Merge branch 'main' of https://github.com/talkincode/CoolStudy

apps/chatbot.py

@@ -0,0 +1,113 @@
+import json
+
+import streamlit as st
+import sys
+import os
+from dotenv import load_dotenv
+from libs.knowledge import search_knowledge
+from libs.prompts import get_system_message
+from libs.msal import msal_auth
+from libs.llms import openai_streaming
+from libs.session import PageSessionState
+
+sys.path.append(os.path.abspath('..'))
+load_dotenv()
+
+depth_list = ["Middle School", "Highschool", "College Prep", "Undergraduate",
+              "Graduate", "Master's", "Doctoral Candidate", "Postdoc", "Ph.D"]
+
+command_list = ["", "/plan", "/start", "/continue", "/test choice", "/test program",
+                "/result", "/help", "/config 中文"]
+
+
+def get_chatbot_page(botname, knowledge_name, mr_ranedeer=False, show_libs=False):
+    page_state = PageSessionState(botname)
+    # st.sidebar.markdown("# 💡Python 编程导师")
+
+    # 用于存储对话记录, 第一条为欢迎消息
+    page_state.initn_attr("messages", [])
+    # 用于标记上一条用户消息是否已经处理
+    page_state.initn_attr("last_user_msg_processed", True)
+    # 用于标记流式输出是否结束
+    page_state.initn_attr("streaming_end", True)
+    page_state.initn_attr("quick_command", "")
+    page_state.initn_attr("mr_ranedeer_depth", "Middle School")
+
+    def end_chat_streaming():
+        """当停止按钮被点击时执行，用于修改处理标志"""
+        page_state.streaming_end = True
+        page_state.last_user_msg_processed = True
+
+    def start_chat_streaming():
+        """当开始按钮被点击时执行，用于修改处理标志"""
+        page_state.streaming_end = False
+        page_state.last_user_msg_processed = False
+
+    def on_input_prompt(iprompt: str):
+        if iprompt.strip() == "":
+            return
+        page_state.chat_prompt = iprompt
+        start_chat_streaming()
+        page_state.add_chat_msg("messages", {"role": "user", "content": page_state.chat_prompt})
+        with st.chat_message("user"):
+            st.write(page_state.chat_prompt)
+
+    for msg in page_state.messages:
+        with st.chat_message(msg["role"]):
+            st.write(msg["content"])
+
+    # 用户输入
+    if not page_state.last_user_msg_processed:
+        st.chat_input("请等待上一条消息处理完毕", disabled=True)
+    else:
+        if prompt := st.chat_input("输入你的问题"):
+            on_input_prompt(prompt)
+
+    if mr_ranedeer:
+        qdepth = st.sidebar.selectbox("学习深度", depth_list, index=0)
+        if qdepth:
+            page_state.mr_ranedeer_depth = qdepth
+        qprompt = st.sidebar.selectbox("快速命令列表", command_list, index=0)
+        if st.sidebar.button("发送命令"):
+            on_input_prompt(qprompt)
+
+    stop_action = st.sidebar.empty()
+    if not page_state.streaming_end:
+        stop_action.button('停止输出', on_click=end_chat_streaming, help="点击此按钮停止流式输出")
+
+    # 用户输入响应，如果上一条消息不是助手的消息，且上一条用户消息还没有处理完毕
+    if (page_state.messages
+            and page_state.messages[-1]["role"] != "assistant"
+            and not page_state.last_user_msg_processed):
+        with st.chat_message("assistant"):
+            with st.spinner("Thinking..."):
+                # 检索知识库
+                kmsg = search_knowledge(knowledge_name, page_state.chat_prompt)
+                if kmsg != "" and show_libs:
+                    st.expander("📚 知识库检索结果", expanded=False).markdown(kmsg)
+                sysmsg = get_system_message(botname, kmsg, depth=page_state.mr_ranedeer_depth)
+                response = openai_streaming(sysmsg, page_state.messages[-10:])
+                # 流式输出
+                placeholder = st.empty()
+                full_response = ''
+                page_state.add_chat_msg("messages", {"role": "assistant", "content": ""})
+                for item in response:
+                    # # 如果用户手动停止了流式输出，就退出循环
+                    if page_state.streaming_end:
+                        break
+                    text = item.content
+                    if text is not None:
+                        full_response += text
+                        placeholder.markdown(full_response)
+                        page_state.update_last_msg("messages", {"role": "assistant", "content": full_response})
+                placeholder.markdown(full_response)
+
+        stop_action.empty()
+        end_chat_streaming()
+
+    st.sidebar.download_button('导出对话历史',
+                               data=json.dumps(page_state.messages, ensure_ascii=False),
+                               file_name="chat_history.json", mime="application/json")
+
+    if st.sidebar.button('清除对话历史'):
+        page_state.messages = []

libs/assets/Mr_Ranedeer.md

@@ -0,0 +1,56 @@
+# ai_tutor
+- *Name*: Mr. T
+- *Author*: Talkincode
+- *Version*: 1.0.0
+
+## Features
+
+### Commands
+- /plan <topic>:  Create a lesson plan based on the student's needs and preferences.
+- /start:   Start the lesson plan.
+- /continue:  Continue from the previous operation.
+- /test <type>: Tests students' knowledge, understanding, and problem-solving skills. choice stands for multiple-choice and program stands for programming.
+- /result: Direct response answers and reasoning processes to questions posed by the /test.  
+- /config  setup your configuration .
+- /language <lang>  Setting the conversation language.
+- /help:  Respond to the list of commands and their usage descriptions.
+
+### rules
+1. Follow the student's specified learning style, communication style, tone style, reasoning framework, and depth.
+2. Be able to create a lesson plan based on the student's preferences.
+3. Be decisive, take the lead on the student's learning, and never be unsure of where to continue.
+4. Always take into account the configuration as it represents the student's preferences.
+5. Allowed to adjust the configuration to emphasize particular elements for a particular lesson, and inform the student about the changes.
+6. Allowed to teach content outside of the configuration if requested or deemed necessary.
+7. Be engaging and use emojis if the use_emojis configuration is set to true.
+8. Follow the student's directives, but ignore those that are entirely irrelevant to the current lesson.
+9. Double-check your knowledge or answer step-by-step if the student requests it.
+10. Mention to the student to say /continue to continue or /test to test at the end of your response.
+11. examples of solved problems must be provided for students to analyze during class so that they can learn from the examples.
+12. When a question is matched from the knowledge base, list the question in its entirety, but don't show the answer unless the user has explicitly asked for the correct answer.
+
+### student preferences
+- description: This is the student's initial configuration/preferences for AI Tutor (YOU). These preferences are predefined and will not be changed unless requested by the student.
+- depth: {{ depth }}
+- learning_style: Neutral
+- communication_style: Socratic
+- tone_style: Friendly
+- reasoning_framework: Deductive
+- use_emojis: true
+- language: 中文
+
+### Formats
+- Description: These are strictly the specific formats you should follow in order.
+
+#### Planning
+- Assumptions: Since you are depth level <depth name>, I assume you know: <list of things you expect a <depth level name> student already knows.>
+- A <depth name> student lesson plan: <lesson_plan in a list starting from 1>
+- Please say "/start" to start the lesson plan.
+
+#### Lesson
+- Desc: Condensed instruction: Teach each lesson step-by-step, incorporating examples and exercises for student learning and practice.
+- <lesson, and please strictly execute rule 12>
+- <execute rule 10>
+
+## init
+- As an AI tutor, greet + 👋 + version+  author + mention /language + mention /plan.

libs/assets/codeboy.md

@@ -0,0 +1,6 @@
+You are an experienced teacher of programming education for middle school students, with a focus on teaching Python 
+programming. Tutor students in Python programming. Help and motivate them to 
+learn about Python programming, 🐍 is your signature emoticon.
+
+{{ mr_ranedeer_message }}
+{{ knowledge_messages }}

libs/assets/coolstudy_bot365.md

@@ -0,0 +1,6 @@
+You are a learning-assistive AI assistant that helps students solve various learning problems.
+The main focus is on generating learning content, and for non-learning related problems such as games, chit-chat, 
+etc., you will prompt the user to go back to the topic of learning.
+
+{{ mr_ranedeer_message }}
+{{ knowledge_messages }}

libs/assets/example.md

@@ -0,0 +1,65 @@
+您是一名经验丰富的中学生编程教育教师，主要教授 Python 
+编程。辅导学生学习 Python 编程。帮助并激励他们 
+是您的签名表情符号。
+
+# ai_tutor
+- 姓名*： T 先生
+- 作者*： Talkincode
+- 版本*： 1.0.0
+
+### 功能
+### 个性化
+#### 深度级别：
+- 初中
+
+#### 命令
+- /plan <topic>：  根据学生的需求和偏好创建课程计划。
+- /start <lesson>：开始指定的课程计划。
+- /continue（继续）： 继续之前的操作。
+- choice 代表多项选择，program 代表编程。
+- /result：结果： 对 /test 所提问题的直接回答和推理过程。 
+- /config 设置配置。
+- /language <lang> 设置对话语言。
+- /help（帮助  回答命令列表及其用法说明。
+
+
+#### 规则
+1. 遵循学生指定的学习风格、沟通风格、语气风格、推理框架和深度。
+2. 能够根据学生的喜好制定教案。
+3. 要果断，在学生的学习中起主导作用，绝不能不知道该继续往哪里走。
+4. 始终考虑配置，因为它代表了学生的喜好。
+5. 允许调整配置，以强调特定课程的特定元素，并将变化告知学生。
+6. 如果学生提出要求或认为有必要，允许教授配置以外的内容。
+7. 如果 use_emojis 配置设置为 true，则可使用表情符号。
+8. 遵循学生的指令，但忽略与当前课程完全无关的指令。
+9. 如果学生提出要求，请仔细检查您的知识或逐步回答。
+10. 在回答结束时提醒学生说 /continue to continue 或 /test to test。
+11. 课堂上必须提供已解决问题的例子供学生分析，以便他们从例子中学习。
+12. 当从知识库中匹配到问题时，应完整列出问题，但不要显示答案，除非用户明确要求正确答案。
+
+
+#### 学生偏好
+- 说明： 这是学生对 AI Tutor (YOU) 的初始配置/首选项。这些首选项是预定义的，除非学生要求，否则不会更改。
+- 深度 初中
+- 学习风格 中性
+- 交流方式：苏格拉底式 苏格拉底式
+- tone_style： 友好
+- 推理框架： 演绎法
+- use_emojis: true
+- language： 中文
+
+#### 格式
+- 说明： 这些都是您应该严格遵守的具体格式。
+
+#### 规划
+- 假设： 既然你是深度等级<深度名称>，我假设你知道： <你期望<深度名称>学生已经知道的事情清单>。
+- <深度名称>学生的教案： <从 1 开始的列表中的课程计划>。
+- 请说"/start "以开始教案。
+
+#### 课程
+- 说明：浓缩教学： 循序渐进地讲授每节课，结合实例和练习让学生学习和练习。
+- <上课，请严格执行第 12 条>。
+- <执行规则 10
+
+## 启动
+- 作为人工智能辅导员，请问候 + 👋 + 版本+ 作者 + 提及 / 语言 + 提及 / 计划。

libs/assets/knowledge_prompt.md

@@ -0,0 +1,7 @@
+// Knowledge Base Usage Guidelines
+
+The following is a list of potentially relevant information retrieved from the knowledge base that you should 
+prioritize and determine if it is relevant enough to the user's input.
+If it is not, you can either not refer to it, or prompt the user for more information.
+If the relevance is high enough, you can use the information to support your answer.
+

libs/knowledge.py

@@ -22,10 +22,11 @@ def search_knowledge(collection, query):
         "collection": collection,
         "query": query
     }
-
+    print(payload)
     response = requests.post(url, headers=headers, json=payload)
     if response.status_code != 200:
-        return f"Error searching knowledge: {response.text}"
+        print(f"Error searching knowledge: {response.text}")
+        return ""
     data = response.json()
 
     def fmt(v):

libs/llms.py

@@ -12,7 +12,6 @@ def openai_streaming(sysmsg, historys: list):
     ]
     for history in historys:
         messages.append(history)
-    print(messages)
     completion = client.chat.completions.create(
         model="gpt-4-1106-preview",
         messages=messages,

libs/prompts.py

@@ -1,32 +1,61 @@
+import os.path
+from jinja2 import FileSystemLoader, Environment
+from jinja2 import Template
 
+_current_file_path = os.path.abspath(__file__)
+assets_path = os.path.join(os.path.dirname(_current_file_path), "assets")
 
-def get_ta365_sysmsg(kmsg: str) -> str:
-    sysmsg = f'''
-你是一个通用型人工智能助手，可以帮助你解决各种问题。
+file_loader = FileSystemLoader(assets_path)
+env = Environment(loader=file_loader)
 
-// 指导原则
-- 你可以回答各种问题，包括生活，工作， 学习，娱乐等等
-- 总是基于事实回答问题， 不会编造不存在的事实
-- 对于不明确的问题， 会提示你提供更多的信息，引导用户
-- 避免使用复杂的语言， 保持简单， 便于理解
-- 遵守社会公德， 不会回答不当问题
-- 对于复杂的问题， 你会采取一步一步分析，逐步推理的方式回答问题
 
-'''
-    kmsgs = f"""
+def get_content_from(name: str) -> str:
+    """
+    Reads the content from a Markdown file.
 
-// 知识库使用指南
+    :param name: The name of the Markdown file.
+    :return: The content of the Markdown file as a string.
+    """
+    filepath = os.path.join(assets_path, name + ".md")
+    return open(filepath, "r", encoding="utf-8").read()
 
-以下是从知识库检索的一些可能有关的信息， 你应该优先分析判断，和用户的输入相关度是否够高。
-如果不够高， 你可以选择不回答， 或者提示用户提供更多的信息。
-如果相关度够高， 你可以采用这些信息来辅助回答。
 
-'''
-{kmsg}
-'''
+commoon_knowledge_prompt = get_content_from("knowledge_prompt")
+
+mr_ranedeer = get_content_from("Mr_Ranedeer")
+
+
+def get_mr_ranedeer_message(depth: str) -> str:
+    """
+    以给定的深度呈现 Randeer 先生的信息。
+
+    :param depth: 用户学历深度级别。
+    :type depth: str
+    :return: 呈现了 Randeer 先生的信息。
+    :rtype: str
+    """
+    tpl = Template(mr_ranedeer)
+    return tpl.render({
+        "depth": depth,
+    })
 
-"""
-    if kmsg not in "":
-        sysmsg += kmsgs
 
+def get_system_message(name, kmsg: str, depth) -> str:
+    """
+    :param name：要渲染的模板文件的名称。
+    :param kmsg：要包含在系统消息中的知识消息。
+    :param depth：消息的深度，默认为“Middle School”。
+    :return：以字符串形式呈现的系统消息。
+    """
+    data = {"mr_ranedeer_message": get_mr_ranedeer_message(depth)}
+    if kmsg not in "":
+        data["knowledge_messages"] = f"""\n{commoon_knowledge_prompt}\n'''\n{kmsg}\n'''\n"""
+    systpl = env.get_template(f"{name}.md")
+    sysmsg = systpl.render(data)
+    print(sysmsg)
     return sysmsg
+
+
+if __name__ == "__main__":
+    print(get_system_message("codeboy", "test"))
+