Upload 7 files

README.md

@@ -1,12 +1,110 @@
----
-title: Ragdoing
-emoji: 💬
-colorFrom: yellow
-colorTo: purple
-sdk: gradio
-sdk_version: 5.0.1
-app_file: app.py
-pinned: false
----
-
-An example chatbot using [Gradio](https://gradio.app), [`huggingface_hub`](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/index), and the [Hugging Face Inference API](https://huggingface.co/docs/api-inference/index).
+# Easy-RAG
+一个适合学习、使用、自主扩展的RAG【检索增强生成】系统，可以联网做AI搜索！
+
+
+![img](https://github.com/yuntianhe2014/Easy-RAG/blob/main/img/zhu.png)
+
+更新历史
+
+        2024/9/04 增加 AI网络搜索 可以联网查询
+        2024/9/04 优化webui异步调用，提高响应速度
+        2024/8/21 增加对 Elasticsearch 支持，在config中设置
+        2024/7/23 参考 meet-libai 项目增加了一个知识图谱的时时提取工具，目前仅是提取，未存储 graph_demo_ui.py
+        2024/7/11 新增faiss向量数据库支持，目前支持(Chroma\FAISS)
+        2024/7/10 更新rerank搜索方式
+        2024/7/09 第一版发布
+![img](https://github.com/yuntianhe2014/Easy-RAG/blob/main/img/zhuye.png)
+
+1、目前已有的功能
+
+    知识库（目前仅支持txt\csv\pdf\md\doc\docx\mp3\mp4\wav\excel\格式数据）：
+
+        1、知识库的创建（目前仅支持Chroma\Faiss\Elasticsearch）
+        2、知识库的更新
+        3、删除知识库中某个文件
+        4、删除知识库
+        5、向量化知识库
+        6、支持音频视频的语音转文本然后向量化 
+            语音转文本 使用的 funasr ，第一次启动时，会从魔塔下载模型，可能会慢一些，之后会自动加载模型
+
+    chat
+
+        1、支持纯大模型聊天多轮
+        2、支持知识库问答 ["复杂召回方式", "简单召回方式","rerank"]
+        
+    AI网络搜索
+        
+        支持网络搜素，大家可以优化 prompt 增加不同 程度的 总结
+        llm基于ollama可以选择不同模型
+        注意：联网基于 searxng，需要先本地或者服务启动 这个项目，我用docker 启动的
+        参考 https://github.com/searxng/searxng-docker
+        
+![img](https://github.com/yuntianhe2014/Easy-RAG/blob/main/img/复杂方式.png)
+        3、通过使用rerank重新排序来提高检索效率
+        
+        本次rerank 使用了bge-reranker-large 模型，需要下载到本地，然后再 rag/rerank.py中配置路径
+            模型地址：https://hf-mirror.com/BAAI/bge-reranker-large
+     
+
+2、后续更新计划
+
+    知识库：
+
+        0、支持Elasticsearch、Milvus,MongoDB等向量数据
+
+
+    chat：
+
+        1、添加 语音回答输出
+        2、增加 问题路由知识库的 功能
+        
+
+安装使用
+
+       Ollma安装，在如下网址选择适合你机器的ollama 安装包，傻瓜式安装即可
+      
+        https://ollama.com/download
+      Ollama 安装模型，本次直接安装我们需要的两个 cmd中执行
+      
+        ollama run qwen2:7b
+        ollama run mofanke/acge_text_embedding:latest
+       
+      下载bge-reranker-large 模型然后在 rag/rerank.py中配置路径
+        
+        https://hf-mirror.com/BAAI/bge-reranker-large
+        
+      选择你想使用的向量数据库 目前仅支持（Chroma和Faiss）
+      
+        在 Config/config.py中配置你想用的 向量数据库
+        如果选择 Elasticsearch 请先启动 Elasticsearch，我是使用docker 启动的
+           docker run -p 9200:9200 -e "discovery.type=single-node" -e "xpack.security.enabled=false" -e "xpack.security.http.ssl.enabled=false" docker.elastic.co/elasticsearch/elasticsearch:8.12.1
+        注意修改 es_url
+        
+      构造python环境
+      
+        conda create -n Easy-RAG python=3.10.9
+        conda activate Easy-RAG
+        
+      项目开发使用的 python3.10.9  经测试 pyhon3.8以上皆可使用
+      
+        git clone https://github.com/yuntianhe2014/Easy-RAG.git
+      安装依赖
+      
+        pip3 install -r requirements.txt -i  https://mirrors.aliyun.com/pypi/simple
+       
+      部署依赖联网项目searxng
+        参考 https://github.com/searxng/searxng-docker
+      项目启动
+      
+        python webui.py
+        
+      知识图谱时时提取工具
+        python graph_demo_ui.py
+  ![img](https://github.com/yuntianhe2014/Easy-RAG/blob/main/img/graph-tool.png)
+
+更多介绍参考 公众号文章：世界大模型
+![img](https://github.com/yuntianhe2014/Easy-RAG/blob/main/img/%E5%BE%AE%E4%BF%A1%E5%9B%BE%E7%89%87_20240524180648.jpg)
+
+项目参考：
+    https://github.com/BinNong/meet-libai
+    https://github.com/searxng/searxng-docker

app.py

@@ -1,64 +1,354 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
-
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
-
-
-if __name__ == "__main__":
-    demo.launch()
+import gradio as gr
+import threading
+import asyncio
+import logging
+from concurrent.futures import ThreadPoolExecutor
+from functools import lru_cache
+import requests
+import json
+
+# 假设这些是您的自定义模块，需要根据实际情况进行调整
+from Config.config import VECTOR_DB, DB_directory
+from Ollama_api.ollama_api import *
+from rag.rag_class import *
+
+# 设置日志
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+# 根据VECTOR_DB选择合适的向量数据库
+if VECTOR_DB == 1:
+    from embeding.chromadb import ChromaDB as vectorDB
+    vectordb = vectorDB(persist_directory=DB_directory)
+elif VECTOR_DB == 2:
+    from embeding.faissdb import FaissDB as vectorDB
+    vectordb = vectorDB(persist_directory=DB_directory)
+elif VECTOR_DB == 3:
+    from embeding.elasticsearchStore import ElsStore as vectorDB
+    vectordb = vectorDB()
+
+# 存储上传的文件
+uploaded_files = []
+
+@lru_cache(maxsize=100)
+def get_knowledge_base_files():
+    cl_dict = {}
+    cols = vectordb.get_all_collections_name()
+    for c_name in cols:
+        cl_dict[c_name] = vectordb.get_collcetion_content_files(c_name)
+    return cl_dict
+
+knowledge_base_files = get_knowledge_base_files()
+
+def upload_files(files):
+    if files:
+        new_files = [file.name for file in files]
+        uploaded_files.extend(new_files)
+        update_knowledge_base_files()
+        logger.info(f"Uploaded files: {new_files}")
+        return update_file_list(), new_files, "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Upload successful!</div>"
+    update_knowledge_base_files()
+    return update_file_list(), [], "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Upload failed!</div>"
+
+def delete_files(selected_files):
+    global uploaded_files
+    uploaded_files = [f for f in uploaded_files if f not in selected_files]
+    if selected_files:
+        update_knowledge_base_files()
+        logger.info(f"Deleted files: {selected_files}")
+        return update_file_list(), "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Delete successful!</div>"
+    update_knowledge_base_files()
+    return update_file_list(), "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Delete failed!</div>"
+
+def delete_collection(selected_knowledge_base):
+    if selected_knowledge_base and selected_knowledge_base != "创建知识库":
+        vectordb.delete_collection(selected_knowledge_base)
+        update_knowledge_base_files()
+        logger.info(f"Deleted collection: {selected_knowledge_base}")
+        return update_knowledge_base_dropdown(), "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Collection deleted successfully!</div>"
+    return update_knowledge_base_dropdown(), "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Delete collection failed!</div>"
+
+async def async_vectorize_files(selected_files, selected_knowledge_base, new_kb_name, chunk_size, chunk_overlap):
+    if selected_files:
+        if selected_knowledge_base == "创建知识库":
+            knowledge_base = new_kb_name
+            vectordb.create_collection(selected_files, knowledge_base, chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+        else:
+            knowledge_base = selected_knowledge_base
+            vectordb.add_chroma(selected_files, knowledge_base, chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+
+        if knowledge_base not in knowledge_base_files:
+            knowledge_base_files[knowledge_base] = []
+        knowledge_base_files[knowledge_base].extend(selected_files)
+
+        logger.info(f"Vectorized files: {selected_files} for knowledge base: {knowledge_base}")
+        await asyncio.sleep(0)  # 允许其他任务执行
+        return f"Vectorized files: {', '.join(selected_files)}\nKnowledge Base: {knowledge_base}\nUploaded Files: {', '.join(uploaded_files)}", "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Vectorization successful!</div>"
+    return "", "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Vectorization failed!</div>"
+
+def update_file_list():
+    return gr.update(choices=uploaded_files, value=[])
+
+def search_knowledge_base(selected_knowledge_base):
+    if selected_knowledge_base in knowledge_base_files:
+        kb_files = knowledge_base_files[selected_knowledge_base]
+        return gr.update(choices=kb_files, value=[])
+    return gr.update(choices=[], value=[])
+
+def update_knowledge_base_files():
+    global knowledge_base_files
+    knowledge_base_files = get_knowledge_base_files()
+
+# 处理聊天消息的函数
+chat_history = []
+
+def safe_chat_response(model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, message):
+    try:
+        return chat_response(model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, message)
+    except Exception as e:
+        logger.error(f"Error in chat response: {str(e)}")
+        return f"<div style='color: red;'>Error: {str(e)}</div>", ""
+
+def chat_response(model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, message):
+    global chat_history
+    if message:
+        chat_history.append(("User", message))
+        if chat_knowledge_base_dropdown == "仅使用模型":
+            rag = RAG_class(model=model_dropdown,persist_directory=DB_directory)
+            answer = rag.mult_chat(chat_history)
+        if chat_knowledge_base_dropdown and chat_knowledge_base_dropdown != "仅使用模型":
+            rag = RAG_class(model=model_dropdown, embed=vector_dropdown, c_name=chat_knowledge_base_dropdown, persist_directory=DB_directory)
+            if chain_dropdown == "复杂召回方式":
+                questions = rag.decomposition_chain(message)
+                answer = rag.rag_chain(questions)
+            elif chain_dropdown == "简单召回方式":
+                answer = rag.simple_chain(message)
+            else:
+                answer = rag.rerank_chain(message)
+
+        response = f" {answer}"
+        chat_history.append(("Bot", response))
+    return format_chat_history(chat_history), ""
+
+def clear_chat():
+    global chat_history
+    chat_history = []
+    return format_chat_history(chat_history)
+
+def format_chat_history(history):
+    formatted_history = ""
+    for user, msg in history:
+        if user == "User":
+            formatted_history += f'''
+            <div style="text-align: right; margin: 10px;">
+                <div style="display: inline-block; background-color: #DCF8C6; padding: 10px; border-radius: 10px; max-width: 60%;">
+                    {msg}
+                </div>
+                <b>:User</b>
+            </div>
+            '''
+        else:
+            if "```" in msg:  # 检测是否包含代码片段
+                code_content = msg.split("```")[1]
+                formatted_history += f'''
+                <div style="text-align: left; margin: 10px;">
+                    <b>Bot:</b>
+                    <div style="display: inline-block; background-color: #F1F0F0; padding: 10px; border-radius: 10px; max-width: 60%;">
+                        <pre><code>{code_content}</code></pre>
+                    </div>
+                </div>
+                '''
+            else:
+                formatted_history += f'''
+                <div style="text-align: left; margin: 10px;">
+                    <b>Bot:</b>
+                    <div style="display: inline-block; background-color: #F1F0F0; padding: 10px; border-radius: 10px; max-width: 60%;">
+                        {msg}
+                    </div>
+                </div>
+                '''
+    return formatted_history
+
+def clear_status():
+    upload_status.update("")
+    delete_status.update("")
+    vectorize_status.update("")
+    delete_collection_status.update("")
+
+def handle_knowledge_base_selection(selected_knowledge_base):
+    if selected_knowledge_base == "创建知识库":
+        return gr.update(visible=True, interactive=True), gr.update(choices=[], value=[]), gr.update(visible=False)
+    elif selected_knowledge_base == "仅使用模型":
+        return gr.update(visible=False, interactive=False), gr.update(choices=[], value=[]), gr.update(visible=False)
+    else:
+        return gr.update(visible=False, interactive=False), search_knowledge_base(selected_knowledge_base), gr.update(visible=True)
+
+def update_knowledge_base_dropdown():
+    global knowledge_base_files
+    choices = ["创建知识库"] + list(knowledge_base_files.keys())
+    return gr.update(choices=choices)
+
+def update_chat_knowledge_base_dropdown():
+    global knowledge_base_files
+    choices = ["仅使用模型"] + list(knowledge_base_files.keys())
+    return gr.update(choices=choices)
+
+
+# SearxNG搜索函数
+def search_searxng(query):
+    searxng_url = 'http://localhost:8080/search'  # 替换为你的SearxNG实例URL
+    params = {
+        'q': query,
+        'format': 'json'
+    }
+    response = requests.get(searxng_url, params=params)
+    response.raise_for_status()
+    return response.json()
+
+
+# Ollama总结函数
+def summarize_with_ollama(model_dropdown,text, question):
+    prompt = """
+        根据下边的内容，回答用户问题，
+        内容为：‘{0}‘\n
+        问题为：{1}
+    """.format(text, question)
+    ollama_url = 'http://localhost:11434/api/generate'  # 替换为你的Ollama实例URL
+    data = {
+        'model': model_dropdown,
+        "prompt": prompt,
+        "stream": False
+    }
+    response = requests.post(ollama_url, json=data)
+    response.raise_for_status()
+    return response.json()
+
+
+# 处理函数
+def ai_web_search(model_dropdown,user_query):
+    # 使用SearxNG进行搜索
+    search_results = search_searxng(user_query)
+    search_texts = [result['title'] + "\n" + result['content'] for result in search_results['results']]
+    combined_text = "\n\n".join(search_texts)
+
+    # 使用Ollama进行总结
+    summary = summarize_with_ollama(model_dropdown,combined_text, user_query)
+    # print(summary)
+    # 返回结果
+    return summary['response']
+# 添加新的函数来处理AI网络搜索
+# def ai_web_search(model_dropdown, query):
+#     try:
+#         # 这里添加实际的网络搜索和AI处理逻辑
+#         # 这只是一个示例，您需要根据实际情况实现
+#         search_result = f"搜索结果: {query}"
+#         ai_response = f"AI回答: 基于搜索结果，对于'{query}'的回答是..."
+#         return f"{search_result}\n\n{ai_response}"
+#     except Exception as e:
+#         logger.error(f"Error in AI web search: {str(e)}")
+#         return f"<div style='color: red;'>Error: {str(e)}</div>"
+
+# 创建 Gradio 界面
+with gr.Blocks() as demo:
+    with gr.Column():
+        # 添加标题
+        title = gr.HTML("<h1 style='text-align: center; font-size: 32px; font-weight: bold;'>RAG精致系统</h1>")
+        # 添加公告栏
+        announcement = gr.HTML("<div style='text-align: center; font-size: 18px; color: red;'>公告栏: RAG精致系统，【检索增强生成】系统！<br/>莫大大</div>")
+
+        with gr.Tabs():
+            with gr.TabItem("知识库"):
+                knowledge_base_dropdown = gr.Dropdown(choices=["创建知识库"] + list(knowledge_base_files.keys()),
+                                                      label="选择知识库")
+                new_kb_input = gr.Textbox(label="输入新的知识库名称", visible=False, interactive=True)
+                file_input = gr.Files(label="Upload files")
+                upload_btn = gr.Button("Upload")
+                file_list = gr.CheckboxGroup(label="Uploaded Files")
+                delete_btn = gr.Button("Delete Selected Files")
+                with gr.Row():
+                    chunk_size_dropdown = gr.Dropdown(choices=[50, 100, 200, 300, 500, 700], label="chunk_size", value=200)
+                    chunk_overlap_dropdown = gr.Dropdown(choices=[20, 50, 100, 200], label="chunk_overlap", value=50)
+                    vectorize_btn = gr.Button("Vectorize Selected Files")
+                delete_collection_btn = gr.Button("Delete Collection")
+                upload_status = gr.HTML()
+                delete_status = gr.HTML()
+                vectorize_status = gr.HTML()
+                delete_collection_status = gr.HTML()
+
+            with gr.TabItem("Chat"):
+                with gr.Row():
+                    model_dropdown = gr.Dropdown(choices=get_llm(), label="模型")
+                    vector_dropdown = gr.Dropdown(choices=get_embeding_model(), label="向量")
+                    chat_knowledge_base_dropdown = gr.Dropdown(choices=["仅使用模型"] + vectordb.get_all_collections_name(), label="知识库")
+                    chain_dropdown = gr.Dropdown(choices=["复杂召回方式", "简单召回方式","rerank"], label="chain方式", visible=False)
+                chat_display = gr.HTML(label="Chat History")
+                chat_input = gr.Textbox(label="Type a message")
+                chat_btn = gr.Button("Send")
+                clear_btn = gr.Button("Clear Chat History")
+
+            with gr.TabItem("AI网络搜索"):
+                with gr.Row():
+                    web_search_model_dropdown = gr.Dropdown(choices=get_llm(), label="模型")
+                web_search_output = gr.Textbox(label="搜索结果和AI回答", lines=10)
+                web_search_input = gr.Textbox(label="输入搜索查询")
+
+                web_search_btn = gr.Button("搜索")
+
+    def handle_upload(files):
+        upload_result, new_files, status = upload_files(files)
+        threading.Thread(target=clear_status).start()
+        return upload_result, new_files, status, update_chat_knowledge_base_dropdown()
+
+    def handle_delete(selected_knowledge_base, selected_files):
+        tmp = []
+        cols_files_tmp = vectordb.get_collcetion_content_files(c_name=selected_knowledge_base)
+        for i in selected_files:
+            if i in cols_files_tmp:
+                tmp.append(i)
+        del cols_files_tmp
+        if tmp:
+            vectordb.del_files(tmp, c_name=selected_knowledge_base)
+        del tmp
+        delete_result, status = delete_files(selected_files)
+        threading.Thread(target=clear_status).start()
+        return delete_result, status, update_chat_knowledge_base_dropdown()
+
+    def handle_vectorize(selected_files, selected_knowledge_base, new_kb_name, chunk_size, chunk_overlap):
+        vectorize_result, status = asyncio.run(async_vectorize_files(selected_files, selected_knowledge_base, new_kb_name, chunk_size, chunk_overlap))
+        threading.Thread(target=clear_status).start()
+        return vectorize_result, status, update_knowledge_base_dropdown(), update_chat_knowledge_base_dropdown()
+
+    def handle_delete_collection(selected_knowledge_base):
+        result, status = delete_collection(selected_knowledge_base)
+        threading.Thread(target=clear_status).start()
+        return result, status, update_chat_knowledge_base_dropdown()
+
+    knowledge_base_dropdown.change(
+        handle_knowledge_base_selection,
+        inputs=knowledge_base_dropdown,
+        outputs=[new_kb_input, file_list, chain_dropdown]
+    )
+    upload_btn.click(handle_upload, inputs=file_input, outputs=[file_list, file_list, upload_status, chat_knowledge_base_dropdown])
+    delete_btn.click(handle_delete, inputs=[knowledge_base_dropdown, file_list], outputs=[file_list, delete_status, chat_knowledge_base_dropdown])
+    vectorize_btn.click(handle_vectorize, inputs=[file_list, knowledge_base_dropdown, new_kb_input, chunk_size_dropdown, chunk_overlap_dropdown],
+                        outputs=[gr.Textbox(visible=False), vectorize_status, knowledge_base_dropdown, chat_knowledge_base_dropdown])
+    delete_collection_btn.click(handle_delete_collection, inputs=knowledge_base_dropdown,
+                                outputs=[knowledge_base_dropdown, delete_collection_status, chat_knowledge_base_dropdown])
+
+    chat_btn.click(chat_response, inputs=[model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, chat_input], outputs=[chat_display, chat_input])
+    clear_btn.click(clear_chat, outputs=chat_display)
+
+    chat_knowledge_base_dropdown.change(
+        fn=lambda selected: gr.update(visible=selected != "仅使用模型"),
+        inputs=chat_knowledge_base_dropdown,
+        outputs=chain_dropdown
+    )
+
+    # 添加新的点击事件处理
+    web_search_btn.click(
+        ai_web_search,
+        inputs=[web_search_model_dropdown, web_search_input],
+        outputs=web_search_output
+    )
+
+demo.launch(debug=True,share=True)

graph_demo_ui.py

@@ -0,0 +1,87 @@
+# -*- coding: utf-8 -*-
+from flask import Flask, render_template, request, jsonify
+import json
+from dotenv import load_dotenv
+from langchain_community.llms import Ollama
+
+
+load_dotenv()
+
+app = Flask(__name__)
+
+# 测试了 llama3：8b,gemma2:9b,qwen2:7b,glm4:9b，arcee-ai/arcee-agent:latest  目前来看 qwen2:7 效果最好
+llm = Ollama(model="qwen2:7b")
+
+
+json_example = {'edges': [{'data': {'color': '#FFA07A',
+                                    'label': 'label 1',
+                                    'source': 'source 1',
+                                    'target': 'target 1'}},
+                          {'data': {'color': '#FFA07A',
+                                    'label': 'label 2',
+                                    'source': 'source 2',
+                                    'target': 'target 2'}}
+                          ],
+                'nodes': [{'data': {'color': '#FFC0CB', 'id': 'id 1', 'label': 'label 1'}},
+                          {'data': {'color': '#90EE90', 'id': 'id 2', 'label': 'label 2'}},
+                          {'data': {'color': '#87CEEB', 'id': 'id 3', 'label': 'label 3'}}]}
+
+
+
+__retriever_prompt = f"""
+            您是一名专门从事知识图谱创建的人工智能专家，目标是根据给定的输入或请求捕获关系。
+            基于各种形式的用户输入，如段落、电子邮件、文本文件等。
+            你的任务是根据输入创建一个知识图谱。
+            nodes必须具有label参数，并且label是来自输入的词语或短语，nodes必须具有id参数,id的格式是"id_数字"，不能重复。
+            edges还必须有一个label参数，其中label是输入中的直接词语或短语,edges中的source和target取自nodes中的id。
+            仅使用JSON进行响应，其格式可以在python中进行jsonify，并直接输入cy.add（data），包括“color”属性，以在前端显示图形。
+            您可以参考给定的示例：{json_example}。存储node和edge的数组中，最后一个元素后边不要有逗号，
+            确保边的目标和源与现有节点匹配。
+            不要在JSON的上方和下方包含markdown三引号，直接用花括号括起来。
+            """
+
+
+def generate_graph_info(raw_text: str) -> str | None:
+    """
+    generate graph info from raw text
+    :param raw_text:
+    :return:
+    """
+    messages = [
+        {"role": "system", "content": "你现在扮演信息抽取的角色，要求根据用户输入和AI的回答，正确提取出信息,记得不多对实体进行翻译。"},
+        {"role": "user", "content": raw_text},
+        {"role": "user", "content": __retriever_prompt}
+    ]
+    print("解析中....")
+    for i in range(3):
+        graph_info_result = llm.invoke(messages)
+        if len(graph_info_result)<10:
+            print("-------",i,"-------------------")
+            continue
+        else:
+            break
+    print(graph_info_result)
+    return graph_info_result
+
+
+@app.route('/')
+def index():
+    return render_template('index.html')
+
+
+@app.route('/update_graph', methods=['POST'])
+def update_graph():
+    raw_text = request.json.get('text', '')
+    try:
+        result = generate_graph_info(raw_text)
+        if '```' in result:
+            graph_data=json.loads(result.split('```',2)[1].replace("json", ''))
+        else:
+            graph_data=json.loads(result)
+        return graph_data
+    except Exception as e:
+        return {'error': f"Error parsing graph data: {str(e)}"}
+
+
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=7860)

requirements.txt

@@ -1 +1,10 @@
-huggingface_hub==0.25.2
+gradio==4.29.0
+langchain-community==0.2.6
+langchain==0.2.6
+langchain-core==0.2.11
+requests
+transformers==4.41.1
+unstructured==0.7.12
+funasr==1.0.24
+modelscope
+chromadb

webui-test-graph.py

@@ -0,0 +1,283 @@
+import gradio as gr
+import threading
+from Config.config import VECTOR_DB,DB_directory
+
+if VECTOR_DB==1:
+    from embeding.chromadb import ChromaDB as vectorDB
+    vectordb = vectorDB(persist_directory=DB_directory)
+elif VECTOR_DB==2:
+    from embeding.faissdb import FaissDB as vectorDB
+    vectordb = vectorDB(persist_directory=DB_directory)
+from Ollama_api.ollama_api import *
+from rag.rag_class import *
+
+# 存储上传的文件
+uploaded_files = []
+
+# 模拟获取最新的知识库文件
+def get_knowledge_base_files():
+    cl_dict = {}
+    cols = vectordb.get_all_collections_name()
+    for c_name in cols:
+        cl_dict[c_name] = vectordb.get_collcetion_content_files(c_name)
+    return cl_dict
+
+knowledge_base_files = get_knowledge_base_files()
+
+def upload_files(files):
+    if files:
+        new_files = [file.name for file in files]
+        uploaded_files.extend(new_files)
+        update_knowledge_base_files()
+        return update_file_list(), new_files, "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Upload successful!</div>"
+    update_knowledge_base_files()
+    return update_file_list(), [], "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Upload failed!</div>"
+
+def delete_files(selected_files):
+    global uploaded_files
+    uploaded_files = [f for f in uploaded_files if f not in selected_files]
+    if selected_files:
+        update_knowledge_base_files()
+        return update_file_list(), "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Delete successful!</div>"
+    update_knowledge_base_files()
+    return update_file_list(), "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Delete failed!</div>"
+
+def delete_collection(selected_knowledge_base):
+    if selected_knowledge_base and selected_knowledge_base != "创建知识库":
+        vectordb.delete_collection(selected_knowledge_base)
+        update_knowledge_base_files()
+        return update_knowledge_base_dropdown(), "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Collection deleted successfully!</div>"
+    return update_knowledge_base_dropdown(), "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Delete collection failed!</div>"
+
+def create_graph(selected_files):
+    from Neo4j.neo4j_op import KnowledgeGraph
+    from Neo4j.graph_extract import update_graph
+    from Config.config import neo4j_host, neo4j_name, neo4j_pwd
+    import tqdm
+
+    kg = KnowledgeGraph(neo4j_host,neo4j_name,neo4j_pwd)
+    data = kg.split_files(selected_files)
+    for doc in tqdm.tqdm(data):
+        text = doc.page_content
+        try:
+            res = update_graph(text)
+            # 批量创建节点
+            nodes = kg.create_nodes("node", res["nodes"])
+
+            # 批量创建关系
+            relationships = kg.create_relationships([
+                ("node", {"name": edge["source"]}, "node", {"name": edge["target"]}, edge["label"]) for edge in res["edges"]
+            ])
+        except:
+            print("错误----------------------------------")
+
+
+def vectorize_files(selected_files, selected_knowledge_base, new_kb_name,choice_graph, chunk_size, chunk_overlap):
+    if selected_files:
+        if selected_knowledge_base == "创建知识库":
+            knowledge_base = new_kb_name
+            vectordb.create_collection(selected_files, knowledge_base, chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+            if choice_graph=='是':
+                create_graph(selected_files)
+        else:
+            knowledge_base = selected_knowledge_base
+            vectordb.add_chroma(selected_files, knowledge_base, chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+            if choice_graph == '是':
+                create_graph(selected_files)
+        if knowledge_base not in knowledge_base_files:
+            knowledge_base_files[knowledge_base] = []
+        knowledge_base_files[knowledge_base].extend(selected_files)
+
+        return f"Vectorized files: {', '.join(selected_files)}\nKnowledge Base: {knowledge_base}\nUploaded Files: {', '.join(uploaded_files)}", "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Vectorization successful!</div>"
+    return "", "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Vectorization failed!</div>"
+
+def update_file_list():
+    return gr.update(choices=uploaded_files, value=[])
+
+def search_knowledge_base(selected_knowledge_base):
+    if selected_knowledge_base in knowledge_base_files:
+        kb_files = knowledge_base_files[selected_knowledge_base]
+        return gr.update(choices=kb_files, value=[])
+    return gr.update(choices=[], value=[])
+
+def update_knowledge_base_files():
+    global knowledge_base_files
+    knowledge_base_files = get_knowledge_base_files()
+
+# 处理聊天消息的函数
+chat_history = []
+
+def chat_response(model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, message):
+    global chat_history
+    if message:
+        chat_history.append(("User", message))
+        if chat_knowledge_base_dropdown == "仅使用模型":
+            rag = RAG_class(model=model_dropdown,persist_directory=DB_directory)
+            answer = rag.mult_chat(chat_history)
+        if chat_knowledge_base_dropdown and chat_knowledge_base_dropdown != "仅使用模型":
+            rag = RAG_class(model=model_dropdown, embed=vector_dropdown, c_name=chat_knowledge_base_dropdown, persist_directory=DB_directory)
+            if chain_dropdown == "复杂召回方式":
+                questions = rag.decomposition_chain(message)
+                answer = rag.rag_chain(questions)
+            elif chain_dropdown == "简单召回方式":
+                answer = rag.simple_chain(message)
+            else:
+                answer = rag.rerank_chain(message)
+
+        response = f" {answer}"
+        chat_history.append(("Bot", response))
+    return format_chat_history(chat_history), ""
+
+def clear_chat():
+    global chat_history
+    chat_history = []
+    return format_chat_history(chat_history)
+
+def format_chat_history(history):
+    formatted_history = ""
+    for user, msg in history:
+        if user == "User":
+            formatted_history += f'''
+            <div style="text-align: right; margin: 10px;">
+                <div style="display: inline-block; background-color: #DCF8C6; padding: 10px; border-radius: 10px; max-width: 60%;">
+                    {msg}
+                </div>
+                <b>:User</b>
+            </div>
+            '''
+        else:
+            if "```" in msg:  # 检测是否包含代码片段
+                code_content = msg.split("```")[1]
+                formatted_history += f'''
+                <div style="text-align: left; margin: 10px;">
+                    <b>Bot:</b>
+                    <div style="display: inline-block; background-color: #F1F0F0; padding: 10px; border-radius: 10px; max-width: 60%;">
+                        <pre><code>{code_content}</code></pre>
+                    </div>
+                </div>
+                '''
+            else:
+                formatted_history += f'''
+                <div style="text-align: left; margin: 10px;">
+                    <b>Bot:</b>
+                    <div style="display: inline-block; background-color: #F1F0F0; padding: 10px; border-radius: 10px; max-width: 60%;">
+                        {msg}
+                    </div>
+                </div>
+                '''
+    return formatted_history
+
+def clear_status():
+    upload_status.update("")
+    delete_status.update("")
+    vectorize_status.update("")
+    delete_collection_status.update("")
+
+def handle_knowledge_base_selection(selected_knowledge_base):
+    if selected_knowledge_base == "创建知识库":
+        return gr.update(visible=True, interactive=True), gr.update(choices=[], value=[]), gr.update(visible=False)
+    elif selected_knowledge_base == "仅使用模型":
+        return gr.update(visible=False, interactive=False), gr.update(choices=[], value=[]), gr.update(visible=False)
+    else:
+        return gr.update(visible=False, interactive=False), search_knowledge_base(selected_knowledge_base), gr.update(visible=True)
+
+def update_knowledge_base_dropdown():
+    global knowledge_base_files
+    choices = ["创建知识库"] + list(knowledge_base_files.keys())
+    return gr.update(choices=choices)
+
+def update_chat_knowledge_base_dropdown():
+    global knowledge_base_files
+    choices = ["仅使用模型"] + list(knowledge_base_files.keys())
+    return gr.update(choices=choices)
+
+# 创建 Gradio 界面
+with gr.Blocks() as demo:
+    with gr.Column():
+        # 添加标题
+        title = gr.HTML("<h1 style='text-align: center; font-size: 32px; font-weight: bold;'>RAG精致系统</h1>")
+        # 添加公告栏
+        announcement = gr.HTML("<div style='text-align: center; font-size: 18px; color: red;'>公告栏: 欢迎使用RAG精致系统</div>")
+
+        with gr.Tabs():
+            with gr.TabItem("知识库"):
+                knowledge_base_dropdown = gr.Dropdown(choices=["创建知识库"] + list(knowledge_base_files.keys()),
+                                                      label="选择知识库")
+                new_kb_input = gr.Textbox(label="输入新的知识库名称", visible=False, interactive=True)
+                choice_graph = gr.Radio(choices=["否", "是"], value="否",label="是否同时提取知识图谱(会比较慢)")
+                file_input = gr.Files(label="Upload files")
+                upload_btn = gr.Button("Upload")
+                file_list = gr.CheckboxGroup(label="Uploaded Files")
+                delete_btn = gr.Button("Delete Selected Files")
+                with gr.Row():
+                    chunk_size_dropdown = gr.Dropdown(choices=[50, 100, 200, 300, 500, 700], label="chunk_size", value=200)
+                    chunk_overlap_dropdown = gr.Dropdown(choices=[20, 50, 100, 200], label="chunk_overlap", value=50)
+                    vectorize_btn = gr.Button("Vectorize Selected Files")
+                delete_collection_btn = gr.Button("Delete Collection")
+                upload_status = gr.HTML()
+                delete_status = gr.HTML()
+                vectorize_status = gr.HTML()
+                delete_collection_status = gr.HTML()
+
+            with gr.TabItem("Chat"):
+                with gr.Row():
+                    model_dropdown = gr.Dropdown(choices=get_llm(), label="模型")
+                    vector_dropdown = gr.Dropdown(choices=get_embeding_model(), label="向量")
+                    chat_knowledge_base_dropdown = gr.Dropdown(choices=["仅使用模型"] + vectordb.get_all_collections_name(), label="知识库")
+                    chain_dropdown = gr.Dropdown(choices=["复杂召回方式", "简单召回方式","rerank"], label="chain方式", visible=False)
+                chat_display = gr.HTML(label="Chat History")
+                chat_input = gr.Textbox(label="Type a message")
+                chat_btn = gr.Button("Send")
+                clear_btn = gr.Button("Clear Chat History")
+
+    def handle_upload(files):
+        upload_result, new_files, status = upload_files(files)
+        threading.Thread(target=clear_status).start()
+        return upload_result, new_files, status, update_chat_knowledge_base_dropdown()
+
+    def handle_delete(selected_knowledge_base, selected_files):
+        tmp = []
+        cols_files_tmp = vectordb.get_collcetion_content_files(c_name=selected_knowledge_base)
+        for i in selected_files:
+            if i in cols_files_tmp:
+                tmp.append(i)
+        del cols_files_tmp
+        if tmp:
+            vectordb.del_files(tmp, c_name=selected_knowledge_base)
+        del tmp
+        delete_result, status = delete_files(selected_files)
+        threading.Thread(target=clear_status).start()
+        return delete_result, status, update_chat_knowledge_base_dropdown()
+
+    def handle_vectorize(selected_files, selected_knowledge_base, new_kb_name, choice_graph,chunk_size, chunk_overlap):
+        vectorize_result, status = vectorize_files(selected_files, selected_knowledge_base, new_kb_name, choice_graph,chunk_size, chunk_overlap)
+        threading.Thread(target=clear_status).start()
+        return vectorize_result, status, update_knowledge_base_dropdown(), update_chat_knowledge_base_dropdown()
+
+    def handle_delete_collection(selected_knowledge_base):
+        result, status = delete_collection(selected_knowledge_base)
+        threading.Thread(target=clear_status).start()
+        return result, status, update_chat_knowledge_base_dropdown()
+
+    knowledge_base_dropdown.change(
+        handle_knowledge_base_selection,
+        inputs=knowledge_base_dropdown,
+        outputs=[new_kb_input, file_list, chain_dropdown]
+    )
+    upload_btn.click(handle_upload, inputs=file_input, outputs=[file_list, file_list, upload_status, chat_knowledge_base_dropdown])
+    delete_btn.click(handle_delete, inputs=[knowledge_base_dropdown, file_list], outputs=[file_list, delete_status, chat_knowledge_base_dropdown])
+    vectorize_btn.click(handle_vectorize, inputs=[file_list, knowledge_base_dropdown, new_kb_input,choice_graph, chunk_size_dropdown, chunk_overlap_dropdown],
+                        outputs=[gr.Textbox(visible=False), vectorize_status, knowledge_base_dropdown, chat_knowledge_base_dropdown])
+    delete_collection_btn.click(handle_delete_collection, inputs=knowledge_base_dropdown,
+                                outputs=[knowledge_base_dropdown, delete_collection_status, chat_knowledge_base_dropdown])
+
+    chat_btn.click(chat_response, inputs=[model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, chat_input], outputs=[chat_display, chat_input])
+    clear_btn.click(clear_chat, outputs=chat_display)
+
+    chat_knowledge_base_dropdown.change(
+        fn=lambda selected: gr.update(visible=selected != "仅使用模型"),
+        inputs=chat_knowledge_base_dropdown,
+        outputs=chain_dropdown
+    )
+
+demo.launch(debug=True,share=True)

webui-test.py

@@ -0,0 +1,354 @@
+import gradio as gr
+import threading
+import asyncio
+import logging
+from concurrent.futures import ThreadPoolExecutor
+from functools import lru_cache
+import requests
+import json
+
+# 假设这些是您的自定义模块，需要根据实际情况进行调整
+from Config.config import VECTOR_DB, DB_directory
+from Ollama_api.ollama_api import *
+from rag.rag_class import *
+
+# 设置日志
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+# 根据VECTOR_DB选择合适的向量数据库
+if VECTOR_DB == 1:
+    from embeding.chromadb import ChromaDB as vectorDB
+    vectordb = vectorDB(persist_directory=DB_directory)
+elif VECTOR_DB == 2:
+    from embeding.faissdb import FaissDB as vectorDB
+    vectordb = vectorDB(persist_directory=DB_directory)
+elif VECTOR_DB == 3:
+    from embeding.elasticsearchStore import ElsStore as vectorDB
+    vectordb = vectorDB()
+
+# 存储上传的文件
+uploaded_files = []
+
+@lru_cache(maxsize=100)
+def get_knowledge_base_files():
+    cl_dict = {}
+    cols = vectordb.get_all_collections_name()
+    for c_name in cols:
+        cl_dict[c_name] = vectordb.get_collcetion_content_files(c_name)
+    return cl_dict
+
+knowledge_base_files = get_knowledge_base_files()
+
+def upload_files(files):
+    if files:
+        new_files = [file.name for file in files]
+        uploaded_files.extend(new_files)
+        update_knowledge_base_files()
+        logger.info(f"Uploaded files: {new_files}")
+        return update_file_list(), new_files, "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Upload successful!</div>"
+    update_knowledge_base_files()
+    return update_file_list(), [], "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Upload failed!</div>"
+
+def delete_files(selected_files):
+    global uploaded_files
+    uploaded_files = [f for f in uploaded_files if f not in selected_files]
+    if selected_files:
+        update_knowledge_base_files()
+        logger.info(f"Deleted files: {selected_files}")
+        return update_file_list(), "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Delete successful!</div>"
+    update_knowledge_base_files()
+    return update_file_list(), "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Delete failed!</div>"
+
+def delete_collection(selected_knowledge_base):
+    if selected_knowledge_base and selected_knowledge_base != "创建知识库":
+        vectordb.delete_collection(selected_knowledge_base)
+        update_knowledge_base_files()
+        logger.info(f"Deleted collection: {selected_knowledge_base}")
+        return update_knowledge_base_dropdown(), "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Collection deleted successfully!</div>"
+    return update_knowledge_base_dropdown(), "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Delete collection failed!</div>"
+
+async def async_vectorize_files(selected_files, selected_knowledge_base, new_kb_name, chunk_size, chunk_overlap):
+    if selected_files:
+        if selected_knowledge_base == "创建知识库":
+            knowledge_base = new_kb_name
+            vectordb.create_collection(selected_files, knowledge_base, chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+        else:
+            knowledge_base = selected_knowledge_base
+            vectordb.add_chroma(selected_files, knowledge_base, chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+
+        if knowledge_base not in knowledge_base_files:
+            knowledge_base_files[knowledge_base] = []
+        knowledge_base_files[knowledge_base].extend(selected_files)
+
+        logger.info(f"Vectorized files: {selected_files} for knowledge base: {knowledge_base}")
+        await asyncio.sleep(0)  # 允许其他任务执行
+        return f"Vectorized files: {', '.join(selected_files)}\nKnowledge Base: {knowledge_base}\nUploaded Files: {', '.join(uploaded_files)}", "<div style='color: green; padding: 10px; border: 2px solid green; border-radius: 5px;'>Vectorization successful!</div>"
+    return "", "<div style='color: red; padding: 10px; border: 2px solid red; border-radius: 5px;'>Vectorization failed!</div>"
+
+def update_file_list():
+    return gr.update(choices=uploaded_files, value=[])
+
+def search_knowledge_base(selected_knowledge_base):
+    if selected_knowledge_base in knowledge_base_files:
+        kb_files = knowledge_base_files[selected_knowledge_base]
+        return gr.update(choices=kb_files, value=[])
+    return gr.update(choices=[], value=[])
+
+def update_knowledge_base_files():
+    global knowledge_base_files
+    knowledge_base_files = get_knowledge_base_files()
+
+# 处理聊天消息的函数
+chat_history = []
+
+def safe_chat_response(model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, message):
+    try:
+        return chat_response(model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, message)
+    except Exception as e:
+        logger.error(f"Error in chat response: {str(e)}")
+        return f"<div style='color: red;'>Error: {str(e)}</div>", ""
+
+def chat_response(model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, message):
+    global chat_history
+    if message:
+        chat_history.append(("User", message))
+        if chat_knowledge_base_dropdown == "仅使用模型":
+            rag = RAG_class(model=model_dropdown,persist_directory=DB_directory)
+            answer = rag.mult_chat(chat_history)
+        if chat_knowledge_base_dropdown and chat_knowledge_base_dropdown != "仅使用模型":
+            rag = RAG_class(model=model_dropdown, embed=vector_dropdown, c_name=chat_knowledge_base_dropdown, persist_directory=DB_directory)
+            if chain_dropdown == "复杂召回方式":
+                questions = rag.decomposition_chain(message)
+                answer = rag.rag_chain(questions)
+            elif chain_dropdown == "简单召回方式":
+                answer = rag.simple_chain(message)
+            else:
+                answer = rag.rerank_chain(message)
+
+        response = f" {answer}"
+        chat_history.append(("Bot", response))
+    return format_chat_history(chat_history), ""
+
+def clear_chat():
+    global chat_history
+    chat_history = []
+    return format_chat_history(chat_history)
+
+def format_chat_history(history):
+    formatted_history = ""
+    for user, msg in history:
+        if user == "User":
+            formatted_history += f'''
+            <div style="text-align: right; margin: 10px;">
+                <div style="display: inline-block; background-color: #DCF8C6; padding: 10px; border-radius: 10px; max-width: 60%;">
+                    {msg}
+                </div>
+                <b>:User</b>
+            </div>
+            '''
+        else:
+            if "```" in msg:  # 检测是否包含代码片段
+                code_content = msg.split("```")[1]
+                formatted_history += f'''
+                <div style="text-align: left; margin: 10px;">
+                    <b>Bot:</b>
+                    <div style="display: inline-block; background-color: #F1F0F0; padding: 10px; border-radius: 10px; max-width: 60%;">
+                        <pre><code>{code_content}</code></pre>
+                    </div>
+                </div>
+                '''
+            else:
+                formatted_history += f'''
+                <div style="text-align: left; margin: 10px;">
+                    <b>Bot:</b>
+                    <div style="display: inline-block; background-color: #F1F0F0; padding: 10px; border-radius: 10px; max-width: 60%;">
+                        {msg}
+                    </div>
+                </div>
+                '''
+    return formatted_history
+
+def clear_status():
+    upload_status.update("")
+    delete_status.update("")
+    vectorize_status.update("")
+    delete_collection_status.update("")
+
+def handle_knowledge_base_selection(selected_knowledge_base):
+    if selected_knowledge_base == "创建知识库":
+        return gr.update(visible=True, interactive=True), gr.update(choices=[], value=[]), gr.update(visible=False)
+    elif selected_knowledge_base == "仅使用模型":
+        return gr.update(visible=False, interactive=False), gr.update(choices=[], value=[]), gr.update(visible=False)
+    else:
+        return gr.update(visible=False, interactive=False), search_knowledge_base(selected_knowledge_base), gr.update(visible=True)
+
+def update_knowledge_base_dropdown():
+    global knowledge_base_files
+    choices = ["创建知识库"] + list(knowledge_base_files.keys())
+    return gr.update(choices=choices)
+
+def update_chat_knowledge_base_dropdown():
+    global knowledge_base_files
+    choices = ["仅使用模型"] + list(knowledge_base_files.keys())
+    return gr.update(choices=choices)
+
+
+# SearxNG搜索函数
+def search_searxng(query):
+    searxng_url = 'http://localhost:8080/search'  # 替换为你的SearxNG实例URL
+    params = {
+        'q': query,
+        'format': 'json'
+    }
+    response = requests.get(searxng_url, params=params)
+    response.raise_for_status()
+    return response.json()
+
+
+# Ollama总结函数
+def summarize_with_ollama(model_dropdown,text, question):
+    prompt = """
+        根据下边的内容，回答用户问题，
+        内容为：‘{0}‘\n
+        问题为：{1}
+    """.format(text, question)
+    ollama_url = 'http://localhost:11434/api/generate'  # 替换为你的Ollama实例URL
+    data = {
+        'model': model_dropdown,
+        "prompt": prompt,
+        "stream": False
+    }
+    response = requests.post(ollama_url, json=data)
+    response.raise_for_status()
+    return response.json()
+
+
+# 处理函数
+def ai_web_search(model_dropdown,user_query):
+    # 使用SearxNG进行搜索
+    search_results = search_searxng(user_query)
+    search_texts = [result['title'] + "\n" + result['content'] for result in search_results['results']]
+    combined_text = "\n\n".join(search_texts)
+
+    # 使用Ollama进行总结
+    summary = summarize_with_ollama(model_dropdown,combined_text, user_query)
+    # print(summary)
+    # 返回结果
+    return summary['response']
+# 添加新的函数来处理AI网络搜索
+# def ai_web_search(model_dropdown, query):
+#     try:
+#         # 这里添加实际的网络搜索和AI处理逻辑
+#         # 这只是一个示例，您需要根据实际情况实现
+#         search_result = f"搜索结果: {query}"
+#         ai_response = f"AI回答: 基于搜索结果，对于'{query}'的回答是..."
+#         return f"{search_result}\n\n{ai_response}"
+#     except Exception as e:
+#         logger.error(f"Error in AI web search: {str(e)}")
+#         return f"<div style='color: red;'>Error: {str(e)}</div>"
+
+# 创建 Gradio 界面
+with gr.Blocks() as demo:
+    with gr.Column():
+        # 添加标题
+        title = gr.HTML("<h1 style='text-align: center; font-size: 32px; font-weight: bold;'>RAG精致系统</h1>")
+        # 添加公告栏
+        announcement = gr.HTML("<div style='text-align: center; font-size: 18px; color: red;'>公告栏: 欢迎使用RAG精致系统，一个适合学习、使用、自主扩展的【检索增强生成】系统！<br/>公众号：世界大模型</div>")
+
+        with gr.Tabs():
+            with gr.TabItem("知识库"):
+                knowledge_base_dropdown = gr.Dropdown(choices=["创建知识库"] + list(knowledge_base_files.keys()),
+                                                      label="选择知识库")
+                new_kb_input = gr.Textbox(label="输入新的知识库名称", visible=False, interactive=True)
+                file_input = gr.Files(label="Upload files")
+                upload_btn = gr.Button("Upload")
+                file_list = gr.CheckboxGroup(label="Uploaded Files")
+                delete_btn = gr.Button("Delete Selected Files")
+                with gr.Row():
+                    chunk_size_dropdown = gr.Dropdown(choices=[50, 100, 200, 300, 500, 700], label="chunk_size", value=200)
+                    chunk_overlap_dropdown = gr.Dropdown(choices=[20, 50, 100, 200], label="chunk_overlap", value=50)
+                    vectorize_btn = gr.Button("Vectorize Selected Files")
+                delete_collection_btn = gr.Button("Delete Collection")
+                upload_status = gr.HTML()
+                delete_status = gr.HTML()
+                vectorize_status = gr.HTML()
+                delete_collection_status = gr.HTML()
+
+            with gr.TabItem("Chat"):
+                with gr.Row():
+                    model_dropdown = gr.Dropdown(choices=get_llm(), label="模型")
+                    vector_dropdown = gr.Dropdown(choices=get_embeding_model(), label="向量")
+                    chat_knowledge_base_dropdown = gr.Dropdown(choices=["仅使用模型"] + vectordb.get_all_collections_name(), label="知识库")
+                    chain_dropdown = gr.Dropdown(choices=["复杂召回方式", "简单召回方式","rerank"], label="chain方式", visible=False)
+                chat_display = gr.HTML(label="Chat History")
+                chat_input = gr.Textbox(label="Type a message")
+                chat_btn = gr.Button("Send")
+                clear_btn = gr.Button("Clear Chat History")
+
+            with gr.TabItem("AI网络搜索"):
+                with gr.Row():
+                    web_search_model_dropdown = gr.Dropdown(choices=get_llm(), label="模型")
+                web_search_output = gr.Textbox(label="搜索结果和AI回答", lines=10)
+                web_search_input = gr.Textbox(label="输入搜索查询")
+
+                web_search_btn = gr.Button("搜索")
+
+    def handle_upload(files):
+        upload_result, new_files, status = upload_files(files)
+        threading.Thread(target=clear_status).start()
+        return upload_result, new_files, status, update_chat_knowledge_base_dropdown()
+
+    def handle_delete(selected_knowledge_base, selected_files):
+        tmp = []
+        cols_files_tmp = vectordb.get_collcetion_content_files(c_name=selected_knowledge_base)
+        for i in selected_files:
+            if i in cols_files_tmp:
+                tmp.append(i)
+        del cols_files_tmp
+        if tmp:
+            vectordb.del_files(tmp, c_name=selected_knowledge_base)
+        del tmp
+        delete_result, status = delete_files(selected_files)
+        threading.Thread(target=clear_status).start()
+        return delete_result, status, update_chat_knowledge_base_dropdown()
+
+    def handle_vectorize(selected_files, selected_knowledge_base, new_kb_name, chunk_size, chunk_overlap):
+        vectorize_result, status = asyncio.run(async_vectorize_files(selected_files, selected_knowledge_base, new_kb_name, chunk_size, chunk_overlap))
+        threading.Thread(target=clear_status).start()
+        return vectorize_result, status, update_knowledge_base_dropdown(), update_chat_knowledge_base_dropdown()
+
+    def handle_delete_collection(selected_knowledge_base):
+        result, status = delete_collection(selected_knowledge_base)
+        threading.Thread(target=clear_status).start()
+        return result, status, update_chat_knowledge_base_dropdown()
+
+    knowledge_base_dropdown.change(
+        handle_knowledge_base_selection,
+        inputs=knowledge_base_dropdown,
+        outputs=[new_kb_input, file_list, chain_dropdown]
+    )
+    upload_btn.click(handle_upload, inputs=file_input, outputs=[file_list, file_list, upload_status, chat_knowledge_base_dropdown])
+    delete_btn.click(handle_delete, inputs=[knowledge_base_dropdown, file_list], outputs=[file_list, delete_status, chat_knowledge_base_dropdown])
+    vectorize_btn.click(handle_vectorize, inputs=[file_list, knowledge_base_dropdown, new_kb_input, chunk_size_dropdown, chunk_overlap_dropdown],
+                        outputs=[gr.Textbox(visible=False), vectorize_status, knowledge_base_dropdown, chat_knowledge_base_dropdown])
+    delete_collection_btn.click(handle_delete_collection, inputs=knowledge_base_dropdown,
+                                outputs=[knowledge_base_dropdown, delete_collection_status, chat_knowledge_base_dropdown])
+
+    chat_btn.click(chat_response, inputs=[model_dropdown, vector_dropdown, chat_knowledge_base_dropdown, chain_dropdown, chat_input], outputs=[chat_display, chat_input])
+    clear_btn.click(clear_chat, outputs=chat_display)
+
+    chat_knowledge_base_dropdown.change(
+        fn=lambda selected: gr.update(visible=selected != "仅使用模型"),
+        inputs=chat_knowledge_base_dropdown,
+        outputs=chain_dropdown
+    )
+
+    # 添加新的点击事件处理
+    web_search_btn.click(
+        ai_web_search,
+        inputs=[web_search_model_dropdown, web_search_input],
+        outputs=web_search_output
+    )
+
+demo.launch(debug=True,share=True)