
@@ -218,18 +228,19 @@ python3 main.py
 - [hello_plugin](https://github.com/RockChinQ/hello_plugin) - `hello_plugin` 的储存库形式，插件开发模板
 - [dominoar/QChatPlugins](https://github.com/dominoar/QchatPlugins) - dominoar编写的诸多新功能插件（语言输出、Ranimg、屏蔽词规则等）
 - [dominoar/QCP-NovelAi](https://github.com/dominoar/QCP-NovelAi) - NovelAI 故事叙述与绘画
+- [oliverkirk-sudo/chat_voice](https://github.com/oliverkirk-sudo/chat_voice) - 文字转语音输出，使用HuggingFace上的[VITS-Umamusume-voice-synthesizer模型](https://huggingface.co/spaces/Plachta/VITS-Umamusume-voice-synthesizer)
+- [RockChinQ/WaitYiYan](https://github.com/RockChinQ/WaitYiYan) - 实时获取百度`文心一言`等待列表人数
+- [QChartGPT_Emoticon_Plugin](https://github.com/chordfish-k/QChartGPT_Emoticon_Plugin) - 使机器人根据回复内容发送表情包
 
 ## 😘致谢
 
 - [@the-lazy-me](https://github.com/the-lazy-me) 为本项目制作[视频教程](https://www.bilibili.com/video/BV15v4y1X7aP)
 - [@mikumifa](https://github.com/mikumifa) 本项目Docker部署仓库开发者
 - [@dominoar](https://github.com/dominoar) 为本项目开发多种插件
-- [@hissincn](https://github.com/hissincn) 本项目贡献者
-- [@LINSTCL](https://github.com/LINSTCL)   GPT-3.5官方模型适配贡献者
-- [@Haibersut](https://github.com/Haibersut)  本项目贡献者
 - [@万神的星空](https://github.com/qq255204159)  整合包发行
+- [@ljcduo](https://github.com/ljcduo)  GPT-4 API内测账号提供
 
-以及其他所有为本项目提供支持的朋友们。
+以及所有[贡献者](https://github.com/RockChinQ/QChatGPT/graphs/contributors)和其他为本项目提供支持的朋友们。
 
 ## 👍赞赏
 

config-template.py

@@ -79,9 +79,33 @@ default_prompt = {
     "default": "如果我之后想获取帮助，请你说“输入!help获取帮助”",
 }
 
-# 实验性设置项: JSON完整情景导入
-# 预设prompt模式
+# 情景预设格式
 # 参考值：旧版本方式：default | 完整情景：full_scenario
+# 旧版本的格式为上述default_prompt中的内容，或prompts目录下的文件名
+#
+# 完整情景预设的格式为JSON，在scenario目录下的JSON文件中列出对话的每个回合，编写方法见scenario/default-template.json
+# 编写方法例如：
+# {
+#   "prompt": [
+#       {
+#           "role": "user",
+#           "content": "之后当我需要帮助时，请说“输入!help获取帮助”"
+#       },{
+#           "role": "assistant",
+#           "content": "好的，当你之后需要帮助时，我会说“输入!help获取帮助”"
+#       },{
+#           "role": "user",
+#           "content": "帮助"
+#       },{
+#           "role": "assistant",
+#           "content": "输入!help获取帮助"
+#       }
+#   ]
+# }
+#
+# 您可以按照上述格式编写自己的情景预设，在prompt中列出对话的每个回合，
+# role为user或assistant，分别表示用户和机器人的回复
+# 每个JSON文件是一个情景预设，文件名即为情景预设的名称
 preset_mode = "default"
 
 # 群内响应规则
@@ -138,12 +162,16 @@ encourage_sponsor_at_start = True
 # 每次向OpenAI接口发送对话记录上下文的字符数
 # 最大不超过(4096 - max_tokens)个字符，max_tokens为下方completion_api_params中的max_tokens
 # 注意：较大的prompt_submit_length会导致OpenAI账户额度消耗更快
-prompt_submit_length = 1024
+prompt_submit_length = 2048
 
 # OpenAI补全API的参数
 # 请在下方填写模型，程序自动选择接口
 # 现已支持的模型有：
 # 
+#    'gpt-4'
+#    'gpt-4-0314'
+#    'gpt-4-32k'
+#    'gpt-4-32k-0314'
 #    'gpt-3.5-turbo'
 #    'gpt-3.5-turbo-0301'
 #    'text-davinci-003'
@@ -155,10 +183,10 @@ prompt_submit_length = 1024
 #    'text-ada-001'
 #
 # 具体请查看OpenAI的文档: https://beta.openai.com/docs/api-reference/completions/create
+# 请将内容修改到config.py中，请勿修改此文件
 completion_api_params = {
     "model": "gpt-3.5-turbo",
     "temperature": 0.9,  # 数值越低得到的回答越理性，取值范围[0, 1]
-    "max_tokens": 1024,  # 每次获取OpenAI接口响应的文字量上限, 不高于4096
     "top_p": 1,  # 生成的文本的文本与要求的符合度, 取值范围[0, 1]
     "frequency_penalty": 0.2,
     "presence_penalty": 1.0,

generate_override_all.py

@@ -0,0 +1,23 @@
+# 使用config-template生成override.json的字段全集模板文件override-all.json
+# 关于override.json机制，请参考：https://github.com/RockChinQ/QChatGPT/pull/271
+import json
+import importlib
+
+
+template = importlib.import_module("config-template")
+output_json = {
+    "comment": "这是override.json支持的字段全集, 关于override.json机制, 请查看https://github.com/RockChinQ/QChatGPT/pull/271"
+}
+
+
+for k, v in template.__dict__.items():
+    if k.startswith("__"):
+        continue
+    # 如果是module
+    if type(v) == type(template):
+        continue
+    print(k, v, type(v))
+    output_json[k] = v
+
+with open("override-all.json", "w", encoding="utf-8") as f:
+    json.dump(output_json, f, indent=4, ensure_ascii=False)

main.py

@@ -1,4 +1,5 @@
 import importlib
+import json
 import os
 import shutil
 import threading
@@ -12,8 +13,8 @@ try:
 except ImportError:
     # 尝试安装
     import pkg.utils.pkgmgr as pkgmgr
-    pkgmgr.install_requirements("requirements.txt")
     try:
+        pkgmgr.install_requirements("requirements.txt")
         import colorlog
     except ImportError:
         print("依赖不满足,请查看 https://github.com/RockChinQ/qcg-installer/issues/15")
@@ -32,7 +33,7 @@ log_colors_config = {
     'INFO': 'white',
     'WARNING': 'yellow',
     'ERROR': 'red',
-    'CRITICAL': 'bold_red',
+    'CRITICAL': 'cyan',
 }
 
 
@@ -143,9 +144,22 @@ def main(first_time_init=False):
                 setattr(config, key, getattr(config_template, key))
                 logging.warning("[{}]不存在".format(key))
                 is_integrity = False
+
         if not is_integrity:
             logging.warning("配置文件不完整，请依据config-template.py检查config.py")
             logging.warning("以上配置已被设为默认值，将在5秒后继续启动... ")
+
+        # 检查override.json覆盖
+        if os.path.exists("override.json"):
+            override_json = json.load(open("override.json", "r", encoding="utf-8"))
+            for key in override_json:
+                if hasattr(config, key):
+                    setattr(config, key, override_json[key])
+                    logging.info("覆写配置[{}]为[{}]".format(key, override_json[key]))
+                else:
+                    logging.error("无法覆写配置[{}]为[{}]，该配置不存在，请检查override.json是否正确".format(key, override_json[key]))
+
+        if not is_integrity:
             time.sleep(5)
 
         import pkg.utils.context
@@ -293,13 +307,21 @@ def main(first_time_init=False):
     import pkg.utils.updater
     try:
         if pkg.utils.updater.is_new_version_available():
-            pkg.utils.context.get_qqbot_manager().notify_admin("新版本可用，请发送 !update 进行自动更新\n更新日志:\n{}".format("\n".join(pkg.utils.updater.get_rls_notes())))
+            logging.info("新版本可用，请发送 !update 进行自动更新\n更新日志:\n{}".format("\n".join(pkg.utils.updater.get_rls_notes())))
         else:
             logging.info("当前已是最新版本")
 
     except Exception as e:
         logging.warning("检查更新失败:{}".format(e))
 
+    try:
+        import pkg.utils.announcement as announcement
+        new_announcement = announcement.fetch_new()
+        if new_announcement != "":
+            logging.critical("[公告] {}".format(new_announcement))
+    except Exception as e:
+        logging.warning("获取公告失败:{}".format(e))
+
     return qqbot
 
 

override-all.json

@@ -0,0 +1,75 @@
+{
+    "comment": "这是override.json支持的字段全集, 关于override.json机制, 请查看https://github.com/RockChinQ/QChatGPT/pull/271",
+    "mirai_http_api_config": {
+        "adapter": "WebSocketAdapter",
+        "host": "localhost",
+        "port": 8080,
+        "verifyKey": "yirimirai",
+        "qq": 1234567890
+    },
+    "openai_config": {
+        "api_key": {
+            "default": "openai_api_key"
+        },
+        "http_proxy": null
+    },
+    "admin_qq": 0,
+    "default_prompt": {
+        "default": "如果我之后想获取帮助，请你说“输入!help获取帮助”"
+    },
+    "preset_mode": "default",
+    "response_rules": {
+        "at": true,
+        "prefix": [
+            "/ai",
+            "!ai",
+            "！ai",
+            "ai"
+        ],
+        "regexp": [],
+        "random_rate": 0.0
+    },
+    "ignore_rules": {
+        "prefix": [
+            "/"
+        ],
+        "regexp": []
+    },
+    "income_msg_check": false,
+    "sensitive_word_filter": true,
+    "baidu_check": false,
+    "baidu_api_key": "",
+    "baidu_secret_key": "",
+    "inappropriate_message_tips": "[百度云]请珍惜机器人，当前返回内容不合规",
+    "encourage_sponsor_at_start": true,
+    "prompt_submit_length": 1024,
+    "completion_api_params": {
+        "model": "gpt-3.5-turbo",
+        "temperature": 0.9,
+        "top_p": 1,
+        "frequency_penalty": 0.2,
+        "presence_penalty": 1.0
+    },
+    "image_api_params": {
+        "size": "256x256"
+    },
+    "quote_origin": true,
+    "include_image_description": true,
+    "process_message_timeout": 30,
+    "show_prefix": false,
+    "blob_message_threshold": 256,
+    "blob_message_strategy": "forward",
+    "font_path": "",
+    "retry_times": 3,
+    "hide_exce_info_to_user": false,
+    "alter_tip_message": "出错了，请稍后再试",
+    "pool_num": 10,
+    "session_expire_time": 1200,
+    "rate_limitation": 60,
+    "rate_limit_strategy": "wait",
+    "rate_limit_drop_tip": "本分钟对话次数超过限速次数，此对话被丢弃",
+    "upgrade_dependencies": true,
+    "report_usage": true,
+    "logging_level": 20,
+    "help_message": "此机器人通过调用OpenAI的GPT-3大型语言模型生成回复，不具有情感。\n你可以用自然语言与其交流，回复的消息中[GPT]开头的为模型生成的语言，[bot]开头的为程序提示。\n了解此项目请找QQ 1010553892 联系作者\n请不要用其生成整篇文章或大段代码，因为每次只会向模型提交少部分文字，生成大部分文字会产生偏题、前后矛盾等问题\n每次会话最后一次交互后20分钟后会自动结束，结束后将开启新会话，如需继续前一次会话请发送 !last 重新开启\n欢迎到github.com/RockChinQ/QChatGPT 给个star\n\n指令帮助信息请查看: https://github.com/RockChinQ/QChatGPT/wiki/%E5%8A%9F%E8%83%BD%E4%BD%BF%E7%94%A8#%E6%9C%BA%E5%99%A8%E4%BA%BA%E6%8C%87%E4%BB%A4"
+}

pkg/database/manager.py

@@ -54,20 +54,27 @@ class DatabaseManager:
             `last_interact_timestamp` bigint not null,
             `status` varchar(255) not null default 'on_going',
             `default_prompt` text not null default '',
-            `prompt` text not null
+            `prompt` text not null,
+            `token_counts` text not null default '[]'
         )
         """)
 
-        # 检查sessions表是否存在`default_prompt`字段
+        # 检查sessions表是否存在`default_prompt`字段, 检查是否存在`token_counts`字段
         self.__execute__("PRAGMA table_info('sessions')")
         columns = self.cursor.fetchall()
         has_default_prompt = False
+        has_token_counts = False
         for field in columns:
             if field[1] == 'default_prompt':
                 has_default_prompt = True
+            if field[1] == 'token_counts':
+                has_token_counts = True
+            if has_default_prompt and has_token_counts:
                 break
         if not has_default_prompt:
             self.__execute__("alter table `sessions` add column `default_prompt` text not null default ''")
+        if not has_token_counts:
+            self.__execute__("alter table `sessions` add column `token_counts` text not null default '[]'")
 
 
         self.__execute__("""
@@ -89,7 +96,7 @@ class DatabaseManager:
 
     # session持久化
     def persistence_session(self, subject_type: str, subject_number: int, create_timestamp: int,
-                            last_interact_timestamp: int, prompt: str, default_prompt: str = ''):
+                            last_interact_timestamp: int, prompt: str, default_prompt: str = '', token_counts: str = ''):
         """持久化指定session"""
 
         # 检查是否已经有了此name和create_timestamp的session
@@ -102,20 +109,20 @@ class DatabaseManager:
         if count == 0:
 
             sql = """
-            insert into `sessions` (`name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `default_prompt`) 
-            values (?, ?, ?, ?, ?, ?, ?)
+            insert into `sessions` (`name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `default_prompt`, `token_counts`) 
+            values (?, ?, ?, ?, ?, ?, ?, ?)
             """
 
             self.__execute__(sql,
                              ("{}_{}".format(subject_type, subject_number), subject_type, subject_number, create_timestamp,
-                          last_interact_timestamp, prompt, default_prompt))
+                          last_interact_timestamp, prompt, default_prompt, token_counts))
         else:
             sql = """
-            update `sessions` set `last_interact_timestamp` = ?, `prompt` = ? 
+            update `sessions` set `last_interact_timestamp` = ?, `prompt` = ?, `token_counts` = ?
             where `type` = ? and `number` = ? and `create_timestamp` = ?
             """
 
-            self.__execute__(sql, (last_interact_timestamp, prompt, subject_type,
+            self.__execute__(sql, (last_interact_timestamp, prompt, token_counts, subject_type,
                                    subject_number, create_timestamp))
 
     # 显式关闭一个session
@@ -140,7 +147,7 @@ class DatabaseManager:
         # 从数据库中加载所有还没过期的session
         config = pkg.utils.context.get_config()
         self.__execute__("""
-        select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`
+        select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`, `token_counts`
         from `sessions` where `last_interact_timestamp` > {}
         """.format(int(time.time()) - config.session_expire_time))
         results = self.cursor.fetchall()
@@ -154,6 +161,7 @@ class DatabaseManager:
             prompt = result[5]
             status = result[6]
             default_prompt = result[7]
+            token_counts = result[8]
 
             # 当且仅当最后一个该对象的会话是on_going状态时，才会被加载
             if status == 'on_going':
@@ -163,7 +171,8 @@ class DatabaseManager:
                     'create_timestamp': create_timestamp,
                     'last_interact_timestamp': last_interact_timestamp,
                     'prompt': prompt,
-                    'default_prompt': default_prompt
+                    'default_prompt': default_prompt,
+                    'token_counts': token_counts
                 }
             else:
                 if session_name in sessions:
@@ -175,7 +184,7 @@ class DatabaseManager:
     def last_session(self, session_name: str, cursor_timestamp: int):
 
         self.__execute__("""
-        select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`
+        select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`, `token_counts`
         from `sessions` where `name` = '{}' and `last_interact_timestamp` < {} order by `last_interact_timestamp` desc
         limit 1
         """.format(session_name, cursor_timestamp))
@@ -192,6 +201,7 @@ class DatabaseManager:
         prompt = result[5]
         status = result[6]
         default_prompt = result[7]
+        token_counts = result[8]
 
         return {
             'subject_type': subject_type,
@@ -199,14 +209,15 @@ class DatabaseManager:
             'create_timestamp': create_timestamp,
             'last_interact_timestamp': last_interact_timestamp,
             'prompt': prompt,
-            'default_prompt': default_prompt
+            'default_prompt': default_prompt,
+            'token_counts': token_counts
         }
 
     # 获取此session_name后一个session的数据
     def next_session(self, session_name: str, cursor_timestamp: int):
 
         self.__execute__("""
-            select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`
+            select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`, `token_counts`
             from `sessions` where `name` = '{}' and `last_interact_timestamp` > {} order by `last_interact_timestamp` asc
             limit 1
             """.format(session_name, cursor_timestamp))
@@ -223,6 +234,7 @@ class DatabaseManager:
         prompt = result[5]
         status = result[6]
         default_prompt = result[7]
+        token_counts = result[8]
 
         return {
             'subject_type': subject_type,
@@ -230,13 +242,14 @@ class DatabaseManager:
             'create_timestamp': create_timestamp,
             'last_interact_timestamp': last_interact_timestamp,
             'prompt': prompt,
-            'default_prompt': default_prompt
+            'default_prompt': default_prompt,
+            'token_counts': token_counts
         }
 
     # 列出与某个对象的所有对话session
     def list_history(self, session_name: str, capacity: int, page: int):
         self.__execute__("""
-        select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`
+        select `name`, `type`, `number`, `create_timestamp`, `last_interact_timestamp`, `prompt`, `status`, `default_prompt`, `token_counts`
         from `sessions` where `name` = '{}' order by `last_interact_timestamp` desc limit {} offset {}
         """.format(session_name, capacity, capacity * page))
         results = self.cursor.fetchall()
@@ -250,6 +263,7 @@ class DatabaseManager:
             prompt = result[5]
             status = result[6]
             default_prompt = result[7]
+            token_counts = result[8]
 
             sessions.append({
                 'subject_type': subject_type,
@@ -257,7 +271,8 @@ class DatabaseManager:
                 'create_timestamp': create_timestamp,
                 'last_interact_timestamp': last_interact_timestamp,
                 'prompt': prompt,
-                'default_prompt': default_prompt
+                'default_prompt': default_prompt,
+                'token_counts': token_counts
             })
 
         return sessions

pkg/openai/manager.py

@@ -34,7 +34,7 @@ class OpenAIInteract:
         pkg.utils.context.set_openai_manager(self)
 
     # 请求OpenAI Completion
-    def request_completion(self, prompts) -> str:
+    def request_completion(self, prompts) -> tuple[str, int]:
         """请求补全接口回复
 
         Parameters:
@@ -60,14 +60,18 @@ class OpenAIInteract:
 
         logging.debug("OpenAI response: %s", response)
 
+        # 记录使用量
+        current_round_token = 0
         if 'model' in config.completion_api_params:
             self.audit_mgr.report_text_model_usage(config.completion_api_params['model'],
                                                    ai.get_total_tokens())
+            current_round_token = ai.get_total_tokens()
         elif 'engine' in config.completion_api_params:
             self.audit_mgr.report_text_model_usage(config.completion_api_params['engine'],
                                                    response['usage']['total_tokens'])
+            current_round_token = response['usage']['total_tokens']
 
-        return ai.get_message()
+        return ai.get_message(), current_round_token
 
     def request_image(self, prompt) -> dict:
         """请求图片接口回复

pkg/openai/modelmgr.py

@@ -21,6 +21,10 @@ COMPLETION_MODELS = {
 CHAT_COMPLETION_MODELS = {
     'gpt-3.5-turbo',
     'gpt-3.5-turbo-0301',
+    'gpt-4',
+    'gpt-4-0314',
+    'gpt-4-32k',
+    'gpt-4-32k-0314'
 }
 
 EDIT_MODELS = {

pkg/openai/session.py

@@ -72,6 +72,7 @@ def load_sessions():
         temp_session.last_interact_timestamp = session_data[session_name]['last_interact_timestamp']
         try:
             temp_session.prompt = json.loads(session_data[session_name]['prompt'])
+            temp_session.token_counts = json.loads(session_data[session_name]['token_counts'])
         except Exception:
             temp_session.prompt = reset_session_prompt(session_name, session_data[session_name]['prompt'])
             temp_session.persistence()
@@ -106,6 +107,9 @@ class Session:
     prompt = []
     """使用list来保存会话中的回合"""
 
+    token_counts = []
+    """每个回合的token数量"""
+
     default_prompt = []
     """本session的默认prompt"""
 
@@ -146,6 +150,8 @@ class Session:
         self.name = name
         self.create_timestamp = int(time.time())
         self.last_interact_timestamp = int(time.time())
+        self.prompt = []
+        self.token_counts = []
         self.schedule()
 
         self.response_lock = threading.Lock()
@@ -209,9 +215,16 @@ class Session:
         config = pkg.utils.context.get_config()
         max_length = config.prompt_submit_length if hasattr(config, "prompt_submit_length") else 1024
 
+        prompts, counts = self.cut_out(text, max_length)
+
+        # 计算请求前的prompt数量
+        total_token_before_query = 0
+        for token_count in counts:
+            total_token_before_query += token_count
+
         # 向API请求补全
-        message = pkg.utils.context.get_openai_manager().request_completion(
-            self.cut_out(text, max_length),
+        message, total_token = pkg.utils.context.get_openai_manager().request_completion(
+            prompts,
         )
 
         # 成功获取，处理回复
@@ -228,6 +241,10 @@ class Session:
         self.prompt.append({'role': 'user', 'content': text})
         self.prompt.append({'role': 'assistant', 'content': res_ans})
 
+        # 向token_counts中添加本回合的token数量
+        self.token_counts.append(total_token-total_token_before_query)
+        logging.debug("本回合使用token: {}, session counts: {}".format(total_token-total_token_before_query, self.token_counts))
+
         if self.just_switched_to_exist_session:
             self.just_switched_to_exist_session = False
             self.set_ongoing()
@@ -244,39 +261,65 @@ class Session:
 
         question = self.prompt[-2]['content']
         self.prompt = self.prompt[:-2]
+        self.token_counts = self.token_counts[:-1]
 
         # 返回上一回合的问题
         return question
 
     # 构建对话体
-    def cut_out(self, msg: str, max_tokens: int) -> list:
-        """将现有prompt进行切割处理，使得新的prompt长度不超过max_tokens"""
-        # 如果用户消息长度超过max_tokens，直接返回
-        temp_prompt: list = []
-        temp_prompt += self.default_prompt
-        temp_prompt.append(
+    def cut_out(self, msg: str, max_tokens: int) -> tuple[list, list]:
+        """将现有prompt进行切割处理，使得新的prompt长度不超过max_tokens
+
+        :return: (新的prompt, 新的token_counts)
+        """
+
+        # 最终由三个部分组成
+        # - default_prompt         情景预设固定值
+        # - changable_prompts      可变部分, 此会话中的历史对话回合
+        # - current_question       当前问题
+
+        # 包装目前的对话回合内容
+        changable_prompts = []
+        changable_counts = []
+        # 倒着来, 遍历prompt的步长为2, 遍历tokens_counts的步长为1
+        changable_index = len(self.prompt) - 1
+        token_count_index = len(self.token_counts) - 1
+
+        packed_tokens = 0
+
+        print(self.prompt)
+
+        while changable_index >= 0 and token_count_index >= 0:
+            if packed_tokens + self.token_counts[token_count_index] > max_tokens:
+                break
+
+            changable_prompts.insert(0, self.prompt[changable_index])
+            changable_prompts.insert(0, self.prompt[changable_index - 1])
+            changable_counts.insert(0, self.token_counts[token_count_index])
+            packed_tokens += self.token_counts[token_count_index]
+
+            changable_index -= 2
+            token_count_index -= 1
+
+        # 将default_prompt和changable_prompts合并
+        result_prompt = self.default_prompt + changable_prompts
+
+        print(changable_prompts)
+
+        # 添加当前问题
+        result_prompt.append(
             {
                 'role': 'user',
                 'content': msg
             }
         )
 
-        token_count = 0
-        for item in temp_prompt:
-            token_count += len(item['content'])
+        logging.debug('cut_out: {}\nchangable section tokens: {}\npacked counts: {}\nsession counts: {}'.format(json.dumps(result_prompt, ensure_ascii=False, indent=4),
+                                                                             packed_tokens,
+                                                                             changable_counts,
+                                                                             self.token_counts))
 
-        # 倒序遍历prompt
-        for i in range(len(self.prompt) - 1, -1, -1):
-            if token_count >= max_tokens:
-                break
-
-            # 将prompt加到temp_prompt倒数第二个位置
-            temp_prompt.insert(len(self.default_prompt), self.prompt[i])
-            token_count += len(self.prompt[i]['content'])
-
-        logging.debug('cut_out: {}'.format(json.dumps(temp_prompt, ensure_ascii=False, indent=4)))
-
-        return temp_prompt
+        return result_prompt, changable_counts
 
     # 持久化session
     def persistence(self):
@@ -291,7 +334,7 @@ class Session:
         subject_number = int(name_spt[1])
 
         db_inst.persistence_session(subject_type, subject_number, self.create_timestamp, self.last_interact_timestamp,
-                                    json.dumps(self.prompt), json.dumps(self.default_prompt))
+                                    json.dumps(self.prompt), json.dumps(self.default_prompt), json.dumps(self.token_counts))
 
     # 重置session
     def reset(self, explicit: bool = False, expired: bool = False, schedule_new: bool = True, use_prompt: str = None):
@@ -314,6 +357,7 @@ class Session:
 
         self.default_prompt = self.get_default_prompt(use_prompt)
         self.prompt = []
+        self.token_counts = []
         self.create_timestamp = int(time.time())
         self.last_interact_timestamp = int(time.time())
         self.just_switched_to_exist_session = False
@@ -339,6 +383,7 @@ class Session:
             self.last_interact_timestamp = last_one['last_interact_timestamp']
             try:
                 self.prompt = json.loads(last_one['prompt'])
+                self.token_counts = json.loads(last_one['token_counts'])
             except json.decoder.JSONDecodeError:
                 self.prompt = reset_session_prompt(self.name, last_one['prompt'])
                 self.persistence()
@@ -359,6 +404,7 @@ class Session:
             self.last_interact_timestamp = next_one['last_interact_timestamp']
             try:
                 self.prompt = json.loads(next_one['prompt'])
+                self.token_counts = json.loads(next_one['token_counts'])
             except json.decoder.JSONDecodeError:
                 self.prompt = reset_session_prompt(self.name, next_one['prompt'])
                 self.persistence()

pkg/qqbot/blob.py

@@ -1,5 +1,4 @@
 # 长消息处理相关
-import logging
 import os
 import time
 import base64
@@ -67,7 +66,7 @@ def check_text(text: str) -> list:
     """检查文本是否为长消息，并转换成该使用的消息链组件"""
     if not hasattr(config, 'blob_message_threshold'):
         return [text]
-    
+
     if len(text) > config.blob_message_threshold:
         if not hasattr(config, 'blob_message_strategy'):
             raise AttributeError('未定义长消息处理策略')

pkg/qqbot/command.py

@@ -284,7 +284,8 @@ def process_command(session_name: str, text_message: str, mgr, config,
                                                                             int(image_count))
                 # 获取此key的额度
                 try:
-                    credit_data = credit.fetch_credit_data(api_keys[key_name])
+                    http_proxy = config.openai_config["http_proxy"] if "http_proxy" in config.openai_config else None
+                    credit_data = credit.fetch_credit_data(api_keys[key_name], http_proxy)
                     reply_str += " - 使用额度:{:.2f}/{:.2f}\n".format(credit_data['total_used'],credit_data['total_granted'])
                 except Exception as e:
                     logging.warning("获取额度失败:{}".format(e))

pkg/qqbot/message.py

@@ -1,6 +1,5 @@
 # 普通消息处理模块
 import logging
-import time
 import openai
 import pkg.utils.context
 import pkg.openai.session
@@ -64,7 +63,7 @@ def process_normal_message(text_message: str, mgr, config, launcher_type: str,
                 reply = event.get_return_value("reply")
 
             if not event.is_prevented_default():
-                reply = blob.check_text(prefix + text)
+                reply = [prefix + text]
         except openai.error.APIConnectionError as e:
             err_msg = str(e)
             if err_msg.__contains__('Error communicating with OpenAI'):
@@ -117,8 +116,7 @@ def process_normal_message(text_message: str, mgr, config, launcher_type: str,
                 reply = handle_exception("{}会话调用API失败:{}".format(session_name, e),
                                          "[bot]err:RateLimitError,请重试或联系作者，或等待修复")
         except openai.error.InvalidRequestError as e:
-            reply = handle_exception("{}API调用参数错误:{}\n\n这可能是由于config.py中的prompt_submit_length参数或"
-                             "completion_api_params中的max_tokens参数数值过大导致的，请尝试将其降低".format(
+            reply = handle_exception("{}API调用参数错误:{}\n".format(
                              session_name, e), "[bot]err:API调用参数错误，请联系管理员，或等待修复")
         except openai.error.ServiceUnavailableError as e:
             reply = handle_exception("{}API调用服务不可用:{}".format(session_name, e), "[bot]err:API调用服务不可用，请重试或联系管理员，或等待修复")

pkg/qqbot/process.py

@@ -26,6 +26,7 @@ import pkg.plugin.host as plugin_host
 import pkg.plugin.models as plugin_models
 import pkg.qqbot.ignore as ignore
 import pkg.qqbot.banlist as banlist
+import pkg.qqbot.blob as blob
 
 processing = []
 
@@ -157,6 +158,7 @@ def process_message(launcher_type: str, launcher_id: int, text_message: str, mes
                                                  reply[0][:min(100, len(reply[0]))] + (
                                                      "..." if len(reply[0]) > 100 else "")))
                 reply = [mgr.reply_filter.process(reply[0])]
+                reply = blob.check_text(reply[0])
             else:
                 logging.info("回复[{}]消息".format(session_name))
 

pkg/utils/announcement.py

@@ -0,0 +1,44 @@
+import base64
+import os
+
+import requests
+
+
+def read_latest() -> str:
+    resp = requests.get(
+        url="https://api.github.com/repos/RockChinQ/QChatGPT/contents/res/announcement",
+    )
+    obj_json = resp.json()
+    b64_content = obj_json["content"]
+    # 解码
+    content = base64.b64decode(b64_content).decode("utf-8")
+    return content
+
+
+def read_saved() -> str:
+    # 已保存的在res/announcement_saved
+    # 检查是否存在
+    if not os.path.exists("res/announcement_saved"):
+        with open("res/announcement_saved", "w") as f:
+            f.write("")
+
+    with open("res/announcement_saved", "r") as f:
+        content = f.read()
+
+    return content
+
+
+def write_saved(content: str):
+    # 已保存的在res/announcement_saved
+    with open("res/announcement_saved", "w") as f:
+        f.write(content)
+
+
+def fetch_new() -> str:
+    latest = read_latest()
+    saved = read_saved()
+    if latest.replace(saved, "").strip() == "":
+        return ""
+    else:
+        write_saved(latest)
+        return latest.replace(saved, "").strip()

pkg/utils/credit.py

@@ -1,13 +1,19 @@
 # OpenAI账号免费额度剩余查询
 import requests
 
-
-def fetch_credit_data(api_key: str) -> dict:
+def fetch_credit_data(api_key: str, http_proxy: str) -> dict:
     """OpenAI账号免费额度剩余查询"""
+    proxies = {
+        "http":http_proxy,
+        "https":http_proxy
+    } if http_proxy is not None else None
+
     resp = requests.get(
         url="https://api.openai.com/dashboard/billing/credit_grants",
         headers={
             "Authorization": "Bearer {}".format(api_key),
-        }
+        },
+        proxies=proxies
     )
+
     return resp.json()

requirements.txt

@@ -1,5 +1,5 @@
 requests~=2.28.1
-openai~=0.27.0
+openai~=0.27.2
 dulwich~=0.21.3
 colorlog~=6.6.0
 yiri-mirai~=0.2.6.1

res/announcement

@@ -0,0 +1 @@
+