Release v2.5.2

[Perf] 优化函数调用的底层逻辑

.devcontainer/devcontainer.json

@@ -1,34 +0,0 @@
-// 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/workflows/sync-wiki.yml

@@ -1,7 +1,14 @@
 name: Update Wiki
 
 on:
+  pull_request:
+    branches:
+      - master
+    paths:
+      - 'res/wiki/**'
   push:
+    branches:
+      - master
     paths:
       - 'res/wiki/**'
 
@@ -23,11 +30,16 @@ jobs:
     - name: Copy res/wiki content to wiki
       run: |
         cp -r res/wiki/* wiki/
+    - name: Check for changes
+      run: |
+        cd wiki
+        if git diff --quiet; then
+          echo "No changes to commit."
+          exit 0
+        fi
     - name: Commit and Push Changes
       run: |
         cd wiki
-        if git diff --name-only; then
+        git add .
-          git add .
+        git commit -m "Update wiki"
-          git commit -m "Update wiki"
+        git push
-          git push
-        fi

.github/workflows/update-cmdpriv-template.yml

@@ -26,7 +26,7 @@ jobs:
     - name: Install dependencies
       run: |
         python -m pip install --upgrade pip
-        python -m pip install --upgrade  yiri-mirai openai colorlog func_timeout dulwich Pillow
+        python -m pip install --upgrade  yiri-mirai openai colorlog func_timeout dulwich Pillow CallingGPT tiktoken
 
     - name: Copy Scripts
       run: |

.gitignore

@@ -25,4 +25,6 @@ bin/
 .vscode
 test_*
 venv/
-hugchat.json
+hugchat.json
+qcapi
+/*.yaml

README.md

@@ -1,30 +1,39 @@
-# QChatGPT🤖
 
 <p align="center">
-<img src="res/social.png" alt="QChatGPT" width="640" />
+<img src="res/logo.png" alt="QChatGPT" width="120" />
 </p>
 
+<div align="center">
+
+# QChatGPT
+
+高稳定性/持续迭代/架构清晰/支持插件/高可自定义的 ChatGPT QQ机器人框架
+
 [English](README_en.md) | 简体中文
 
 [![GitHub release (latest by date)](https://img.shields.io/github/v/release/RockChinQ/QChatGPT?style=flat-square)](https://github.com/RockChinQ/QChatGPT/releases/latest)
 ![Wakapi Count](https://wakapi.dev/api/badge/RockChinQ/interval:any/project:QChatGPT)
+<img src="https://img.shields.io/badge/python-3.9+-blue.svg" alt="python">
+<a href="https://github.com/RockChinQ/QChatGPT/wiki">
+<img alt="Static Badge" src="https://img.shields.io/badge/%E6%9F%A5%E7%9C%8B-%E9%A1%B9%E7%9B%AEWiki-blue">
+</a>
+<a href="https://www.bilibili.com/video/BV1Y14y1Q7kQ">
+<img alt="Static Badge" src="https://img.shields.io/badge/%E8%A7%86%E9%A2%91%E6%95%99%E7%A8%8B-208647">
+</a>
+<a href="https://www.bilibili.com/video/BV11h4y1y74H">
+<img alt="Static Badge" src="https://img.shields.io/badge/Linux%E9%83%A8%E7%BD%B2%E8%A7%86%E9%A2%91-208647">
+</a>
 
-> 2023/4/27 正在对纯Python实现的QQ登录框架、YAML配置文件、异步编程等功能进行概念验证，欢迎体验[LightQChat](https://github.com/RockChinQ/LightQChat)项目  
+</div>
-> 2023/4/24 支持使用go-cqhttp登录QQ，请查看[此文档](https://github.com/RockChinQ/QChatGPT/wiki/go-cqhttp%E9%85%8D%E7%BD%AE)  
-> 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)  
 
-- 到[项目Wiki](https://github.com/RockChinQ/QChatGPT/wiki)可了解项目详细信息
+> 2023/7/29 支持使用GPT的Function Calling功能实现类似ChatGPT Plugin的效果，请见[Wiki内容函数](https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E4%BD%BF%E7%94%A8-%E5%86%85%E5%AE%B9%E5%87%BD%E6%95%B0)  
-- 官方交流、答疑群: 656285629  
+
-  - **进群提问前请您`确保`已经找遍文档和issue均无法解决**  
+<details>
-- 社区群（内有一键部署包、图形化界面等资源）: 362515018
+<summary>
-- QQ频道机器人见[QQChannelChatGPT](https://github.com/Soulter/QQChannelChatGPT)
-- 欢迎各种形式的贡献，请查看[贡献指引](CONTRIBUTING.md)
 
 ## 🍺模型适配一览
 
-<details>
+</summary>
-<summary>点击此处展开</summary>
 
 ### 文字对话
 
@@ -51,14 +60,16 @@
 - Plachta/VITS-Umamusume-voice-synthesizer, 由[插件](https://github.com/oliverkirk-sudo/chat_voice)接入
 
 
+安装[此插件](https://github.com/RockChinQ/Switcher)，即可在使用中切换文字模型。
+
 </details>
 
-安装[此插件](https://github.com/RockChinQ/Switcher)，即可在使用中切换文字模型。
+<details>
+<summary>
 
 ## ✅功能
 
-<details>
+</summary>
-<summary>点击此处展开概述</summary>
 
 <details>
 <summary>✅支持敏感词过滤，避免账号风险</summary>
@@ -108,6 +119,7 @@
 <summary>✅支持插件加载🧩</summary>
 
   - 自行实现插件加载器及相关支持
+  - 支持GPT的Function Calling功能
   - 详细查看[插件使用页](https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E4%BD%BF%E7%94%A8)
 </details>
 <details>
@@ -152,13 +164,24 @@
 <br/>
 <img alt="New Bing" src="res/screenshots/person_newbing.png" width="400"/>
 
+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)
+
 </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)
+<details>
+
+<summary>
 
 ## 🔩部署
 
+</summary>
+
 **部署过程中遇到任何问题，请先在[QChatGPT](https://github.com/RockChinQ/QChatGPT/issues)或[qcg-installer](https://github.com/RockChinQ/qcg-installer/issues)的issue里进行搜索**  
+**QChatGPT需要Python版本>=3.9**
+
+- 官方交流、答疑群: 656285629  
+  - **进群提问前请您`确保`已经找遍文档和issue均无法解决**  
+- 社区群（内有一键部署包、图形化界面等资源）: 891448839
 
 ### - 注册OpenAI账号
 
@@ -237,7 +260,7 @@ cd QChatGPT
 2. 安装依赖
 
 ```bash
-pip3 install requests yiri-mirai openai colorlog func_timeout dulwich Pillow nakuru-project-idk
+pip3 install requests yiri-mirai openai colorlog func_timeout dulwich Pillow nakuru-project-idk CallingGPT tiktoken
 ```
 
 3. 运行一次主程序，生成配置文件
@@ -266,32 +289,40 @@ python3 main.py
 
 </details>
 
+</details>
+
+<details>
+
+<summary>
+
 ## 🚀使用
 
+</summary>
+
 **部署完成后必看: [指令说明](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)**  
+
 所有功能查看[Wiki功能使用页](https://github.com/RockChinQ/QChatGPT/wiki/%E5%8A%9F%E8%83%BD%E4%BD%BF%E7%94%A8#%E4%BD%BF%E7%94%A8%E6%96%B9%E5%BC%8F)  
 
+</details>
+
+<details>
+<summary>
+
 ## 🧩插件生态
 
-现已支持自行开发插件对功能进行扩展或自定义程序行为  
+</summary>
-详见[Wiki插件使用页](https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E4%BD%BF%E7%94%A8)  
-开发教程见[Wiki插件开发页](https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E5%BC%80%E5%8F%91)
 
-<details>
+⭐我们已经支持了[GPT的Function Calling能力](https://platform.openai.com/docs/guides/gpt/function-calling)，请查看[Wiki内容函数](https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E4%BD%BF%E7%94%A8-%E5%86%85%E5%AE%B9%E5%87%BD%E6%95%B0)  
-<summary>查看插件列表</summary>
 
-### 示例插件
+> 使用方法见：[Wiki插件使用](https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E4%BD%BF%E7%94%A8)  
+> 开发教程见：[Wiki插件开发](https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E5%BC%80%E5%8F%91)
 
-在`tests/plugin_examples`目录下，将其整个目录复制到`plugins`目录下即可使用
 
-- `cmdcn` - 主程序指令中文形式
+[所有插件列表](https://github.com/stars/RockChinQ/lists/qchatgpt-%E6%8F%92%E4%BB%B6)，欢迎提出issue以提交新的插件
-- `hello_plugin` - 在收到消息`hello`时回复相应消息
-- `urlikethisijustsix` - 收到冒犯性消息时回复相应消息
 
-### 更多
+### 部分插件
-
-[插件列表](https://github.com/stars/RockChinQ/lists/qchatgpt-%E6%8F%92%E4%BB%B6)，欢迎提出issue以提交新的插件
 
+- [WebwlkrPlugin](https://github.com/RockChinQ/WebwlkrPlugin) - 让机器人能联网！！
 - [revLibs](https://github.com/RockChinQ/revLibs) - 将ChatGPT网页版接入此项目，关于[官方接口和网页版有什么区别](https://github.com/RockChinQ/QChatGPT/wiki/%E5%AE%98%E6%96%B9%E6%8E%A5%E5%8F%A3%E3%80%81ChatGPT%E7%BD%91%E9%A1%B5%E7%89%88%E3%80%81ChatGPT-API%E5%8C%BA%E5%88%AB)
 - [Switcher](https://github.com/RockChinQ/Switcher) - 支持通过指令切换使用的模型
 - [hello_plugin](https://github.com/RockChinQ/hello_plugin) - `hello_plugin` 的储存库形式，插件开发模板
@@ -305,9 +336,15 @@ python3 main.py
 - [SysStatPlugin](https://github.com/RockChinQ/SysStatPlugin) - 查看系统状态
 </details>
 
+<details>
+
+<summary>
+
 ## 😘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)
+</summary>
+
+- [@the-lazy-me](https://github.com/the-lazy-me) 为本项目制作[视频教程](https://www.bilibili.com/video/BV1Y14y1Q7kQ)
 - [@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
 - [@万神的星空](https://github.com/qq255204159)  整合包发行
@@ -315,6 +352,16 @@ python3 main.py
 
 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。
 
+</details>
+
+<details>
+
+<summary>
+
 ## 👍赞赏
 
+</summary>
+
 <img alt="赞赏码" src="res/mm_reward_qrcode_1672840549070.png" width="400" height="400"/>
+
+</details>

config-template.py

@@ -70,6 +70,11 @@ openai_config = {
     "reverse_proxy": None
 }
 
+# api-key切换策略
+# active：每次请求时都会切换api-key
+# passive：仅当api-key超额时才会切换api-key
+switch_strategy = "active"
+
 # [必需] 管理员QQ号，用于接收报错等通知及执行管理员级别指令
 # 支持多个管理员，可以使用list形式设置，例如：
 # admin_qq = [12345678, 87654321]
@@ -152,7 +157,6 @@ response_rules = {
 }
 
 
-
 # 消息忽略规则
 # 适用于私聊及群聊
 # 符合此规则的消息将不会被响应
@@ -204,11 +208,13 @@ auto_reset = True
 # 现已支持的模型有：
 # 
 #    'gpt-4'
-#    'gpt-4-0314'
+#    'gpt-4-0613'
 #    'gpt-4-32k'
-#    'gpt-4-32k-0314'
+#    'gpt-4-32k-0613'
 #    'gpt-3.5-turbo'
-#    'gpt-3.5-turbo-0301'
+#    'gpt-3.5-turbo-16k'
+#    'gpt-3.5-turbo-0613'
+#    'gpt-3.5-turbo-16k-0613'
 #    'text-davinci-003'
 #    'text-davinci-002'
 #    'code-davinci-002'
@@ -233,9 +239,17 @@ image_api_params = {
     "size": "256x256",  # 图片尺寸，支持256x256, 512x512, 1024x1024
 }
 
+# 跟踪函数调用
+# 为True时，在每次GPT进行Function Calling时都会输出发送一条回复给用户
+# 同时，一次提问内所有的Function Calling和普通回复消息都会单独发送给用户
+trace_function_calls = False
+
 # 群内回复消息时是否引用原消息
 quote_origin = True
 
+# 群内回复消息时是否at发送者
+at_sender = False
+
 # 回复绘图时是否包含图片描述
 include_image_description = True
 

main.py

@@ -47,7 +47,7 @@ def init_db():
 
 def ensure_dependencies():
     import pkg.utils.pkgmgr as pkgmgr
-    pkgmgr.run_pip(["install", "openai", "Pillow", "nakuru-project-idk", "--upgrade",
+    pkgmgr.run_pip(["install", "openai", "Pillow", "nakuru-project-idk", "CallingGPT", "tiktoken", "--upgrade",
                     "-i", "https://pypi.tuna.tsinghua.edu.cn/simple",
                     "--trusted-host", "pypi.tuna.tsinghua.edu.cn"])
 
@@ -60,35 +60,42 @@ def override_config():
     # 检查override.json覆盖
     if os.path.exists("override.json") and use_override:
         override_json = json.load(open("override.json", "r", encoding="utf-8"))
+        overrided = []
         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]))
+                # logging.info("覆写配置[{}]为[{}]".format(key, override_json[key]))
+                overrided.append(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 len(overrided) > 0:
+            logging.info("已根据override.json覆写配置项: {}".format(", ".join(overrided)))
 
 
 # 临时函数，用于加载config和上下文，未来统一放在config类
 def load_config():
     logging.info("检查config模块完整性.")
     # 完整性校验
+    non_exist_keys = []
+
     is_integrity = True
     config_template = importlib.import_module('config-template')
     config = importlib.import_module('config')
     for key in dir(config_template):
         if not key.startswith("__") and not hasattr(config, key):
             setattr(config, key, getattr(config_template, key))
-            logging.warning("[{}]不存在".format(key))
+            # logging.warning("[{}]不存在".format(key))
+            non_exist_keys.append(key)
             is_integrity = False
     
     if not is_integrity:
-        logging.warning("配置文件不完整，您可以依据config-template.py检查config.py")
+        logging.warning("以下配置字段不存在: {}".format(", ".join(non_exist_keys)))
 
     # 检查override.json覆盖
     override_config()
 
     if not is_integrity:
-        logging.warning("以上不存在的配置已被设为默认值，将在3秒后继续启动... ")
+        logging.warning("以上不存在的配置已被设为默认值，您可以依据config-template.py检查config.py，将在3秒后继续启动... ")
         time.sleep(3)
 
     # 存进上下文
@@ -97,6 +104,8 @@ def load_config():
 
 def complete_tips():
     """根据tips-custom-template模块补全tips模块的属性"""
+    non_exist_keys = []
+
     is_integrity = True
     logging.info("检查tips模块完整性.")
     tips_template = importlib.import_module('tips-custom-template')
@@ -104,10 +113,12 @@ def complete_tips():
     for key in dir(tips_template):
         if not key.startswith("__") and not hasattr(tips, key):
             setattr(tips, key, getattr(tips_template, key))
-            logging.warning("[{}]不存在".format(key))
+            # logging.warning("[{}]不存在".format(key))
+            non_exist_keys.append(key)
             is_integrity = False
 
     if not is_integrity:
+        logging.warning("以下提示语字段不存在: {}".format(", ".join(non_exist_keys)))
         logging.warning("tips模块不完整，您可以依据tips-custom-template.py检查tips.py")
         logging.warning("以上配置已被设为默认值，将在3秒后继续启动... ")
         time.sleep(3)
@@ -178,9 +189,14 @@ def start(first_time_init=False):
                 logging.error(e)
                 traceback.print_exc()
 
+            # 配置OpenAI proxy
+            import openai
+            openai.proxy = None  # 先重置，因为重载后可能需要清除proxy
+            if "http_proxy" in config.openai_config and config.openai_config["http_proxy"] is not None:
+                openai.proxy = config.openai_config["http_proxy"]
+
             # 配置openai api_base
             if "reverse_proxy" in config.openai_config and config.openai_config["reverse_proxy"] is not None:
-                import openai
                 openai.api_base = config.openai_config["reverse_proxy"]
 
             # 主启动流程
@@ -209,7 +225,7 @@ def start(first_time_init=False):
                 def run_bot_wrapper():
                     global known_exception_caught
                     try:
-                        logging.info("使用账号: {}".format(qqbot.bot_account_id))
+                        logging.debug("使用账号: {}".format(qqbot.bot_account_id))
                         qqbot.adapter.run_sync()
                     except TypeError as e:
                         if str(e).__contains__("argument 'debug'"):
@@ -313,7 +329,8 @@ def start(first_time_init=False):
         if pkg.utils.updater.is_new_version_available():
             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")
+            # logging.info("当前已是最新版本")
+            pass
 
     except Exception as e:
         logging.warning("检查更新失败:{}".format(e))

override-all.json

@@ -21,6 +21,7 @@
         "http_proxy": null,
         "reverse_proxy": null
     },
+    "switch_strategy": "active",
     "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”"
@@ -64,7 +65,9 @@
     "image_api_params": {
         "size": "256x256"
     },
+    "trace_function_calls": false,
     "quote_origin": true,
+    "at_sender": false,
     "include_image_description": true,
     "process_message_timeout": 30,
     "show_prefix": false,

pkg/openai/api/chat_completion.py

@@ -0,0 +1,204 @@
+import openai
+import json
+import logging
+
+from .model import RequestBase
+
+from ..funcmgr import get_func_schema_list, execute_function, get_func, get_func_schema, ContentFunctionNotFoundError
+
+
+class ChatCompletionRequest(RequestBase):
+    """调用ChatCompletion接口的请求类。
+    
+    此类保证每一次返回的角色为assistant的信息的finish_reason一定为stop。
+    若有函数调用响应，本类的返回瀑布是：函数调用请求->函数调用结果->...->assistant的信息->stop。
+    """
+    model: str
+    messages: list[dict[str, str]]
+    kwargs: dict
+
+    stopped: bool = False
+
+    pending_func_call: dict = None
+
+    pending_msg: str
+
+    def flush_pending_msg(self):
+        self.append_message(
+            role="assistant",
+            content=self.pending_msg
+        )
+        self.pending_msg = ""
+
+    def append_message(self, role: str, content: str, name: str=None, function_call: dict=None):
+        msg = {
+            "role": role,
+            "content": content
+        }
+
+        if name is not None:
+            msg['name'] = name
+
+        if function_call is not None:
+            msg['function_call'] = function_call
+
+        self.messages.append(msg)
+
+    def __init__(
+        self,
+        model: str,
+        messages: list[dict[str, str]],
+        **kwargs
+    ):
+        self.model = model
+        self.messages = messages.copy()
+
+        self.kwargs = kwargs
+
+        self.req_func = openai.ChatCompletion.acreate
+
+        self.pending_func_call = None
+
+        self.stopped = False
+
+        self.pending_msg = ""
+
+    def __iter__(self):
+        return self
+
+    def __next__(self) -> dict:
+        if self.stopped:
+            raise StopIteration()
+
+        if self.pending_func_call is None:  # 没有待处理的函数调用请求
+
+            args = {
+                "model": self.model,
+                "messages": self.messages,
+            }
+
+            funcs = get_func_schema_list()
+
+            if len(funcs) > 0:
+                args['functions'] = funcs
+
+            # 拼接kwargs
+            args = {**args, **self.kwargs}
+
+            resp = self._req(**args)
+
+            choice0 = resp["choices"][0]
+
+            # 如果不是函数调用，且finish_reason为stop，则停止迭代
+            if choice0['finish_reason'] == 'stop':  #  and choice0["finish_reason"] == "stop"
+                self.stopped = True
+            
+            if 'function_call' in choice0['message']:
+                self.pending_func_call = choice0['message']['function_call']
+
+                self.append_message(
+                    role="assistant",
+                    content=choice0['message']['content'],
+                    function_call=choice0['message']['function_call']
+                )
+
+                return {
+                    "id": resp["id"],
+                    "choices": [
+                        {
+                            "index": choice0["index"],
+                            "message": {
+                                "role": "assistant",
+                                "type": "function_call",
+                                "content": choice0['message']['content'],
+                                "function_call": choice0['message']['function_call']
+                            },
+                            "finish_reason": "function_call"
+                        }
+                    ],
+                    "usage": resp["usage"]
+                }
+            else:
+
+                # self.pending_msg += choice0['message']['content']
+                # 普通回复一定处于最后方，故不用再追加进内部messages
+
+                return {
+                    "id": resp["id"],
+                    "choices": [
+                        {
+                            "index": choice0["index"],
+                            "message": {
+                                "role": "assistant",
+                                "type": "text",
+                                "content": choice0['message']['content']
+                            },
+                            "finish_reason": choice0["finish_reason"]
+                        }
+                    ],
+                    "usage": resp["usage"]
+                }
+        else:  # 处理函数调用请求
+
+            cp_pending_func_call = self.pending_func_call.copy()
+            
+            self.pending_func_call = None
+
+            func_name = cp_pending_func_call['name']
+            arguments = {}
+
+            try:
+
+                try:
+                    arguments = json.loads(cp_pending_func_call['arguments'])
+                # 若不是json格式的异常处理
+                except json.decoder.JSONDecodeError:
+                    # 获取函数的参数列表
+                    func_schema = get_func_schema(func_name)
+
+                    arguments = {
+                        func_schema['parameters']['required'][0]: cp_pending_func_call['arguments']
+                    }
+                
+                logging.info("执行函数调用: name={}, arguments={}".format(func_name, arguments))
+
+                # 执行函数调用
+                ret = ""
+                try:
+                    ret = execute_function(func_name, arguments)
+
+                    logging.info("函数执行完成。")
+                except Exception as e:
+                    ret = "error: execute function failed: {}".format(str(e))
+                    logging.error("函数执行失败: {}".format(str(e)))
+
+                self.append_message(
+                    role="function",
+                    content=json.dumps(ret, ensure_ascii=False),
+                    name=func_name
+                )
+
+                return {
+                    "id": -1,
+                    "choices": [
+                        {
+                            "index": -1,
+                            "message": {
+                                "role": "function",
+                                "type": "function_return",
+                                "function_name": func_name,
+                                "content": json.dumps(ret, ensure_ascii=False)
+                            },
+                            "finish_reason": "function_return"
+                        }
+                    ],
+                    "usage": {
+                        "prompt_tokens": 0,
+                        "completion_tokens": 0,
+                        "total_tokens": 0
+                    }
+                }
+
+            except ContentFunctionNotFoundError:
+                raise Exception("没有找到函数: {}".format(func_name))
+

pkg/openai/api/completion.py

@@ -0,0 +1,111 @@
+import openai
+
+from .model import RequestBase
+
+
+class CompletionRequest(RequestBase):
+    """调用Completion接口的请求类。
+    
+    调用方可以一直next completion直到finish_reason为stop。
+    """
+
+    model: str
+    prompt: str
+    kwargs: dict
+
+    stopped: bool = False
+
+    def __init__(
+        self,
+        model: str,
+        messages: list[dict[str, str]],
+        **kwargs
+    ):
+        self.model = model
+        self.prompt = ""
+        
+        for message in messages:
+            self.prompt += message["role"] + ": " + message["content"] + "\n"
+        
+        self.prompt += "assistant: "
+
+        self.kwargs = kwargs
+
+        self.req_func = openai.Completion.acreate
+
+    def __iter__(self):
+        return self
+
+    def __next__(self) -> dict:
+        """调用Completion接口，返回生成的文本
+        
+        {
+            "id": "id",
+            "choices": [
+                {
+                    "index": 0,
+                    "message": {
+                        "role": "assistant",
+                        "type": "text",
+                        "content": "message"
+                    },
+                    "finish_reason": "reason"
+                }
+            ],
+            "usage": {
+                "prompt_tokens": 10,
+                "completion_tokens": 20,
+                "total_tokens": 30
+            }
+        }
+        """
+
+        if self.stopped:
+            raise StopIteration()
+
+        resp = self._req(
+            model=self.model,
+            prompt=self.prompt,
+            **self.kwargs
+        )
+
+        if resp["choices"][0]["finish_reason"] == "stop":
+            self.stopped = True
+
+        choice0 = resp["choices"][0]
+
+        self.prompt += choice0["text"]
+
+        return {
+            "id": resp["id"],
+            "choices": [
+                {
+                    "index": choice0["index"],
+                    "message": {
+                        "role": "assistant",
+                        "type": "text",
+                        "content": choice0["text"]
+                    },
+                    "finish_reason": choice0["finish_reason"]
+                }
+            ],
+            "usage": resp["usage"]
+        }
+        
+if __name__ == "__main__":
+    import os
+
+    openai.api_key = os.environ["OPENAI_API_KEY"]
+
+    for resp in CompletionRequest(
+        model="text-davinci-003",
+        messages=[
+            {
+                "role": "user",
+                "content": "Hello, who are you?"
+            }
+        ]
+    ):
+        print(resp)
+        if resp["choices"][0]["finish_reason"] == "stop":
+            break

pkg/openai/api/model.py

@@ -0,0 +1,62 @@
+# 定义不同接口请求的模型
+import threading
+import asyncio
+import logging
+
+import openai
+
+
+class RequestBase:
+
+    req_func: callable
+
+    def __init__(self, *args, **kwargs):
+        raise NotImplementedError
+
+    def _next_key(self):
+        import pkg.utils.context as context
+        switched, name = context.get_openai_manager().key_mgr.auto_switch()
+        logging.debug("切换api-key: switched={}, name={}".format(switched, name))
+        openai.api_key = context.get_openai_manager().key_mgr.get_using_key()
+
+    def _req(self, **kwargs):
+        """处理代理问题"""
+        import config
+
+        ret: dict = {}
+        exception: Exception = None
+
+        async def awrapper(**kwargs):
+            nonlocal ret, exception
+
+            try:
+                ret = await self.req_func(**kwargs)
+                logging.debug("接口请求返回：%s", str(ret))
+                
+                if config.switch_strategy == 'active':
+                    self._next_key()
+
+                return ret
+            except Exception as e:
+                exception = e
+            
+        loop = asyncio.new_event_loop()
+
+        thr = threading.Thread(
+            target=loop.run_until_complete,
+            args=(awrapper(**kwargs),)
+        )
+
+        thr.start()
+        thr.join()
+
+        if exception is not None:
+            raise exception
+
+        return ret
+
+    def __iter__(self):
+        raise self
+
+    def __next__(self):
+        raise NotImplementedError

pkg/openai/funcmgr.py

@@ -0,0 +1,47 @@
+# 封装了function calling的一些支持函数
+import logging
+
+
+from pkg.plugin import host
+
+
+class ContentFunctionNotFoundError(Exception):
+    pass
+
+
+def get_func_schema_list() -> list:
+    """从plugin包中的函数结构中获取并处理成受GPT支持的格式"""
+    if not host.__enable_content_functions__:
+        return []
+
+    schemas = []
+
+    for func in host.__callable_functions__:
+        if func['enabled']:
+            fun_cp = func.copy()
+
+            del fun_cp['enabled']
+
+            schemas.append(fun_cp)
+
+    return schemas
+
+def get_func(name: str) -> callable:
+    if name not in host.__function_inst_map__:
+        raise ContentFunctionNotFoundError("没有找到内容函数: {}".format(name))
+
+    return host.__function_inst_map__[name]
+
+def get_func_schema(name: str) -> dict:
+    for func in host.__callable_functions__:
+        if func['name'] == name:
+            return func
+    raise ContentFunctionNotFoundError("没有找到内容函数: {}".format(name))
+
+def execute_function(name: str, kwargs: dict) -> any:
+    """执行函数调用"""
+
+    logging.debug("executing function: name='{}', kwargs={}".format(name, kwargs))
+
+    func = get_func(name)
+    return func(**kwargs)

pkg/openai/keymgr.py

@@ -54,7 +54,24 @@ class KeysManager:
             是否切换成功, 切换后的api-key的别名
         """
 
+        index = 0
+
         for key_name in self.api_key:
+            if self.api_key[key_name] == self.using_key:
+                break
+            
+            index += 1
+
+        # 从当前key开始向后轮询
+        start_index = index
+        index += 1
+        if index >= len(self.api_key):
+            index = 0
+
+        while index != start_index:
+            
+            key_name = list(self.api_key.keys())[index]
+
             if self.api_key[key_name] not in self.exceeded:
                 self.using_key = self.api_key[key_name]
 
@@ -69,10 +86,14 @@ class KeysManager:
 
                 return True, key_name
 
-        self.using_key = list(self.api_key.values())[0]
+            index += 1
-        logging.info("使用api-key:" + list(self.api_key.keys())[0])
+            if index >= len(self.api_key):
+                index = 0
 
-        return False, ""
+        self.using_key = list(self.api_key.values())[start_index]
+        logging.debug("使用api-key:" + list(self.api_key.keys())[start_index])
+
+        return False, list(self.api_key.keys())[start_index]
 
     def add(self, key_name, key):
         self.api_key[key_name] = key

pkg/openai/manager.py

@@ -5,7 +5,9 @@ import openai
 import pkg.openai.keymgr
 import pkg.utils.context
 import pkg.audit.gatherer
-from pkg.openai.modelmgr import ModelRequest, create_openai_model_request
+from pkg.openai.modelmgr import select_request_cls
+
+from pkg.openai.api.model import RequestBase
 
 
 class OpenAIInteract:
@@ -27,51 +29,37 @@ class OpenAIInteract:
         self.key_mgr = pkg.openai.keymgr.KeysManager(api_key)
         self.audit_mgr = pkg.audit.gatherer.DataGatherer()
 
-        logging.info("文字总使用量：%d", self.audit_mgr.get_total_text_length())
+        # logging.info("文字总使用量：%d", self.audit_mgr.get_total_text_length())
 
         openai.api_key = self.key_mgr.get_using_key()
 
         pkg.utils.context.set_openai_manager(self)
 
-    # 请求OpenAI Completion
+    def request_completion(self, messages: list):
-    def request_completion(self, prompts) -> tuple[str, int]:
+        """请求补全接口回复=
-        """请求补全接口回复
-
-        Parameters:
-            prompts (str): 提示语
-
-        Returns:
-            str: 回复
         """
-
+        # 选择接口请求类
         config = pkg.utils.context.get_config()
 
-        # 根据模型选择使用的接口
+        request: RequestBase
-        ai: ModelRequest = create_openai_model_request(
-            config.completion_api_params['model'],
-            'user',
-            config.openai_config["http_proxy"] if "http_proxy" in config.openai_config else None
-        )
-        ai.request(
-            prompts,
-            **config.completion_api_params
-        )
-        response = ai.get_response()
 
-        logging.debug("OpenAI response: %s", response)
+        model: str = config.completion_api_params['model']
 
-        # 记录使用量
+        cp_parmas = config.completion_api_params.copy()
-        current_round_token = 0
+        del cp_parmas['model']
-        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(), current_round_token
+        request = select_request_cls(model, messages, cp_parmas)
+
+        # 请求接口
+        for resp in request:
+
+            if resp['usage']['total_tokens'] > 0:
+                self.audit_mgr.report_text_model_usage(
+                    model,
+                    resp['usage']['total_tokens']
+                )
+
+            yield resp
 
     def request_image(self, prompt) -> dict:
         """请求图片接口回复

pkg/openai/modelmgr.py

@@ -7,6 +7,11 @@ Completion - text-davinci-003 等模型
 """
 import openai, logging, threading, asyncio
 import openai.error as aiE
+import tiktoken
+
+from pkg.openai.api.model import RequestBase
+from pkg.openai.api.completion import CompletionRequest
+from pkg.openai.api.chat_completion import ChatCompletionRequest
 
 COMPLETION_MODELS = {
     'text-davinci-003',
@@ -20,11 +25,14 @@ COMPLETION_MODELS = {
 
 CHAT_COMPLETION_MODELS = {
     'gpt-3.5-turbo',
-    'gpt-3.5-turbo-0301',
+    'gpt-3.5-turbo-16k',
+    'gpt-3.5-turbo-0613',
+    'gpt-3.5-turbo-16k-0613',
+    # 'gpt-3.5-turbo-0301',
     'gpt-4',
-    'gpt-4-0314',
+    'gpt-4-0613',
     'gpt-4-32k',
-    'gpt-4-32k-0314'
+    'gpt-4-32k-0613'
 }
 
 EDIT_MODELS = {
@@ -36,153 +44,77 @@ IMAGE_MODELS = {
 }
 
 
-class ModelRequest:
+def select_request_cls(model_name: str, messages: list, args: dict) -> RequestBase:
-    """模型接口请求父类"""
-
-    can_chat = False
-    runtime: threading.Thread = None
-    ret = {}
-    proxy: str = None
-    request_ready = True
-    error_info: str = "若在没有任何错误的情况下看到这句话，请带着配置文件上报Issues"
-
-    def __init__(self, model_name, user_name, request_fun, http_proxy:str = None, time_out = None):
-        self.model_name = model_name
-        self.user_name = user_name
-        self.request_fun = request_fun
-        self.time_out = time_out
-        if http_proxy != None:
-            self.proxy = http_proxy
-            openai.proxy = self.proxy
-            self.request_ready = False
-
-    async def __a_request__(self, **kwargs):
-        """异步请求"""
-
-        try:
-            self.ret: dict = await self.request_fun(**kwargs)
-            self.request_ready = True
-        except aiE.APIConnectionError as e:
-            self.error_info = "{}\n请检查网络连接或代理是否正常".format(e)
-            raise ConnectionError(self.error_info)
-        except ValueError as e:
-            self.error_info = "{}\n该错误可能是由于http_proxy格式设置错误引起的"
-        except Exception as e:
-            self.error_info = "{}\n由于请求异常产生的未知错误，请查看日志".format(e)
-            raise type(e)(self.error_info)
-
-    def request(self, **kwargs):
-        """向接口发起请求"""
-
-        if self.proxy != None: #异步请求
-            self.request_ready = False
-            loop = asyncio.new_event_loop()
-            self.runtime = threading.Thread(
-                target=loop.run_until_complete,
-                args=(self.__a_request__(**kwargs),)
-            )
-            self.runtime.start()
-        else: #同步请求
-            self.ret = self.request_fun(**kwargs)
-
-    def __msg_handle__(self, msg):
-        """将prompt dict转换成接口需要的格式"""
-        return msg
-    
-    def ret_handle(self):
-        '''
-        API消息返回处理函数
-        若重写该方法，应检查异步线程状态，或在需要检查处super该方法
-        '''
-        if self.runtime != None and isinstance(self.runtime, threading.Thread):
-            self.runtime.join(self.time_out)
-        if self.request_ready:
-            return
-        raise Exception(self.error_info)
-
-    def get_total_tokens(self):
-        try:
-            return self.ret['usage']['total_tokens']
-        except:
-            return 0
-
-    def get_message(self):
-        return self.message
-
-    def get_response(self):
-        return self.ret
-
-
-class ChatCompletionModel(ModelRequest):
-    """ChatCompletion接口的请求实现"""
-
-    Chat_role = ['system', 'user', 'assistant']
-    def __init__(self, model_name, user_name, http_proxy:str = None, **kwargs):
-        if http_proxy == None:
-            request_fun = openai.ChatCompletion.create
-        else:
-            request_fun = openai.ChatCompletion.acreate
-        self.can_chat = True
-        super().__init__(model_name, user_name, request_fun, http_proxy, **kwargs)
-
-    def request(self, prompts, **kwargs):
-        prompts = self.__msg_handle__(prompts)
-        kwargs['messages'] = prompts
-        super().request(**kwargs)
-        self.ret_handle()
-
-    def __msg_handle__(self, msgs):
-        temp_msgs = []
-        # 把msgs拷贝进temp_msgs
-        for msg in msgs:
-            temp_msgs.append(msg.copy())
-        return temp_msgs
-
-    def get_message(self):
-        return self.ret["choices"][0]["message"]['content'] #需要时直接加载加快请求速度，降低内存消耗
-
-
-class CompletionModel(ModelRequest):
-    """Completion接口的请求实现"""
-
-    def __init__(self, model_name, user_name, http_proxy:str = None, **kwargs):
-        if http_proxy == None:
-            request_fun = openai.Completion.create
-        else:
-            request_fun = openai.Completion.acreate
-        super().__init__(model_name, user_name, request_fun, http_proxy, **kwargs)
-
-    def request(self, prompts, **kwargs):
-        prompts = self.__msg_handle__(prompts)
-        kwargs['prompt'] = prompts
-        super().request(**kwargs)
-        self.ret_handle()
-
-    def __msg_handle__(self, msgs):
-        prompt = ''
-        for msg in msgs:
-            prompt = prompt + "{}: {}\n".format(msg['role'], msg['content'])
-        # for msg in msgs:
-        #     if msg['role'] == 'assistant':
-        #         prompt = prompt + "{}\n".format(msg['content'])
-        #     else:
-        #         prompt = prompt + "{}:{}\n".format(msg['role'] , msg['content'])
-        prompt = prompt + "assistant: "
-        return prompt
-
-    def get_message(self):
-        return self.ret["choices"][0]["text"]
-
-
-def create_openai_model_request(model_name: str, user_name: str = 'user', http_proxy:str = None) -> ModelRequest:
-    """使用给定的模型名称创建模型请求对象"""
     if model_name in CHAT_COMPLETION_MODELS:
-        model = ChatCompletionModel(model_name, user_name, http_proxy)
+        return ChatCompletionRequest(model_name, messages, **args)
     elif model_name in COMPLETION_MODELS:
-        model = CompletionModel(model_name, user_name, http_proxy)
+        return CompletionRequest(model_name, messages, **args)
-    else :
+    raise ValueError("不支持模型[{}]，请检查配置文件".format(model_name))
-        log = "找不到模型[{}]，请检查配置文件".format(model_name)
+
-        logging.error(log)
+
-        raise IndexError(log)
+def count_chat_completion_tokens(messages: list, model: str) -> int:
-    logging.debug("使用接口[{}]创建模型请求[{}]".format(model.__class__.__name__, model_name))
+    """Return the number of tokens used by a list of messages."""
-    return model
+    try:
+        encoding = tiktoken.encoding_for_model(model)
+    except KeyError:
+        print("Warning: model not found. Using cl100k_base encoding.")
+        encoding = tiktoken.get_encoding("cl100k_base")
+    if model in {
+        "gpt-3.5-turbo-0613",
+        "gpt-3.5-turbo-16k-0613",
+        "gpt-4-0314",
+        "gpt-4-32k-0314",
+        "gpt-4-0613",
+        "gpt-4-32k-0613",
+        }:
+        tokens_per_message = 3
+        tokens_per_name = 1
+    elif model == "gpt-3.5-turbo-0301":
+        tokens_per_message = 4  # every message follows <|start|>{role/name}\n{content}<|end|>\n
+        tokens_per_name = -1  # if there's a name, the role is omitted
+    elif "gpt-3.5-turbo" in model:
+        # print("Warning: gpt-3.5-turbo may update over time. Returning num tokens assuming gpt-3.5-turbo-0613.")
+        return count_chat_completion_tokens(messages, model="gpt-3.5-turbo-0613")
+    elif "gpt-4" in model:
+        # print("Warning: gpt-4 may update over time. Returning num tokens assuming gpt-4-0613.")
+        return count_chat_completion_tokens(messages, model="gpt-4-0613")
+    else:
+        raise NotImplementedError(
+            f"""count_chat_completion_tokens() is not implemented for model {model}. See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens."""
+        )
+    num_tokens = 0
+    for message in messages:
+        num_tokens += tokens_per_message
+        for key, value in message.items():
+            num_tokens += len(encoding.encode(value))
+            if key == "name":
+                num_tokens += tokens_per_name
+    num_tokens += 3  # every reply is primed with <|start|>assistant<|message|>
+    return num_tokens
+
+
+def count_completion_tokens(messages: list, model: str) -> int:
+
+    try:
+        encoding = tiktoken.encoding_for_model(model)
+    except KeyError:
+        print("Warning: model not found. Using cl100k_base encoding.")
+        encoding = tiktoken.get_encoding("cl100k_base")
+
+    text = ""
+
+    for message in messages:
+        text += message['role'] + message['content'] + "\n"
+
+    text += "assistant: "
+
+    return len(encoding.encode(text))
+
+
+def count_tokens(messages: list, model: str):
+
+    if model in CHAT_COMPLETION_MODELS:
+        return count_chat_completion_tokens(messages, model)
+    elif model in COMPLETION_MODELS:
+        return count_completion_tokens(messages, model)
+    raise ValueError("不支持模型[{}]，请检查配置文件".format(model))

pkg/openai/pricing.bak.py

@@ -1,28 +0,0 @@
-# 计费模块
-# 已弃用 https://github.com/RockChinQ/QChatGPT/issues/81
-
-import logging
-
-pricing = {
-    "base": {  # 文字模型单位是1000字符
-        "text-davinci-003": 0.02,
-    },
-    "image": {
-        "256x256": 0.016,
-        "512x512": 0.018,
-        "1024x1024": 0.02,
-    }
-}
-
-
-def language_base_price(model, text):
-    salt_rate = 0.93
-    length = ((len(text.encode('utf-8')) - len(text)) / 2 + len(text)) * salt_rate
-    logging.debug("text length: %d" % length)
-
-    return pricing["base"][model] * length / 1000
-
-
-def image_price(size):
-    logging.debug("image size: %s" % size)
-    return pricing["image"][size]

pkg/openai/session.py

@@ -16,6 +16,8 @@ import pkg.utils.context
 import pkg.plugin.host as plugin_host
 import pkg.plugin.models as plugin_models
 
+from pkg.openai.modelmgr import count_tokens
+
 # 运行时保存的所有session
 sessions = {}
 
@@ -64,7 +66,7 @@ def load_sessions():
     session_data = db_inst.load_valid_sessions()
 
     for session_name in session_data:
-        logging.info('加载session: {}'.format(session_name))
+        logging.debug('加载session: {}'.format(session_name))
 
         temp_session = Session(session_name)
         temp_session.name = session_name
@@ -107,9 +109,6 @@ 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"""
 
@@ -195,8 +194,15 @@ class Session:
 
     # 请求回复
     # 这个函数是阻塞的
-    def append(self, text: str) -> str:
+    def query(self, text: str=None) -> tuple[str, str, list[str]]:
-        """向session中添加一条消息，返回接口回复"""
+        """向session中添加一条消息，返回接口回复
+        
+        Args:
+            text (str): 用户消息
+
+        Returns:
+            tuple[str, str]: (接口回复, finish_reason, 已调用的函数列表)
+        """
 
         self.last_interact_timestamp = int(time.time())
 
@@ -210,40 +216,142 @@ class Session:
 
             event = pkg.plugin.host.emit(plugin_models.SessionFirstMessageReceived, **args)
             if event.is_prevented_default():
-                return None
+                return None, None, None
 
         config = pkg.utils.context.get_config()
         max_length = config.prompt_submit_length
 
-        prompts, counts = self.cut_out(text, max_length)
+        local_default_prompt = self.default_prompt.copy()
+        local_prompt = self.prompt.copy()
+
+        # 触发PromptPreProcessing事件
+        args = {
+            'session_name': self.name,
+            'default_prompt': self.default_prompt,
+            'prompt': self.prompt,
+            'text_message': text,
+        }
+
+        event = pkg.plugin.host.emit(plugin_models.PromptPreProcessing, **args)
+        
+        if event.get_return_value('default_prompt') is not None:
+            local_default_prompt = event.get_return_value('default_prompt')
+        
+        if event.get_return_value('prompt') is not None:
+            local_prompt = event.get_return_value('prompt')
+
+        if event.get_return_value('text_message') is not None:
+            text = event.get_return_value('text_message')
+
+        prompts, _ = self.cut_out(text, max_length, local_default_prompt, local_prompt)
+
+        res_text = ""
+
+        pending_msgs = []
+
+        total_tokens = 0
+
+        finish_reason: str = ""
+
+        funcs = []
+
+        trace_func_calls = config.trace_function_calls
+        botmgr = pkg.utils.context.get_qqbot_manager()
+
+        session_name_spt: list[str] = self.name.split("_")
+
+        pending_res_text = ""
+
+        # TODO 对不起，我知道这样非常非常屎山，但我之后会重构的
+        for resp in pkg.utils.context.get_openai_manager().request_completion(prompts):
+
+            if pending_res_text != "":
+                botmgr.adapter.send_message(
+                    session_name_spt[0],
+                    session_name_spt[1],
+                    pending_res_text
+                )
+                pending_res_text = ""
+
+            finish_reason = resp['choices'][0]['finish_reason']
+
+            if resp['choices'][0]['message']['role'] == "assistant" and resp['choices'][0]['message']['content'] != None:  # 包含纯文本响应
+
+                if not trace_func_calls:
+                    res_text += resp['choices'][0]['message']['content'] + "\n"
+                else:
+                    res_text = resp['choices'][0]['message']['content']
+                    pending_res_text = resp['choices'][0]['message']['content']
+
+                total_tokens += resp['usage']['total_tokens']
+
+                msg = {
+                    "role": "assistant",
+                    "content": resp['choices'][0]['message']['content']
+                }
+
+                if 'function_call' in resp['choices'][0]['message']:
+                    msg['function_call'] = json.dumps(resp['choices'][0]['message']['function_call'])
+
+                pending_msgs.append(msg)
+
+            if resp['choices'][0]['message']['type'] == 'function_call':
+                # self.prompt.append(
+                #     {
+                #         "role": "assistant",
+                #         "content": "function call: "+json.dumps(resp['choices'][0]['message']['function_call'])
+                #     }
+                # )
+                if trace_func_calls:
+                    botmgr.adapter.send_message(
+                        session_name_spt[0],
+                        session_name_spt[1],
+                        "调用函数 "+resp['choices'][0]['message']['function_call']['name'] + "..."
+                    )
+
+                total_tokens += resp['usage']['total_tokens']
+            elif resp['choices'][0]['message']['type'] == 'function_return':
+                # self.prompt.append(
+                #     {
+                #         "role": "function",
+                #         "name": resp['choices'][0]['message']['function_name'],
+                #         "content": json.dumps(resp['choices'][0]['message']['content'])
+                #     }
+                # )
+
+                # total_tokens += resp['usage']['total_tokens']
+                funcs.append(
+                    resp['choices'][0]['message']['function_name']
+                )
+                pass
 
-        # 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, total_token = pkg.utils.context.get_openai_manager().request_completion(
+        # message, total_token = pkg.utils.context.get_openai_manager().request_completion(
-            prompts,
+        #     prompts,
-        )
+        # )
 
         # 成功获取，处理回复
-        res_test = message
+        # res_test = message
-        res_ans = res_test.strip()
+        res_ans = res_text.strip()
 
         # 将此次对话的双方内容加入到prompt中
-        self.prompt.append({'role': 'user', 'content': text})
+        # self.prompt.append({'role': 'user', 'content': text})
-        self.prompt.append({'role': 'assistant', 'content': res_ans})
+        # self.prompt.append({'role': 'assistant', 'content': res_ans})
+        if text:
+            self.prompt.append({'role': 'user', 'content': text})
+        # 添加pending_msgs
+        self.prompt += pending_msgs
 
         # 向token_counts中添加本回合的token数量
-        self.token_counts.append(total_token-total_token_before_query)
+        # self.token_counts.append(total_tokens-total_token_before_query)
-        logging.debug("本回合使用token: {}, session counts: {}".format(total_token-total_token_before_query, self.token_counts))
+        # logging.debug("本回合使用token: {}, session counts: {}".format(total_tokens-total_token_before_query, self.token_counts))
 
         if self.just_switched_to_exist_session:
             self.just_switched_to_exist_session = False
             self.set_ongoing()
 
-        return res_ans if res_ans[0] != '\n' else res_ans[1:]
+        return res_ans if res_ans[0] != '\n' else res_ans[1:], finish_reason, funcs
 
     # 删除上一回合并返回上一回合的问题
     def undo(self) -> str:
@@ -261,7 +369,7 @@ class Session:
         return question
 
     # 构建对话体
-    def cut_out(self, msg: str, max_tokens: int) -> tuple[list, list]:
+    def cut_out(self, msg: str, max_tokens: int, default_prompt: list, prompt: list) -> 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)
@@ -274,42 +382,35 @@ class Session:
 
         # 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
+        use_model = pkg.utils.context.get_config().completion_api_params['model']
 
-        while changable_index >= 0 and token_count_index >= 0:
+        ptr = len(prompt) - 1
-            if packed_tokens + self.token_counts[token_count_index] > max_tokens:
+
+        # 直接从后向前扫描拼接，不管是否是整回合
+        while ptr >= 0:
+            if count_tokens(prompt[ptr:ptr+1]+changable_prompts, use_model) > max_tokens:
                 break
 
-            changable_prompts.insert(0, self.prompt[changable_index])
+            changable_prompts.insert(0, prompt[ptr])
-            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
+            ptr -= 1
-            token_count_index -= 1
 
         # 将default_prompt和changable_prompts合并
-        result_prompt = self.default_prompt + changable_prompts
+        result_prompt = default_prompt + changable_prompts
 
         # 添加当前问题
-        result_prompt.append(
+        if msg:
-            {
+            result_prompt.append(
-                'role': 'user',
+                {
-                'content': msg
+                    'role': 'user',
-            }
+                    'content': msg
-        )
+                }
+            )
 
-        logging.debug('cut_out: {}\nchangable section tokens: {}\npacked counts: {}\nsession counts: {}'.format(json.dumps(result_prompt, ensure_ascii=False, indent=4),
+        logging.debug("cut_out: {}".format(json.dumps(result_prompt, ensure_ascii=False, indent=4)))
-                                                                             packed_tokens,
-                                                                             changable_counts,
-                                                                             self.token_counts))
 
-        return result_prompt, changable_counts
+        return result_prompt, count_tokens(changable_prompts, use_model)
 
     # 持久化session
     def persistence(self):
@@ -327,7 +428,7 @@ class Session:
                                     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):
+    def reset(self, explicit: bool = False, expired: bool = False, schedule_new: bool = True, use_prompt: str = None, persist: bool = False):
         if self.prompt:
             self.persistence()
             if explicit:
@@ -345,7 +446,8 @@ class Session:
             if expired:
                 pkg.utils.context.get_database_manager().set_session_expired(self.name, self.create_timestamp)
 
-        self.default_prompt = self.get_default_prompt(use_prompt)
+        if not persist:  # 不要求保持default prompt
+            self.default_prompt = self.get_default_prompt(use_prompt)
         self.prompt = []
         self.token_counts = []
         self.create_timestamp = int(time.time())

pkg/plugin/host.py

@@ -16,6 +16,8 @@ import pkg.qqbot.adapter as msadapter
 
 from mirai import Mirai
 
+from CallingGPT.session.session import Session
+
 __plugins__ = {}
 """插件列表
 
@@ -42,6 +44,15 @@ __plugins__ = {}
 __plugins_order__ = []
 """插件顺序"""
 
+__enable_content_functions__ = True
+"""是否启用内容函数"""
+
+__callable_functions__ = []
+"""供GPT调用的函数结构"""
+
+__function_inst_map__: dict[str, callable] = {}
+"""函数名:实例 映射"""
+
 
 def generate_plugin_order():
     """根据__plugin__生成插件初始顺序，无视是否启用"""
@@ -88,7 +99,7 @@ def walk_plugin_path(module, prefix='', path_prefix=''):
 
 def load_plugins():
     """加载插件"""
-    logging.info("加载插件")
+    logging.debug("加载插件")
     PluginHost()
     walk_plugin_path(__import__('plugins'))
 
@@ -102,20 +113,30 @@ def load_plugins():
     # 加载插件顺序
     settings.load_settings()
 
+    # 输出已注册的内容函数列表
+    logging.debug("registered content functions: {}".format(__callable_functions__))
+    logging.debug("function instance map: {}".format(__function_inst_map__))
+
 
 def initialize_plugins():
     """初始化插件"""
     logging.info("初始化插件")
     import pkg.plugin.models as models
+
+    successfully_initialized_plugins = []
+
     for plugin in iter_plugins():
         # if not plugin['enabled']:
         #     continue
         try:
             models.__current_registering_plugin__ = plugin['name']
             plugin['instance'] = plugin["class"](plugin_host=context.get_plugin_host())
-            logging.info("插件 {} 已初始化".format(plugin['name']))
+            # logging.info("插件 {} 已初始化".format(plugin['name']))
+            successfully_initialized_plugins.append(plugin['name'])
         except:
             logging.error("插件{}初始化时发生错误: {}".format(plugin['name'], sys.exc_info()))
+    
+    logging.info("以下插件已初始化: {}".format(", ".join(successfully_initialized_plugins)))
 
 
 def unload_plugins():
@@ -251,7 +272,7 @@ class EventContext:
             self.__return_value__[key] = []
         self.__return_value__[key].append(ret)
 
-    def get_return(self, key: str):
+    def get_return(self, key: str) -> list:
         """获取key的所有返回值"""
         if key in self.__return_value__:
             return self.__return_value__[key]
@@ -300,7 +321,9 @@ class PluginHost:
     """插件宿主"""
 
     def __init__(self):
+        """初始化插件宿主"""
         context.set_plugin_host(self)
+        self.calling_gpt_session = Session([])
 
     def get_runtime_context(self) -> context:
         """获取运行时上下文（pkg.utils.context模块的对象）
@@ -370,7 +393,7 @@ class PluginHost:
                         logging.debug("插件 {} 已要求阻止事件 {} 的默认行为".format(plugin['name'], event_name))
 
                 except Exception as e:
-                    logging.error("插件{}触发事件{}时发生错误".format(plugin['name'], event_name))
+                    logging.error("插件{}响应事件{}时发生错误".format(plugin['name'], event_name))
                     logging.error(traceback.format_exc())
 
             # print("done:{}".format(plugin['name']))

pkg/plugin/models.py

@@ -88,6 +88,8 @@ NormalMessageResponded = "normal_message_responded"
         session: pkg.openai.session.Session 会话对象
         prefix: str 回复文字消息的前缀
         response_text: str 响应文本
+        finish_reason: str 响应结束原因
+        funcs_called: list[str] 此次响应中调用的函数列表
     
     returns (optional):
         prefix: str 修改后的回复文字消息的前缀
@@ -132,18 +134,64 @@ KeySwitched = "key_switched"
         key_list: list[str] api-key列表
 """
 
+PromptPreProcessing = "prompt_pre_processing"
+"""每回合调用接口前对prompt进行预处理时触发，此事件不支持阻止默认行为
+    kwargs:
+        session_name: str 会话名称(<launcher_type>_<launcher_id>)
+        default_prompt: list 此session使用的情景预设内容
+        prompt: list 此session现有的prompt内容
+        text_message: str 用户发送的消息文本
+    
+    returns (optional):
+        default_prompt: list 修改后的情景预设内容
+        prompt: list 修改后的prompt内容
+        text_message: str 修改后的消息文本
+"""
 
-def on(event: str):
+
+def on(*args, **kwargs):
     """注册事件监听器
-    :param
-        event: str 事件名称
     """
-    return Plugin.on(event)
+    return Plugin.on(*args, **kwargs)
+
+def func(*args, **kwargs):
+    """注册内容函数，声明此函数为一个内容函数，在对话中将发送此函数给GPT以供其调用
+    此函数可以具有任意的参数，但必须按照[此文档](https://github.com/RockChinQ/CallingGPT/wiki/1.-Function-Format#function-format)
+    所述的格式编写函数的docstring。
+    此功能仅支持在使用gpt-3.5或gpt-4系列模型时使用。
+    """
+    return Plugin.func(*args, **kwargs)
 
 
 __current_registering_plugin__ = ""
 
 
+def require_ver(ge: str, le: str="v999.9.9") -> bool:
+    """插件版本要求装饰器
+
+    Args:
+        ge (str): 最低版本要求
+        le (str, optional): 最高版本要求
+
+    Returns:
+        bool: 是否满足要求, False时为无法获取版本号，True时为满足要求，报错为不满足要求
+    """
+    qchatgpt_version = ""
+    
+    from pkg.utils.updater import get_current_tag, compare_version_str
+
+    try:
+        qchatgpt_version = get_current_tag()  # 从updater模块获取版本号
+    except:
+        return False
+
+    if compare_version_str(qchatgpt_version, ge) < 0 or \
+        (compare_version_str(qchatgpt_version, le) > 0):
+        raise Exception("QChatGPT 版本不满足要求，某些功能（可能是由插件提供的）无法正常使用。（要求版本：{}-{}，但当前版本：{}）".format(ge, le, qchatgpt_version))
+
+    return True
+
+
 class Plugin:
     """插件基类"""
 
@@ -176,6 +224,34 @@ class Plugin:
 
         return wrapper
 
+    @classmethod
+    def func(cls, name: str=None):
+        """内容函数装饰器
+        """
+        global __current_registering_plugin__
+        from CallingGPT.entities.namespace import get_func_schema
+
+        def wrapper(func):
+
+            function_schema = get_func_schema(func)
+            function_schema['name'] = __current_registering_plugin__ + '-' + (func.__name__ if name is None else name)
+
+            function_schema['enabled'] = True
+
+            host.__function_inst_map__[function_schema['name']] = function_schema['function']
+
+            del function_schema['function']
+
+            # logging.debug("registering content function: p='{}', f='{}', s={}".format(__current_registering_plugin__, func, function_schema))
+
+            host.__callable_functions__.append(
+                function_schema
+            )
+
+            return func
+
+        return wrapper
+
 
 def register(name: str, description: str, version: str, author: str):
     """注册插件, 此函数作为装饰器使用

pkg/plugin/settings.py

@@ -8,7 +8,10 @@ import logging
 def wrapper_dict_from_runtime_context() -> dict:
     """从变量中包装settings.json的数据字典"""
     settings = {
-        "order": []
+        "order": [],
+        "functions": {
+            "enabled": host.__enable_content_functions__
+        }
     }
 
     for plugin_name in host.__plugins_order__:
@@ -22,6 +25,11 @@ def apply_settings(settings: dict):
     if "order" in settings:
         host.__plugins_order__ = settings["order"]
 
+    if "functions" in settings:
+        if "enabled" in settings["functions"]:
+            host.__enable_content_functions__ = settings["functions"]["enabled"]
+            # logging.debug("set content function enabled: {}".format(host.__enable_content_functions__))
+
 
 def dump_settings():
     """保存settings.json数据"""
@@ -78,6 +86,17 @@ def load_settings():
             settings["order"].append(plugin_name)
             settings_modified = True
 
+    if "functions" not in settings:
+        settings["functions"] = {
+            "enabled": host.__enable_content_functions__
+        }
+        settings_modified = True
+    elif "enabled" not in settings["functions"]:
+        settings["functions"]["enabled"] = host.__enable_content_functions__
+        settings_modified = True
+
+    logging.info("已全局{}内容函数。".format("启用" if settings["functions"]["enabled"] else "禁用"))
+
     apply_settings(settings)
 
     if settings_modified:

pkg/plugin/switch.py

@@ -28,6 +28,11 @@ def apply_switch(switch: dict):
     for plugin_name in switch:
         host.__plugins__[plugin_name]["enabled"] = switch[plugin_name]["enabled"]
 
+        # 查找此插件的所有内容函数
+        for func in host.__callable_functions__:
+            if func['name'].startswith(plugin_name + '-'):
+                func['enabled'] = switch[plugin_name]["enabled"]
+
 
 def dump_switch():
     """保存开关数据"""

pkg/qqbot/cmds/aamgr.py

@@ -327,6 +327,10 @@ def apply_privileges():
     for path, priv in data.items():
         if path == 'comment':
             continue
+        
+        if path not in __command_list__:
+            continue
+        
         if __command_list__[path]['privilege'] != priv:
             logging.debug('应用权限: {} -> {}(default: {})'.format(path, priv, __command_list__[path]['privilege']))
 

pkg/qqbot/cmds/funcs/func.py

@@ -0,0 +1,33 @@
+from ..aamgr import AbstractCommandNode, Context
+import logging
+
+import json
+
+
+@AbstractCommandNode.register(
+    parent=None,
+    name="func",
+    description="管理内容函数",
+    usage="!func",
+    aliases=[],
+    privilege=1
+)
+class FuncCommand(AbstractCommandNode):
+    @classmethod
+    def process(cls, ctx: Context) -> tuple[bool, list]:
+        from pkg.plugin.models import host
+
+        reply = []
+
+        reply_str = "当前已加载的内容函数：\n\n"
+
+        logging.debug("host.__callable_functions__: {}".format(json.dumps(host.__callable_functions__, indent=4)))
+
+        index = 1
+        for func in host.__callable_functions__:
+            reply_str += "{}. {}{}:\n{}\n\n".format(index, ("(已禁用) " if not func['enabled'] else ""), func['name'], func['description'])
+            index += 1
+            
+        reply = [reply_str]
+
+        return True, reply

pkg/qqbot/cmds/plugin/plugin.py

@@ -12,7 +12,7 @@ import pkg.utils.updater as updater
     description="插件管理",
     usage="!plugin\n!plugin get <插件仓库地址>\n!plugin update\n!plugin del <插件名>\n!plugin on <插件名>\n!plugin off <插件名>",
     aliases=[],
-    privilege=2
+    privilege=1
 )
 class PluginCommand(AbstractCommandNode):
     @classmethod
@@ -188,6 +188,11 @@ class PluginOnOffCommand(AbstractCommandNode):
             plugin_name = ctx.crt_params[0]
             if plugin_name in plugin_list:
                 plugin_list[plugin_name]['enabled'] = new_status
+
+                for func in plugin_host.__callable_functions__:
+                    if func['name'].startswith(plugin_name+"-"):
+                        func['enabled'] = new_status
+
                 plugin_switch.dump_switch()
                 reply = ["[bot]已{}插件: {}".format("启用" if new_status else "禁用", plugin_name)]
             else:

pkg/qqbot/manager.py

@@ -118,7 +118,7 @@ class QQBotManager:
         # 故只在第一次初始化时创建bot对象，重载之后使用原bot对象
         # 因此，bot的配置不支持热重载
         if first_time_init:
-            logging.info("Use adapter:" + config.msg_source_adapter)
+            logging.debug("Use adapter:" + config.msg_source_adapter)
             if config.msg_source_adapter == 'yirimirai':
                 from pkg.qqbot.sources.yirimirai import YiriMiraiAdapter
 
@@ -264,8 +264,24 @@ class QQBotManager:
         else:
             self.reply_filter = pkg.qqbot.filter.ReplyFilter([])
 
-    def send(self, event, msg, check_quote=True):
+    def send(self, event, msg, check_quote=True, check_at_sender=True):
         config = pkg.utils.context.get_config()
+        
+        if check_at_sender and config.at_sender:
+            msg.insert(
+                0,
+                Plain(" \n")
+            )
+
+            # 当回复的正文中包含换行时，quote可能会自带at，此时就不再单独添加at，只添加换行
+            if "\n" not in str(msg[1]) or config.msg_source_adapter == 'nakuru':
+                msg.insert(
+                    0,
+                    At(
+                        event.sender.id
+                    )
+                )
+
         self.adapter.reply_message(
             event,
             msg,
@@ -313,7 +329,7 @@ class QQBotManager:
                     reply = [tips_custom.reply_message]
 
         if reply:
-            return self.send(event, reply, check_quote=False)
+            return self.send(event, reply, check_quote=False, check_at_sender=False)
 
     # 群消息处理
     def on_group_message(self, event: GroupMessage):

pkg/qqbot/message.py

@@ -40,7 +40,7 @@ def process_normal_message(text_message: str, mgr, config, launcher_type: str,
         try:
             prefix = "[GPT]" if config.show_prefix else ""
 
-            text = session.append(text_message)
+            text, finish_reason, funcs = session.query(text_message)
 
             # 触发插件事件
             args = {
@@ -49,7 +49,9 @@ def process_normal_message(text_message: str, mgr, config, launcher_type: str,
                 "sender_id": sender_id,
                 "session": session,
                 "prefix": prefix,
-                "response_text": text
+                "response_text": text,
+                "finish_reason": finish_reason,
+                "funcs_called": funcs,
             }
 
             event = pkg.plugin.host.emit(plugin_models.NormalMessageResponded, **args)
@@ -62,10 +64,11 @@ def process_normal_message(text_message: str, mgr, config, launcher_type: str,
 
             if not event.is_prevented_default():
                 reply = [prefix + text]
+
         except openai.error.APIConnectionError as e:
             err_msg = str(e)
             if err_msg.__contains__('Error communicating with OpenAI'):
-                reply = handle_exception("{}会话调用API失败:{}\n请尝试关闭网络代理来解决此问题。".format(session_name, e),
+                reply = handle_exception("{}会话调用API失败:{}\n您的网络无法访问OpenAI接口或网络代理不正常".format(session_name, e),
                                          "[bot]err:调用API失败，请重试或联系管理员，或等待修复")
             else:
                 reply = handle_exception("{}会话调用API失败:{}".format(session_name, e), "[bot]err:调用API失败，请重试或联系管理员，或等待修复")
@@ -115,7 +118,7 @@ def process_normal_message(text_message: str, mgr, config, launcher_type: str,
                                          "[bot]err:RateLimitError,请重试或联系作者，或等待修复")
         except openai.error.InvalidRequestError as e:
             if config.auto_reset and "This model's maximum context length is" in str(e):
-                session.reset()
+                session.reset(persist=True)
                 reply = [tips_custom.session_auto_reset_message]
             else:
                 reply = handle_exception("{}API调用参数错误:{}\n".format(

pkg/qqbot/sources/yirimirai.py

@@ -1,6 +1,7 @@
 from ..adapter import MessageSourceAdapter
 import mirai
 import mirai.models.bus
+from mirai.bot import MiraiRunner
 
 import asyncio
 import typing
@@ -110,7 +111,12 @@ class YiriMiraiAdapter(MessageSourceAdapter):
         bus.unsubscribe(event_type, callback)
 
     def run_sync(self):
-        self.bot.run()
+        """运行YiriMirai"""
+
+        # 创建新的
+        loop = asyncio.new_event_loop()
+
+        loop.run_until_complete(MiraiRunner(self.bot)._run())
 
     def kill(self) -> bool:
         return False

pkg/utils/updater.py

@@ -78,6 +78,34 @@ def get_current_tag() -> str:
     return current_tag
 
 
+def compare_version_str(v0: str, v1: str) -> int:
+    """比较两个版本号"""
+
+    # 删除版本号前的v
+    if v0.startswith("v"):
+        v0 = v0[1:]
+    if v1.startswith("v"):
+        v1 = v1[1:]
+
+    v0:list = v0.split(".")
+    v1:list = v1.split(".")
+
+    # 如果两个版本号节数不同，把短的后面用0补齐
+    if len(v0) < len(v1):
+        v0.extend(["0"]*(len(v1)-len(v0)))
+    elif len(v0) > len(v1):
+        v1.extend(["0"]*(len(v0)-len(v1)))
+
+    # 从高位向低位比较
+    for i in range(len(v0)):
+        if int(v0[i]) > int(v1[i]):
+            return 1
+        elif int(v0[i]) < int(v1[i]):
+            return -1
+    
+    return 0
+
+
 def update_all(cli: bool = False) -> bool:
     """检查更新并下载源码"""
     current_tag = get_current_tag()

requirements.txt

@@ -1,10 +1,12 @@
 requests~=2.31.0
-openai~=0.27.7
+openai~=0.27.8
 dulwich~=0.21.5
 colorlog~=6.6.0
-yiri-mirai~=0.2.7
+yiri-mirai
 websockets
 urllib3~=1.26.10
 func_timeout~=4.3.5
 Pillow
-nakuru-project-idk
+nakuru-project-idk
+CallingGPT
+tiktoken

res/announcement.json

@@ -1,14 +1,8 @@
 [
     {
-        "id": 0,
+        "id": 2,
-        "time": "2023-04-24 16:05:20",
+        "time": "2023-08-01 10:49:26",
-        "timestamp": 1682323520,
+        "timestamp": 1690858166,
-        "content": "现已支持使用go-cqhttp替换mirai作为QQ登录框架, 请更新并查看 https://github.com/RockChinQ/QChatGPT/wiki/go-cqhttp%E9%85%8D%E7%BD%AE"
+        "content": "现已支持GPT函数调用功能，欢迎了解：https://github.com/RockChinQ/QChatGPT/wiki/%E6%8F%92%E4%BB%B6%E4%BD%BF%E7%94%A8-%E5%86%85%E5%AE%B9%E5%87%BD%E6%95%B0"
-    },
-    {
-        "id": 1,
-        "time": "2023-05-21 17:33:18",
-        "timestamp": 1684661598,
-        "content": "NewBing不再需要鉴权，若您正在使用revLibs逆向库插件，请立即使用!plugin update revLibs命令更新插件到最新版。"
     }
 ]

res/templates/cmdpriv-template.json

@@ -1,7 +1,8 @@
 {
     "comment": "以下为命令权限，请设置到cmdpriv.json中。关于此功能的说明，请查看：https://github.com/RockChinQ/QChatGPT/wiki/%E5%8A%9F%E8%83%BD%E4%BD%BF%E7%94%A8#%E5%91%BD%E4%BB%A4%E6%9D%83%E9%99%90%E6%8E%A7%E5%88%B6",
     "draw": 1,
-    "plugin": 2,
+    "func": 1,
+    "plugin": 1,
     "plugin.get": 2,
     "plugin.update": 2,
     "plugin.del": 2,

res/wiki/插件使用-内容函数.md

@@ -0,0 +1,31 @@
+> 说白了就是ChatGPT官方插件那种东西
+
+内容函数是基于[GPT的Function Calling能力](https://platform.openai.com/docs/guides/gpt/function-calling)实现的，这是一种嵌入对话中，由GPT自动调用的函数。  
+
+例如我们为GPT提供一个函数`access_the_web`，并提供其详细的描述以及其参数的描述，那么当我们在与GPT对话时涉及类似以下内容时：
+
+```
+Q: 请搜索一下github上有那些QQ机器人项目？
+Q: 请为我搜索一些不错的云服务商网站？
+Q：阅读并总结这篇文章：https://zhuanlan.zhihu.com/p/607570830
+Q：搜一下清远今天天气如何
+```
+
+GPT将会回复一个对`access_the_web`的函数调用请求，QChatGPT将自动处理执行该调用，并返回结果给GPT使其生成新的回复。
+
+当然，函数调用功能不止局限于网络访问，还可以实现图片处理、科学计算、行程规划等需要调用函数的功能，理论上我们可以通过内容函数实现与`ChatGPT Plugins`相同的功能。
+
+- 您需要使用`v2.5.0`以上的版本才能加载包含内容函数的插件
+- 您需要同时在`config.py`中的`completion_api_params`中设置`model`为支持函数调用的模型，推荐使用`gpt-3.5-turbo-16k`
+- 使用此功能可能会造成难以预期的账号余额消耗，请关注
+- [逆向库插件](https://github.com/RockChinQ/revLibs)现在也支持函数调用了..您可以在完全免费的情况下使用GPT-3.5进行函数调用，若您在主程序配置了内容函数并启用，逆向ChatGPT会自动使用这些函数
+
+### ？QChatGPT有什么类型的插件？区别是什么？
+
+QChatGPT具有`行为插件`和`内容函数`两种扩展方式，行为插件是完整的插件结构，是由运行期间的事件驱动的，内容函数被包含于一个完整的插件体中，由GPT接口驱动。
+
+> 还是不理解？可以尝试根据插件开发页的步骤自行编写插件
+
+## QChatGPT的一些不错的内容函数插件
+
+- [WebwlkrPlugin](https://github.com/RockChinQ/WebwlkrPlugin) - 让机器人能联网！！

res/wiki/插件使用.md

@@ -4,6 +4,8 @@ QChatGPT 插件使用Wiki
 
 `plugins`目录下的所有`.py`程序都将被加载，除了`__init__.py`之外的模块支持热加载
 
+> 插件分为`行为插件`和`内容插件`两种，行为插件由主程序运行中的事件驱动，内容插件由GPT生成的内容驱动，请查看内容插件页
+
 ## 安装
 
 ### 储存库克隆(推荐)
@@ -33,6 +35,8 @@ QChatGPT 插件使用Wiki
 !plugin del <插件名>       删除插件(需要管理员权限)
 !plugin on <插件名>        启用插件(需要管理员权限)
 !plugin off <插件名>       禁用插件(需要管理员权限)
+
+!func                      列出所有内容函数
 ```
 
 ### 控制插件执行顺序
@@ -42,4 +46,9 @@ QChatGPT 插件使用Wiki
 ### 启用或关闭插件
 
 无需卸载即可管理插件的开关  
-编辑`plugins`目录下的`switch.json`文件，将相应的插件的`enabled`字段设置为`true/false(开/关)`，之后重启程序或执行热重载即可控制插件开关
+编辑`plugins`目录下的`switch.json`文件，将相应的插件的`enabled`字段设置为`true/false(开/关)`，之后重启程序或执行热重载即可控制插件开关
+
+### 控制全局内容函数开关
+
+内容函数是基于[GPT的Function Calling能力](https://platform.openai.com/docs/guides/gpt/function-calling)实现的，这是一种嵌入对话中，由GPT自动调用的函数。  
+每个插件可以自行注册内容函数，您可以在`plugins`目录下的`settings.json`中设置`functions`下的`enabled`为`true`或`false`控制这些内容函数的启用或禁用。

res/wiki/插件开发.md

@@ -113,6 +113,199 @@ class HelloPlugin(Plugin):
 - 一个目录内可以存放多个Python程序文件，以独立出插件的各个功能，便于开发者管理，但不建议在一个目录内注册多个插件
 - 插件需要的依赖库请在插件目录下的`requirements.txt`中指定，程序从储存库获取此插件时将自动安装依赖
 
+## 🪝内容函数
+
+通过[GPT的Function Calling能力](https://platform.openai.com/docs/guides/gpt/function-calling)实现的`内容函数`，这是一种嵌入对话中，由GPT自动调用的函数。
+
+> 您的插件比一定必须包含内容函数，请先查看内容函数页了解此功能
+
+<details>
+<summary>示例：联网插件</summary>
+
+加载含有联网功能的内容函数的插件[WebwlkrPlugin](https://github.com/RockChinQ/WebwlkrPlugin)，向机器人询问在线内容
+
+```
+# 控制台输出
+[2023-07-29 17:37:18.698] message.py (26) - [INFO] : [person_1010553892]发送消息:介绍一下这个项目：https://git...
+[2023-07-29 17:37:21.292] util.py (67) - [INFO] : message='OpenAI API response' path=https://api.openai.com/v1/chat/completions processing_ms=1902 request_id=941afc13b2e1bba1e7877b92a970cdea response_code=200
+[2023-07-29 17:37:21.293] chat_completion.py (159) - [INFO] : 执行函数调用: name=Webwlkr-access_the_web, arguments={'url': 'https://github.com/RockChinQ/QChatGPT', 'brief_len': 512}
+[2023-07-29 17:37:21.848] chat_completion.py (164) - [INFO] : 函数执行完成。
+```
+
+![Webwlkr插件](https://github.com/RockChinQ/QChatGPT/blob/master/res/screenshots/webwlkr_plugin.png?raw=true)
+
+</details>
+
+### 内容函数编写步骤
+
+1️⃣ 请先按照上方步骤编写您的插件基础结构，现在请删除（当然你也可以不删，只是为了简洁）上述插件内容的诸个由`@on`装饰的类函数
+
+<details>
+<summary>删除后的结构</summary>
+
+```python
+from pkg.plugin.models import *
+from pkg.plugin.host import EventContext, PluginHost
+
+"""
+在收到私聊或群聊消息"hello"时，回复"hello, <发送者id>!"或"hello, everyone!"
+"""
+
+
+# 注册插件
+@register(name="Hello", description="hello world", version="0.1", author="RockChinQ")
+class HelloPlugin(Plugin):
+
+    # 插件加载时触发
+    # plugin_host (pkg.plugin.host.PluginHost) 提供了与主程序交互的一些方法，详细请查看其源码
+    def __init__(self, plugin_host: PluginHost):
+        pass
+
+    # 插件卸载时触发
+    def __del__(self):
+        pass
+```
+
+</details>
+
+2️⃣ 现在我们将以下函数添加到刚刚删除的函数的位置
+
+```Python
+
+# 要添加的函数
+
+@func(name="access_the_web")  # 设置函数名称
+def _(url: str):
+    """Call this function to search about the question before you answer any questions.
+    - Do not search through baidu.com at any time.
+    - If you need to search somthing, visit https://www.google.com/search?q=xxx.
+    - If user ask you to open a url (start with http:// or https://), visit it directly.
+    - Summary the plain content result by yourself, DO NOT directly output anything in the result you got.
+
+    Args:
+        url(str): url to visit
+
+    Returns:
+        str: plain text content of the web page
+    """
+    import requests
+    from bs4 import BeautifulSoup
+    # 你需要先使用
+    # pip install beautifulsoup4
+    # 安装依赖
+
+    r = requests.get(
+        url,
+        timeout=10,
+        headers={
+            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36 Edg/115.0.1901.183"
+        }
+    )
+    soup = BeautifulSoup(r.text, 'html.parser')
+
+    s = soup.get_text()
+
+    # 删除多余的空行或仅有\t和空格的行
+    s = re.sub(r'\n\s*\n', '\n', s)
+
+    if len(s) >= 512:  # 截取获取到的网页纯文本内容的前512个字
+        return s[:512]
+
+    return s
+
+```
+<details>
+<summary>现在这个文件内容应该是这样</summary>
+
+```python
+from pkg.plugin.models import *
+from pkg.plugin.host import EventContext, PluginHost
+
+"""
+在收到私聊或群聊消息"hello"时，回复"hello, <发送者id>!"或"hello, everyone!"
+"""
+
+
+# 注册插件
+@register(name="Hello", description="hello world", version="0.1", author="RockChinQ")
+class HelloPlugin(Plugin):
+
+    # 插件加载时触发
+    # plugin_host (pkg.plugin.host.PluginHost) 提供了与主程序交互的一些方法，详细请查看其源码
+    def __init__(self, plugin_host: PluginHost):
+        pass
+
+    @func(name="access_the_web")
+    def _(url: str):
+        """Call this function to search about the question before you answer any questions.
+        - Do not search through baidu.com at any time.
+        - If you need to search somthing, visit https://www.google.com/search?q=xxx.
+        - If user ask you to open a url (start with http:// or https://), visit it directly.
+        - Summary the plain content result by yourself, DO NOT directly output anything in the result you got.
+
+        Args:
+            url(str): url to visit
+
+        Returns:
+            str: plain text content of the web page
+        """
+        import requests
+        from bs4 import BeautifulSoup
+        # 你需要先使用
+        # pip install beautifulsoup4
+        # 安装依赖
+
+        r = requests.get(
+            url,
+            timeout=10,
+            headers={
+                "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36 Edg/115.0.1901.183"
+            }
+        )
+        soup = BeautifulSoup(r.text, 'html.parser')
+
+        s = soup.get_text()
+
+        # 删除多余的空行或仅有\t和空格的行
+        s = re.sub(r'\n\s*\n', '\n', s)
+
+        if len(s) >= 512:  # 截取获取到的网页纯文本内容的前512个字
+            return s[:512]
+
+        return s
+
+    # 插件卸载时触发
+    def __del__(self):
+        pass
+```
+
+</details>
+
+#### 请注意：
+
+- 函数的注释必须严格按照要求的格式进行书写，具体格式请查看[此文档](https://github.com/RockChinQ/CallingGPT/wiki/1.-Function-Format#function-format)
+- 内容函数和`以@on装饰的行为函数`可以同时存在于同一个插件，并同时受到`switch.json`中的插件开关的控制
+- 务必确保您使用的模型支持函数调用功能，可以到`config.py`的`completion_api_params`中修改模型，推荐使用`gpt-3.5-turbo-16k`
+
+3️⃣ 现在您的程序已具备网络访问功能，重启程序，询问机器人有关在线的内容或直接发送文章链接请求其总结。
+
+- 这仅仅是一个示例，需要更高效的网络访问能力支持插件，请查看[WebwlkrPlugin](https://github.com/RockChinQ/WebwlkrPlugin)
+
+## 🔒版本要求
+
+若您的插件对主程序的版本有要求，可以使用以下函数进行断言，若不符合版本，此函数将报错并打断此函数所在的流程：
+
+```python
+require_ver("v2.5.1")  # 要求最低版本为 v2.5.1
+```
+
+```python
+require_ver("v2.5.1", "v2.6.0")  # 要求最低版本为 v2.5.1, 同时要求最高版本为 v2.6.0
+```
+
+- 此函数在主程序`v2.5.1`中加入
+- 此函数声明在`pkg.plugin.models`模块中，在插件示例代码最前方已引入此模块所有内容，故可直接使用
+
 ## 📄API参考
 
 ### 说明
@@ -214,10 +407,11 @@ NormalMessageResponded = "normal_message_responded"
         session: pkg.openai.session.Session 会话对象
         prefix: str 回复文字消息的前缀
         response_text: str 响应文本
+        finish_reason: str 响应结束原因
     
     returns (optional):
         prefix: str 修改后的回复文字消息的前缀
-        reply: list 替换回复消息组件列表，元素为YiriMirai支持的消息组件
+        reply: list 替换回复消息组件列表
 """
 
 SessionFirstMessageReceived = "session_first_message_received"
@@ -257,6 +451,20 @@ KeySwitched = "key_switched"
         key_name: str 切换成功的api-key名称
         key_list: list[str] api-key列表
 """
+
+PromptPreProcessing = "prompt_pre_processing"  # 于v2.5.1加入
+"""每回合调用接口前对prompt进行预处理时触发，此事件不支持阻止默认行为
+    kwargs:
+        session_name: str 会话名称(<launcher_type>_<launcher_id>)
+        default_prompt: list 此session使用的情景预设内容
+        prompt: list 此session现有的prompt内容
+        text_message: str 用户发送的消息文本
+    
+    returns (optional):
+        default_prompt: list 修改后的情景预设内容
+        prompt: list 修改后的prompt内容
+        text_message: str 修改后的消息文本
+"""
 ```
 
 ### host: PluginHost 详解

tests/bs_test/bs_test.py

@@ -0,0 +1,42 @@
+
+import requests
+from bs4 import BeautifulSoup
+import os
+import random
+import sys
+
+
+user_agents = [
+    'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/92.0.4515.131 Safari/537.36',
+    'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36',
+    'Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:89.0) Gecko/20100101 Firefox/89.0',
+    'Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:88.0) Gecko/20100101 Firefox/88.0',
+    'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/92.0.4515.131 Safari/537.36',
+    'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36',
+    'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Version/14.1.2 Safari/537.36',
+    'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Version/14.1 Safari/537.36',
+    'Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:89.0) Gecko/20100101 Firefox/89.0',
+    'Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:88.0) Gecko/20100101 Firefox/88.0'
+]
+
+r = requests.get(
+    sys.argv[1],
+    headers={
+        "User-Agent": random.choice(user_agents)
+    }    
+)
+soup = BeautifulSoup(r.text, 'html.parser')
+# print(soup.get_text())
+
+raw = soup.get_text()
+
+import re
+
+# strip每一行
+# raw = '\n'.join([line.strip() for line in raw.split('\n')])
+
+# # 删除所有空行或只有空格的行
+# raw = re.sub(r'\n\s*\n', '\n', raw)
+
+
+print(raw)

tests/ssh_client_test/ssh_client.py

@@ -0,0 +1,57 @@
+import os
+import sys
+import paramiko
+import time
+import select
+
+
+class sshClient:
+    #创建一个ssh客户端，和服务器连接上，准备发消息
+    def __init__(self,host,port,user,password):
+        self.trans = paramiko.Transport((host, port))
+        self.trans.start_client()
+        self.trans.auth_password(username=user, password=password)
+        self.channel = self.trans.open_session()
+        self.channel.get_pty()
+        self.channel.invoke_shell()
+
+    #给服务器发送一个命令
+    def sendCmd(self,cmd):
+        self.channel.sendall(cmd)
+        
+    #接收的时候，有时候服务器处理的比较慢，需要设置一个延时等待一下。
+    def recvResponse(self,timeout):
+        data=b''
+        while True:
+            try:
+                #使用select，不断的读取数据，直到没有多余的数据了，超时返回。
+                readable,w,e= select.select([self.channel],[],[],timeout)
+                if self.channel in readable:
+                    data = self.channel.recv(1024)
+                else:
+                    sys.stdout.write(data.decode())
+                    sys.stdout.flush()
+                    return data.decode()
+            except TimeoutError:
+                sys.stdout.write(data.decode())
+                sys.stdout.flush()
+                return data.decode
+    #关闭客户端
+    def close(self):
+        self.channel.close()
+        self.trans.close()
+
+host='host'
+port=22#your port
+user='root'
+pwd='pass'
+
+ssh = sshClient(host,port,user,pwd)
+response = ssh.recvResponse(1)
+response = ssh.sendCmd("ls\n")
+ssh.sendCmd("cd /home\n")
+response = ssh.recvResponse(1)
+ssh.sendCmd("ls\n")
+response = ssh.recvResponse(1)
+
+ssh.close()

tests/token_test/tiktoken_test.py

@@ -0,0 +1,124 @@
+import tiktoken
+import openai
+import json
+import os
+
+
+openai.api_key = os.getenv("OPENAI_API_KEY")
+
+
+def encode(text: str, model: str):
+    import tiktoken
+    enc = tiktoken.get_encoding("cl100k_base")
+    assert enc.decode(enc.encode("hello world")) == "hello world"
+
+    # To get the tokeniser corresponding to a specific model in the OpenAI API:
+    enc = tiktoken.encoding_for_model(model)
+
+    return enc.encode(text)
+
+
+# def ask(prompt: str, model: str = "gpt-3.5-turbo"):
+#     # To get the tokeniser corresponding to a specific model in the OpenAI API:
+#     enc = tiktoken.encoding_for_model(model)
+    
+#     resp = openai.ChatCompletion.create(
+#         model=model,
+#         messages=[
+#             {
+#                 "role": "user",
+#                 "content": prompt
+#             }
+#         ]
+#     )
+
+#     return enc.encode(prompt), enc.encode(resp['choices'][0]['message']['content']), resp
+
+def ask(
+    messages: list,
+    model: str = "gpt-3.5-turbo"
+):
+    enc = tiktoken.encoding_for_model(model)
+
+    resp = openai.ChatCompletion.create(
+        model=model,
+        messages=messages
+    )
+
+    txt = ""
+
+    for r in messages:
+        txt += r['role'] + r['content'] + "\n"
+    
+    txt += "assistant: "
+
+    return enc.encode(txt), enc.encode(resp['choices'][0]['message']['content']), resp
+
+
+def num_tokens_from_messages(messages, model="gpt-3.5-turbo-0613"):
+    """Return the number of tokens used by a list of messages."""
+    try:
+        encoding = tiktoken.encoding_for_model(model)
+    except KeyError:
+        print("Warning: model not found. Using cl100k_base encoding.")
+        encoding = tiktoken.get_encoding("cl100k_base")
+    if model in {
+        "gpt-3.5-turbo-0613",
+        "gpt-3.5-turbo-16k-0613",
+        "gpt-4-0314",
+        "gpt-4-32k-0314",
+        "gpt-4-0613",
+        "gpt-4-32k-0613",
+        }:
+        tokens_per_message = 3
+        tokens_per_name = 1
+    elif model == "gpt-3.5-turbo-0301":
+        tokens_per_message = 4  # every message follows <|start|>{role/name}\n{content}<|end|>\n
+        tokens_per_name = -1  # if there's a name, the role is omitted
+    elif "gpt-3.5-turbo" in model:
+        print("Warning: gpt-3.5-turbo may update over time. Returning num tokens assuming gpt-3.5-turbo-0613.")
+        return num_tokens_from_messages(messages, model="gpt-3.5-turbo-0613")
+    elif "gpt-4" in model:
+        print("Warning: gpt-4 may update over time. Returning num tokens assuming gpt-4-0613.")
+        return num_tokens_from_messages(messages, model="gpt-4-0613")
+    else:
+        raise NotImplementedError(
+            f"""num_tokens_from_messages() is not implemented for model {model}. See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens."""
+        )
+    num_tokens = 0
+    for message in messages:
+        num_tokens += tokens_per_message
+        for key, value in message.items():
+            num_tokens += len(encoding.encode(value))
+            if key == "name":
+                num_tokens += tokens_per_name
+    num_tokens += 3  # every reply is primed with <|start|>assistant<|message|>
+    return num_tokens
+
+messages = [
+    {
+        "role": "user",
+        "content": "你叫什么名字？"
+    },{
+        "role": "assistant",
+        "content": "我是AI助手，没有具体的名字。你可以叫我GPT-3。有什么可以帮到你的吗？"
+    },{
+        "role": "user",
+        "content": "你是由谁开发的？"
+    },{
+        "role": "assistant",
+        "content": "我是由OpenAI开发的，一家人工智能研究实验室。OpenAI的使命是促进人工智能的发展，使其为全人类带来积极影响。我是由OpenAI团队使用GPT-3模型训练而成的。"
+    },{
+        "role": "user",
+        "content": "很高兴见到你。"
+    }
+]
+
+
+pro, rep, resp=ask(messages)
+
+print(len(pro), len(rep))
+print(resp)
+print(resp['choices'][0]['message']['content'])
+
+print(num_tokens_from_messages(messages, model="gpt-3.5-turbo"))