app.py

import json, requests
import gradio as gr
import uuid
import time
import random
import os

proxies = None
proxies = {
    'http': 'http://127.0.0.1:7680',
    'https': 'https://127.0.0.1:7680',
    # 'socks5':"socks5://127.0.0.1:10808",
}

# 创建新的聊天频道
def generate_uuid():
    random_uuid = uuid.uuid4()
    random_uuid_str = str(random_uuid)
    formatted_uuid = f"{random_uuid_str[0:8]}-{random_uuid_str[9:13]}-{random_uuid_str[14:18]}-{random_uuid_str[19:23]}-{random_uuid_str[24:]}"
    return formatted_uuid


# Deletes the conversation
def delete_conversation(user_id,channel_id,cookie):
    url = f"https://claude.ai/api/organizations/{user_id}/chat_conversations/{channel_id}"

    payload = json.dumps(f"{channel_id}")
    headers = {
        'User-Agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36 Edg/114.0.1823.82',
        'Accept-Language': 'en-US,en;q=0.5',
        'Content-Type': 'application/json',
        'Content-Length': '38',
        'Referer': 'https://claude.ai/chats',
        'Origin': 'https://claude.ai',
        'Sec-Fetch-Dest': 'empty',
        'Sec-Fetch-Mode': 'cors',
        'Sec-Fetch-Site': 'same-origin',
        'Connection': 'keep-alive',
        'Cookie': f'{cookie}',
        'TE': 'trailers'
    }

    response = requests.delete(url, headers=headers, data=payload)
    # Returns True if deleted or False if any error in deleting
    if response.status_code == 204:
        return True
    else:
        return False

def create_new_chat(user_id,cookie):
    url = f"https://claude.ai/api/organizations/{user_id}/chat_conversations"
    uuid = generate_uuid()

    payload = json.dumps({"uuid": uuid, "name": ""})
    headers = {
        'User-Agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36 Edg/114.0.1823.82',
        'Accept-Language': 'zh-CN,zh;q=0.9,en;q=0.8,en-GB;q=0.7,en-US;q=0.6',
        'Referer': 'https://claude.ai/chats',
        'Content-Type': 'application/json',
        'Origin': 'https://claude.ai',
        'Connection': 'keep-alive',
        'Cookie': cookie,
        'Sec-Fetch-Dest': 'empty',
        'Sec-Fetch-Mode': 'cors',
        'Sec-Fetch-Site': 'same-origin',
    }

    # response = requests.request("POST", url, headers=headers, data=payload)
    # 使用代理，解决国内网络不能访问问题
    # response = request("POST", url, headers=headers, data=payload, proxies=proxies)
    response = requests.post(url, headers=headers, data=payload)
    print(f"新的对话页面：{response.json()}")

    # Returns JSON of the newly created conversation information
    return response.json()

def get_content_type(file_path):
    # Function to determine content type based on file extension
    extension = os.path.splitext(file_path)[-1].lower()
    if extension == '.pdf':
        return 'application/pdf'
    elif extension == '.txt':
        return 'text/plain'
    elif extension == '.csv':
        return 'text/csv'
    # Add more content types as needed for other file types
    else:
        return 'application/octet-stream'

def upload_attachment(cookie ,organization_id,file_path):
    ## 自己解析
    if file_path.endswith(('.txt', '.pdf', '.csv')):
        file_name = os.path.basename(file_path)
        file_size = os.path.getsize(file_path)
        file_type = "text/plain"
        with open(file_path, 'r', encoding='utf-8') as file:
            file_content = file.read()
            print({
            "file_name": file_name,
            "file_type": file_type,
            "file_size": file_size,
            "extracted_content": file_content})

        return {
            "file_name": file_name,
            "file_type": file_type,
            "file_size": file_size,
            "extracted_content": file_content
        }



    url = 'https://claude.ai/api/convert_document'
    headers = {
        'User-Agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36 Edg/114.0.1823.82',
        'Accept-Language': 'zh-CN,zh;q=0.9,en;q=0.8,en-GB;q=0.7,en-US;q=0.6',
        'Referer': 'https://claude.ai/chats',
        'Origin': 'https://claude.ai',
        'Sec-Fetch-Dest': 'empty',
        'Sec-Fetch-Mode': 'cors',
        'Sec-Fetch-Site': 'same-origin',
        'Connection': 'keep-alive',
        'Cookie': f'{cookie}',
        'TE': 'trailers'
    }

    file_name = os.path.basename(file_path)
    print(file_name)
    content_type = get_content_type(file_path)
    print(content_type)
    files = {
        'file': (file_name, open(file_path, 'rb'), content_type),
        'orgUuid': (None, organization_id)
        # 'orgUuid': organization_id
    }
    print(files)
    response = requests.post(url, headers=headers, files=files)
    if response.status_code == 200:
        return response.json()
    else:
        return False



def request_to_v2(message, cookie, user_id,attachment=None,delst=True,context=[]):
    context = [message]
    
    
    ## 是否加入文件
    # Upload attachment if provided
    
    attachments = []
    if attachment:
        attachment = attachment.name
        attachment_response = upload_attachment(cookie,user_id,attachment)
        if attachment_response:
            attachments = [attachment_response]
            print(f"attachment_response:{attachment_response}")
        else:
            bots = "\nError: Invalid file format. Please try again."
            print(bots)
            context += [bots]
            responses = [(u, b) for u, b in zip(context[::2], context[1::2])]
            return responses, context

    # Ensure attachments is an empty list when no attachment is provided
    if not attachment:
        attachments = []        



    response_json = create_new_chat(user_id,cookie)
    channel_id = response_json.get('uuid',"")
    print(channel_id)

    
    timeout = 5*60 #5分钟不回复，显示超时
    context = [message]
    headers = {
        # ':Authority': 'claude.ai',
        # ':Method': 'POST',
        # ':Path': '/api/append_message',
        # ':Scheme': 'https',
        'Content-Type': 'application/json',
        'Origin': 'https://claude.ai',
        'Referer': f'https://claude.ai/chat/{channel_id}',
        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36',
        'Cookie': cookie,
        'Accept-Language': 'zh-CN,zh;q=0.9,en;q=0.8',
        # 'Accept-Encoding': 'gzip, deflate, br',
        'Accept': 'text/event-stream, text/event-stream',
        'Sec-Ch-Ua': '"Not.A/Brand";v="8", "Chromium";v="114", "Google Chrome";v="114"',
        'Sec-Ch-Ua-Mobile': '?0',
        'Sec-Ch-Ua-Platform': '"Windows"',
        'Sec-Fetch-Dest': 'empty',
        'Sec-Fetch-Mode': 'cors',
        'Sec-Fetch-Site': 'same-origin',
        'Connection': 'keep-alive',
    }
    post_msg_url = 'https://claude.ai/api/append_message'
    post_msg_data = {
        "completion":{
            "incremental":True,
            "prompt": message,
            "timezone":"Asia/Shanghai",
            "model":"claude-2"
        },
        "organization_uuid": user_id,
        "conversation_uuid": channel_id,
        "text": message,
        "attachments": attachments
    }
    post_msg_data = json.dumps(post_msg_data)
    # print(post_msg_data)

    response = {
        "status_code":-1,
        "text":"报错啦"
    }

    try:
        print(f"输入的问题：{message}")
        if proxies:
            # response = requests.post(post_msg_url, headers=headers, data=post_msg_data, verify=False, stream=False,timeout = timeout)
            print("使用代理.")
            response = requests.post(post_msg_url, headers=headers, data=post_msg_data, verify=False, stream=False,timeout = timeout, proxies=proxies)
        else:
            print("******不使用代理.")
            response = requests.post(post_msg_url, headers=headers, data=post_msg_data, verify=False, stream=False,timeout = timeout)
            
        bots = ""
        print("a"*100)
        for data in response.iter_lines():
            if data:
                # 将字节字符串转换为字符串
                data_str = data.decode('utf-8')
                # 删除开头的 "data: " 部分
                data_str = data_str.replace("data: ", "")
                data_json = json.loads(data_str)
                # print(data_json)
                bots +=  data_json["completion"]
        print(bots)
        context += [bots]
        responses = [(u, b) for u, b in zip(context[::2], context[1::2])]
        return responses, context

    except Exception as e:
        print(">>>>>> 查询失败")
        print(e)
        print(response.text)

        response =">>>>>> 查询失败\n报错信息为："+ str(e) + f"\n response信息:\n错误代码：{response.status_code}\n错误信息{response.text}"
        context += [response]
        responses = [(u, b) for u, b in zip(context[::2], context[1::2])]

        return responses, context

    finally:
        if delst:
            time.sleep(random.uniform(1,2))
            ds = delete_conversation(user_id,channel_id,cookie)
            if ds:
                print(f"成功删除")
            else:
                print(f"删除失败")

        print("*-"*100)



with gr.Blocks() as dialog_app:
    gr.HTML("""<h1 align="center">Claude2-API-xiaolv-第3版-单轮对话-支持文件知识库</h1>""")
    with gr.Tab("Claude2 API xiaolv"):
        gr.Markdown("""
        ## 需要传入的参数有：
        1.cookie：intercom-device-id-lupk8zyo=8888
        2.user_id：对应 organization_uuid
        3.file：上传文件，可传可不传，如果需要基于文件进行问答，可以选择上传（支持：txt、csv、pdf格式）
        ## 特点：
        1. 根据传入的cookie 和 user_id 自动创建一个新的聊天频道（conversation_uuid/channel_id）；
        2. 在新的聊天聊天频道进行对话；
        3. 请求放回后，自动删除新的聊天频道，避免频繁调用导致页面太多记录。
        4. 模型只能进行一次聊天，对于特定场景比较有用，防止模型记住历史数据，影响模型对实际数据的输出额判断。
        """)
        gr.HTML(
        """<p>5.如果想多轮对话的话，可以参考以下这个地址：<a href="https://huggingface.co/spaces/xiaolv/claude2_xiaolv">Claude2-API-xiaolv-第1版-多轮对话</a></p>"""
        """<p>6.如果只想回复一次的话，可以参考以下这个地址：<a href="https://huggingface.co/spaces/xiaolv/claude2_xiaolv_api_updata">Claude2-API-xiaolv-第2版-单轮对话</a></p>"""
        """<p>如果喜欢请给个 💖 </p>"""
        )
        

        with gr.Row():
            with gr.Column(scale=0.4):
                cookies = gr.Textbox(lines=2, label="输入cookies")
                user_id = gr.Textbox(lines=2, label="输入user_id/organization_uuid")
                file = gr.File(label='请上传知识库文件, 目前支持txt、csv、pdf格式,不超过10M，可传可不传',file_types=['.txt', '.csv', '.pdf'])
                delst = gr.Checkbox(label="是", info="在该对话回复后，是否要删除该对话？（默认会删除）",value = True)
                # channel_id = gr.Textbox(lines=2, label="输入channel_id/conversation_uuid")
                # chatbot = gr.Chatbot()
            with gr.Column(scale=0.6):
                chatbot = gr.Chatbot([])

            state = gr.State([])

        with gr.Row():
            inputs = gr.Textbox(
                label="输入问题",
                placeholder="请输入你的文本，确保已经正确填入cookies、user_id"
            )
        inputs.submit(request_to_v2, [inputs, cookies, user_id,file,delst,state], [chatbot, state])
        send = gr.Button("发送请求.....")
        send.click(request_to_v2, [inputs, cookies, user_id,file,delst,state], [chatbot, state], api_name="xiaolvgpt", show_progress=True)


        gr.Markdown("""
        清除历史记录是，请输入：```/resrt```""")   
        
# 为网站设置密码防止滥用
# dialog_app.launch(auth=("admin", "2240560729"),show_error=True,show_api=True,share=True)
# dialog_app.launch(auth=("admin", "2240560729"),show_error=True)
# dialog_app.launch(show_error=True,server_port = 7861)
dialog_app.launch(show_error=True)
# dialog_app.launch(show_error=True,show_api=True,share=True)