How can I change from OpenAI to ChatOpenAI in langchain and Flask?

营赢盈英2024-09-07 15:10

题意："在 LangChain 和 Flask 中，如何将 OpenAI 更改为 ChatOpenAI？"

问题背景：

This is an implementation based on langchain and flask and refers to an implementation to be able to stream responses from the OpenAI server in langchain to a page with javascript that can show the streamed response.

"这是一个基于 LangChain 和 Flask 的实现，用于将 OpenAI 服务器的响应流式传输到一个带有 JavaScript 的页面，该页面可以显示流式响应。"

I tried all ways to modify the code below to replace the langchain library from openai to chatopenai without success, i upload below both implementations (the one with openai working) and the one chatopenai with error. thank you to all the community and those who can help me to understand the problem, it would be very helpful if you could also show me how to solve it since I have been trying for days and the error it shows has really no significance.

"我尝试了所有方法修改下面的代码，将 LangChain 库从 OpenAI 替换为 ChatOpenAI，但没有成功。下面上传了两个实现（一个是使用 OpenAI 正常工作的版本，另一个是带有错误的 ChatOpenAI 版本）。感谢所有社区成员以及能够帮助我理解问题的人。如果你们能告诉我如何解决这个问题，那将非常有帮助，因为我已经尝试了好几天，而显示的错误并没有什么实际意义。"

Code version with library that works but reports as deprecated:

"这是使用库的代码版本，它可以正常工作但报告为已弃用："

python 复制代码 
from flask import Flask, Response
import threading
import queue

from langchain.llms import OpenAI
from langchain.callbacks.base import BaseCallbackManager
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler

app = Flask(__name__)

@app.route('/')
def index():
    return Response('''<!DOCTYPE html>
<html>
<head><title>Flask Streaming Langchain Example</title></head>
<body>
    <div id="output"></div>
    <script>
const outputEl = document.getElementById('output');

(async function() {
    try {
        const controller = new AbortController();
        const signal = controller.signal;
        const timeout = 120000; // Imposta il timeout su 120 secondi

        setTimeout(() => controller.abort(), timeout);

        const response = await fetch('/chain', {method: 'POST', signal});
        const reader = response.body.getReader();
        const decoder = new TextDecoder();
        let buffer = '';

        while (true) {
            const { done, value } = await reader.read();
            if (done) { break; }

            const text = decoder.decode(value, {stream: true});
            outputEl.innerHTML += text;
        }
    } catch (err) {
        console.error(err);
    }
})();

    </script>
</body>
</html>''', mimetype='text/html')


class ThreadedGenerator:
    def __init__(self):
        self.queue = queue.Queue()

    def __iter__(self):
        return self

    def __next__(self):
        item = self.queue.get()
        if item is StopIteration: raise item
        return item

    def send(self, data):
        self.queue.put(data)

    def close(self):
        self.queue.put(StopIteration)

class ChainStreamHandler(StreamingStdOutCallbackHandler):
    def __init__(self, gen):
        super().__init__()
        self.gen = gen

    def on_llm_new_token(self, token: str, **kwargs):
        self.gen.send(token)

def llm_thread(g, prompt):
    try:
        llm = OpenAI(
            model_name="gpt-4",
            verbose=True,
            streaming=True,

            callback_manager=BaseCallbackManager([ChainStreamHandler(g)]),
            temperature=0.7,
        )
        llm(prompt)
    finally:
        g.close()


def chain(prompt):
    g = ThreadedGenerator()
    threading.Thread(target=llm_thread, args=(g, prompt)).start()
    return g


@app.route('/chain', methods=['POST'])
def _chain():
    return Response(chain("Create a poem about the meaning of life \n\n"), mimetype='text/plain')

if __name__ == '__main__':
    app.run(threaded=True, debug=True)

Version with error (OpenAI replaced with ChatOpenAI)"

这是带有错误的版本（将 OpenAI 替换为 ChatOpenAI）："

python 复制代码 
import threading
import queue

from langchain.chat_models import ChatOpenAI
from langchain.callbacks.base import BaseCallbackManager
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler

app = Flask(__name__)

@app.route('/')
def index():
    return Response('''<!DOCTYPE html>
<html>
<head><title>Flask Streaming Langchain Example</title></head>
<body>
    <div id="output"></div>
    <script>
const outputEl = document.getElementById('output');

(async function() {
    try {
        const controller = new AbortController();
        const signal = controller.signal;
        const timeout = 120000; // Imposta il timeout su 120 secondi

        setTimeout(() => controller.abort(), timeout);

        const response = await fetch('/chain', {method: 'POST', signal});
        const reader = response.body.getReader();
        const decoder = new TextDecoder();
        let buffer = '';

        while (true) {
            const { done, value } = await reader.read();
            if (done) { break; }

            const text = decoder.decode(value, {stream: true});
            outputEl.innerHTML += text;
        }
    } catch (err) {
        console.error(err);
    }
})();

    </script>
</body>
</html>''', mimetype='text/html')


class ThreadedGenerator:
    def __init__(self):
        self.queue = queue.Queue()

    def __iter__(self):
        return self

    def __next__(self):
        item = self.queue.get()
        if item is StopIteration: raise item
        return item

    def send(self, data):
        self.queue.put(data)

    def close(self):
        self.queue.put(StopIteration)

class ChainStreamHandler(StreamingStdOutCallbackHandler):
    def __init__(self, gen):
        super().__init__()
        self.gen = gen

    def on_llm_new_token(self, token: str, **kwargs):
        self.gen.send(token)

    def on_chat_model_start(self, token: str):
        print("started")

def llm_thread(g, prompt):
    try:
        llm = ChatOpenAI(
            model_name="gpt-4",
            verbose=True,
            streaming=True,

            callback_manager=BaseCallbackManager([ChainStreamHandler(g)]),
            temperature=0.7,
        )
        llm(prompt)
    finally:
        g.close()


def chain(prompt):
    g = ThreadedGenerator()
    threading.Thread(target=llm_thread, args=(g, prompt)).start()
    return g


@app.route('/chain', methods=['POST'])
def _chain():
    return Response(chain("parlami dei 5 modi di dire in inglese che gli italiani conoscono meno \n\n"), mimetype='text/plain')

if __name__ == '__main__':
    app.run(threaded=True, debug=True)

Error showing the console at startup and at the time I enter the web page:

"启动时和进入网页时控制台显示的错误："

python 复制代码 
Error in ChainStreamHandler.on_chat_model_start callback: ChainStreamHandler.on_chat_model_start() got an unexpected keyword argument 'run_id'
Exception in thread Thread-4 (llm_thread):
127.0.0.1 - - [09/Sep/2023 18:09:29] "POST /chain HTTP/1.1" 200 -
Traceback (most recent call last):
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\callbacks\manager.py", line 300, in _handle_event
    getattr(handler, event_name)(*args, **kwargs)
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\callbacks\base.py", line 168, in on_chat_model_start
    raise NotImplementedError(
NotImplementedError: StdOutCallbackHandler does not implement `on_chat_model_start`

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\user22\AppData\Local\Programs\Python\Python311\Lib\threading.py", line 1038, in _bootstrap_inner    
    self.run()
  File "C:\Users\user22\AppData\Local\Programs\Python\Python311\Lib\threading.py", line 975, in run
    self._target(*self._args, **self._kwargs)
  File "c:\Users\user22\Desktop\Work\TESTPROJ\streamresp.py", line 90, in llm_thread
    llm(prompt)
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\chat_models\base.py", line 552, in __call__
    generation = self.generate(
                 ^^^^^^^^^^^^^^
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\chat_models\base.py", line 293, in generate
    run_managers = callback_manager.on_chat_model_start(
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\callbacks\manager.py", line 1112, in on_chat_model_start
    _handle_event(
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\callbacks\manager.py", line 304, in _handle_event
    message_strings = [get_buffer_string(m) for m in args[1]]
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\callbacks\manager.py", line 304, in <listcomp>
    message_strings = [get_buffer_string(m) for m in args[1]]
                       ^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\user22\Desktop\Work\TESTPROJ\env\Lib\site-packages\langchain\schema\messages.py", line 52, in get_buffer_string
    raise ValueError(f"Got unsupported message type: {m}")
ValueError: Got unsupported message type: p

thank you very much for the support!

"非常感谢您的支持！"

问题解决：

Thanks to python273 user on github I've resolved:

"感谢 GitHub 上的用户 python273，我已经解决了这个问题。"

python 复制代码 
import os
os.environ["OPENAI_API_KEY"] = ""

from flask import Flask, Response, request
import threading
import queue

from langchain.chat_models import ChatOpenAI
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
from langchain.schema import AIMessage, HumanMessage, SystemMessage

app = Flask(__name__)

@app.route('/')
def index():
    # just for the example, html is included directly, move to .html file
    return Response('''
<!DOCTYPE html>
<html>
<head><title>Flask Streaming Langchain Example</title></head>
<body>
    <form id="form">
        <input name="prompt" value="write a short koan story about seeing beyond"/>
        <input type="submit"/>
    </form>
    <div id="output"></div>
    <script>
        const formEl = document.getElementById('form');
        const outputEl = document.getElementById('output');

        let aborter = new AbortController();
        async function run() {
            aborter.abort();  // cancel previous request
            outputEl.innerText = '';
            aborter = new AbortController();
            const prompt = new FormData(formEl).get('prompt');
            try {
                const response = await fetch(
                    '/chain', {
                        signal: aborter.signal,
                        method: 'POST',
                        headers: {'Content-Type': 'application/json'},
                        body: JSON.stringify({
                            prompt
                        }),
                    }
                );
                const reader = response.body.getReader();
                const decoder = new TextDecoder();
                while (true) {
                    const { done, value } = await reader.read();
                    if (done) { break; }
                    const decoded = decoder.decode(value, {stream: true});
                    outputEl.innerText += decoded;
                }
            } catch (err) {
                console.error(err);
            }
        }
        run();  // run on initial prompt
        formEl.addEventListener('submit', function(event) {
            event.preventDefault();
            run();
        });
    </script>
</body>
</html>
''', mimetype='text/html')

class ThreadedGenerator:
    def __init__(self):
        self.queue = queue.Queue()

    def __iter__(self):
        return self

    def __next__(self):
        item = self.queue.get()
        if item is StopIteration: raise item
        return item

    def send(self, data):
        self.queue.put(data)

    def close(self):
        self.queue.put(StopIteration)

class ChainStreamHandler(StreamingStdOutCallbackHandler):
    def __init__(self, gen):
        super().__init__()
        self.gen = gen

    def on_llm_new_token(self, token: str, **kwargs):
        self.gen.send(token)

def llm_thread(g, prompt):
    try:
        chat = ChatOpenAI(
            verbose=True,
            streaming=True,
            callbacks=[ChainStreamHandler(g)],
            temperature=0.7,
        )
        chat([HumanMessage(content=prompt)])
    finally:
        g.close()


def chain(prompt):
    g = ThreadedGenerator()
    threading.Thread(target=llm_thread, args=(g, prompt)).start()
    return g


@app.route('/chain', methods=['POST'])
def _chain():
    return Response(chain(request.json['prompt']), mimetype='text/plain')

if __name__ == '__main__':
    app.run(threaded=True, debug=True)

Link to the original reply: https://gist.github.com/python273/563177b3ad5b9f74c0f8f3299ec13850

"原始回复的链接： https://gist.github.com/python273/563177b3ad5b9f74c0f8f3299ec13850"

相关推荐
阡之尘埃1 小时前
Python数据分析案例61——信贷风控评分卡模型(A卡)（scorecardpy 全面解析）
人工智能·python·机器学习·数据分析·智能风控·信贷风控
丕羽5 小时前
【Pytorch】基本语法
人工智能·pytorch·python
bryant_meng5 小时前
【python】Distribution
开发语言·python·分布函数·常用分布
m0_594526306 小时前
Python批量合并多个PDF
java·python·pdf
工业互联网专业6 小时前
Python毕业设计选题：基于Hadoop的租房数据分析系统的设计与实现
vue.js·hadoop·python·flask·毕业设计·源码·课程设计
钱钱钱端6 小时前
【压力测试】如何确定系统最大并发用户数？
自动化测试·软件测试·python·职场和发展·压力测试·postman
慕卿扬6 小时前
基于python的机器学习（二）—— 使用Scikit-learn库
笔记·python·学习·机器学习·scikit-learn
Json____7 小时前
python的安装环境Miniconda（Conda 命令管理依赖配置）
开发语言·python·conda·miniconda
小袁在上班7 小时前
Python 单元测试中的 Mocking 与 Stubbing：提高测试效率的关键技术
python·单元测试·log4j
白狐欧莱雅7 小时前
使用python中的pygame简单实现飞机大战游戏
经验分享·python·游戏·pygame
热门推荐
01【HarmonyOS】HUAWEI DevEco Studio 下载地址汇总02(欧拉)openEuler系统添加网卡文件配置流程、(欧拉)openEuler系统手动配置ipv6地址流程、(欧拉)openEuler系统网络管理说明03组基轨迹建模 GBTM的介绍与实现（Stata 或 R）04【AIGC】重塑未来的科技巨轮05全面解析：构建基于深度学习的安全帽检测系统（UI界面+YOLO代码+数据集）06【经验分享】Ubuntu22.04安装微信(linux官方版)07基于YOLOv10深度学习的CT扫描图像肾结石智能检测系统【python源码+Pyqt5界面+数据集+训练代码】深度学习实战、目标检测08Ubuntu 20.04使用Livox mid 360 测试 FAST_LIO09RAG 实践- Ollama+RagFlow 部署本地知识库10红米手机使用google play