Chainlit快速实现AI对话应用将聊天数据的持久化到内存中

泰山AI

于 2024-08-09 20:02:26 发布

阅读量500

点赞数 12

分类专栏：微软GraphRAG 文章标签：人工智能 python 开发语言 chatlit llm

本文链接：https://blog.csdn.net/weixin_40986713/article/details/141069960

版权

微软GraphRAG 专栏收录该内容

12 篇文章 1 订阅

订阅专栏

概述

默认情况下，Chainlit 应用不会保留其生成的聊天和元素。即网页一刷新，所有的聊天记录，页面上的所有聊天记录都会消失。但是，存储和利用这些数据的能力可能是您的项目或组织的重要组成部分。

之前讲到通过Literal AI可以一分钟快速实现聊天记录数据持久化《Chainlit快速实现AI对话应用1 分钟内实现聊天数据的持久化保存》，但是Literal AI在国内无法正常访问，所以还需自定义聊天数据到本地，才能方便国内使用。

教程

安装chainlit依赖

pip install chainlit

配置环境变量
在项目根目录下，创建.env文件，内容如下：

OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1"
OPENAI_API_KEY="your api_key"

由于国内无法访问open ai的chatgpt，所以需要配置 OPENAI_BASE_URL的代理地址，如果使用国内的LLM大模型接口，可以使用兼容open ai的接口地址

创建代码
在项目根目录下，创建一个app.py的文件，代码如下：

import uuid
from typing import Dict, List, Optional

import chainlit.data as cl_data
from chainlit.step import StepDict
from literalai.helper import utc_now
from openai import AsyncOpenAI
import chainlit as cl

client = AsyncOpenAI()

now = utc_now()

thread_history = []  # type: List[cl_data.ThreadDict]
deleted_thread_ids = []  # type: List[str]


class CustomDataLayer(cl_data.BaseDataLayer):
    async def get_user(self, identifier: str):
        return cl.PersistedUser(id="",createdAt=now, identifier=identifier)

    async def create_user(self, user: cl.User):
        return cl.PersistedUser(id="",createdAt=now, identifier=user.identifier)

    async def update_thread(
            self,
            thread_id: str,
            name: Optional[str] = None,
            user_id: Optional[str] = None,
            metadata: Optional[Dict] = None,
            tags: Optional[List[str]] = None,
    ):
        thread = next((t for t in thread_history if t["id"] == thread_id), None)
        print('name', name)
        print('thread', thread)
        print('user_id', user_id)
        print('metadata', metadata)
        if thread:
            if name:
                thread["name"] = name
            if user_id:
                thread["user_id"] = user_id
            if metadata:
                thread["metadata"] = metadata
            if tags:
                thread["tags"] = tags
        else:
            thread_history.append(
                {
                    "id": thread_id,
                    "name": name,
                    "metadata": metadata,
                    "tags": tags,
                    "createdAt": utc_now(),
                    "userId": user_id,
                    "userIdentifier": user_id,
                    "steps": [],
                }
            )

    @cl_data.queue_until_user_message()
    async def create_step(self, step_dict: StepDict):
        thread = next(
            (t for t in thread_history if t["id"] == step_dict.get("threadId")), None
        )
        if thread:
            thread["steps"].append(step_dict)

    @cl_data.queue_until_user_message()
    async def update_step(self, step_dict: "StepDict"):
        await self.create_step(step_dict)

    async def get_thread_author(self, thread_id: str) -> str:
        thread = next(
            (t for t in thread_history if t["id"] == thread_id), None
        )
        if not thread:
            return ""
        user_identifier = thread.get("userIdentifier")
        if not user_identifier:
            return ""

        return user_identifier

    async def list_threads(
            self, pagination: cl_data.Pagination, filters: cl_data.ThreadFilter
    ) -> cl_data.PaginatedResponse[cl_data.ThreadDict]:
        return cl_data.PaginatedResponse(
            data=[t for t in thread_history if t["id"] not in deleted_thread_ids],
            pageInfo=cl_data.PageInfo(
                hasNextPage=False, startCursor=None, endCursor=None
            ),
        )

    async def get_thread(self, thread_id: str):
        thread = next((t for t in thread_history if t["id"] == thread_id), None)
        if not thread:
            return None
        thread["steps"] = sorted(thread["steps"], key=lambda x: x["createdAt"])
        return thread

    async def delete_thread(self, thread_id: str):
        deleted_thread_ids.append(thread_id)


cl_data._data_layer = CustomDataLayer()


@cl.on_chat_start
async def main():
    content = "你好，我是泰山AI智能客服，有什么可以帮助您吗？"
    await cl.Message(content).send()


@cl.on_message
async def handle_message():
    # Wait for queue to be flushed
    await cl.sleep(1)
    msg = cl.Message(content="")
    await msg.send()

    stream = await client.chat.completions.create(
        model="qwen-turbo", messages=cl.chat_context.to_openai(), stream=True
    )

    async for part in stream:
        if token := part.choices[0].delta.content or "":
            await msg.stream_token(token)
    await msg.update()


@cl.password_auth_callback
def auth_callback(username: str, password: str) -> Optional[cl.User]:
    if (username, password) == ("admin", "admin"):
        return cl.User(identifier="admin")
    else:
        return cl.User(identifier=uuid.uuid4().hex)


@cl.on_chat_resume
async def on_chat_resume():
    pass

执行命令创建 AUTH_SECRET 鉴权

chainlit create-secret

在这里插入图片描述
复制最后一行代码到.env环境配置文件中

CHAINLIT_AUTH_SECRET="$b?/v0NeJlAU~I5As1WSCa,j8wJ3w%agTyIFlUt4408?mfC*,/wovlfA%3O/751U"
OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1"
OPENAI_API_KEY=""