[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 크롬 브라우저 웹 주소를 구하거나 윈도우즈 메모장 실행하기

■ AgentExecutor 클래스를 사용해 크롬 브라우저 웹 주소를 구하거나 윈도우즈 메모장을 실행하는 방법을 보여준다.

※ Ollama 설치와 benedict/linkbricks-llama3.1-korean:8b 모델이 다운로드 되어 있어야 한다.

▶ main.py


import win32gui
import win32process
import psutil
import requests
import subprocess

from pywinauto             import Application
from langchain.chat_models import ChatOllama
from langchain.memory      import ConversationBufferMemory
from langchain.tools       import Tool
from langchain.agents      import AgentType
from langchain.agents      import initialize_agent

def findMainChromeBrowserWindowHandle():
    chromeBrowserList = []
    def enumerateWindowsCallback(windowHandle, _):
        if win32gui.IsWindowVisible(windowHandle) and win32gui.GetWindowText(windowHandle):
            _, processID = win32process.GetWindowThreadProcessId(windowHandle)
            try:
                process = psutil.Process(processID)
                if process.name().lower() == "chrome.exe":
                    chromeBrowserList.append((windowHandle, process))
            except psutil.NoSuchProcess:
                pass
        return True
    win32gui.EnumWindows(enumerateWindowsCallback, None)
    for windowHandle, process in chromeBrowserList:
        if not any("--type=" in arg for arg in process.cmdline()):
            return windowHandle
    return None

def getChromeBrowserAddressBarText():
    try:
        mainChromeBrowserWindowHandle = findMainChromeBrowserWindowHandle()
        if mainChromeBrowserWindowHandle is None:
            print("No main Chrome window found.")
            return None
        application = Application(backend="uia").connect(handle = mainChromeBrowserWindowHandle)
        window = application.window(handle = mainChromeBrowserWindowHandle)
        for className in ["OmniboxViewViews", "EditControl", "Edit"]:
            try:
                addressEdit = window.child_window(class_name=className)
                if addressEdit.exists():
                    return addressEdit.legacy_properties()["Value"]
            except Exception:
                continue
        print("Address bar element not found.")
        return None
    except Exception as exception:
        print(f"Error in getChromeAddressBarText : {str(exception)}")
        return None

def addProtocolToURL(url):
    if url.startswith("http://") or url.startswith("https://"):
        return url
    httpsURL = f"https://{url}"
    try:
        response = requests.head(httpsURL, allow_redirects = True, timeout = 5)
        if response.status_code == 200:
            return httpsURL
    except requests.RequestException:
        pass
    httpURL = f"http://{url}"
    try:
        response = requests.head(httpURL, allow_redirects = True, timeout = 5)
        if response.status_code == 200:
            return httpURL
    except requests.RequestException:
        pass
    return httpsURL

def getChromeBrowserActiveTabURL(*argumentTuple, **keywordDictionary):
    addressBarText = getChromeBrowserAddressBarText()
    if addressBarText:
        fullURL = addProtocolToURL(addressBarText)
        return fullURL
    return "크롬 브라우저의 URL을 가져오지 못했습니다."

def runNotepad(*arugmentTuple, **keywordDictionary):
    try:
        subprocess.Popen("notepad.exe")
        return "메모장이 성공적으로 실행되었습니다."
    except Exception as e:
        return f"메모장 실행 중 오류가 발생했습니다 : {str(e)}"

chatOllama = ChatOllama(model = "benedict/linkbricks-llama3.1-korean:8b")

conversationBufferMemory = ConversationBufferMemory(memory_key = "chat_history", return_messages = True)

chromeURLTool = Tool(
    name        = "ChromeURLTool",
    func        = getChromeBrowserActiveTabURL,
    description = "현재 크롬 브라우저의 활성 탭 URL을 가져옵니다."
)

runNotepadTool = Tool(
    name        = "NotepadTool",
    func        = runNotepad,
    description = "윈도우즈 메모장을 실행합니다."
)

agentExecutor = initialize_agent(
    agent   = AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,
    llm     = chatOllama,
    memory  = conversationBufferMemory,
    tools   = [chromeURLTool,runNotepadTool],
    verbose = False
)

print("채팅을 시작합니다. '종료'를 입력하면 프로그램이 종료됩니다.")

while True:
    user_input = input("사용자 : ")
    if user_input.lower() == '종료':
        print("채팅을 종료합니다.")
        break
    response = agentExecutor.invoke({"input" : user_input})
    print(f"AI : {response['output']}")

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

import win32gui

import win32process

import psutil

import requests

import subprocess

from pywinauto import Application

from langchain.chat_models import ChatOllama

from langchain.memory import ConversationBufferMemory

from langchain.tools import Tool

from langchain.agents import AgentType

from langchain.agents import initialize_agent

def findMainChromeBrowserWindowHandle():

chromeBrowserList = []

def enumerateWindowsCallback(windowHandle, _):

if win32gui.IsWindowVisible(windowHandle) and win32gui.GetWindowText(windowHandle):

_, processID = win32process.GetWindowThreadProcessId(windowHandle)

try:

process = psutil.Process(processID)

if process.name().lower() == "chrome.exe":

chromeBrowserList.append((windowHandle, process))

except psutil.NoSuchProcess:

pass

return True

win32gui.EnumWindows(enumerateWindowsCallback, None)

for windowHandle, process in chromeBrowserList:

if not any("--type=" in arg for arg in process.cmdline()):

return windowHandle

return None

def getChromeBrowserAddressBarText():

try:

mainChromeBrowserWindowHandle = findMainChromeBrowserWindowHandle()

if mainChromeBrowserWindowHandle is None:

print("No main Chrome window found.")

return None

application = Application(backend="uia").connect(handle = mainChromeBrowserWindowHandle)

window = application.window(handle = mainChromeBrowserWindowHandle)

for className in ["OmniboxViewViews", "EditControl", "Edit"]:

try:

addressEdit = window.child_window(class_name=className)

if addressEdit.exists():

return addressEdit.legacy_properties()["Value"]

except Exception:

continue

print("Address bar element not found.")

return None

except Exception as exception:

print(f"Error in getChromeAddressBarText : {str(exception)}")

return None

def addProtocolToURL(url):

if url.startswith("http://") or url.startswith("https://"):

return url

httpsURL = f"https://{url}"

try:

response = requests.head(httpsURL, allow_redirects = True, timeout = 5)

if response.status_code == 200:

return httpsURL

except requests.RequestException:

pass

httpURL = f"http://{url}"

try:

response = requests.head(httpURL, allow_redirects = True, timeout = 5)

if response.status_code == 200:

return httpURL

except requests.RequestException:

pass

return httpsURL

def getChromeBrowserActiveTabURL(*argumentTuple, **keywordDictionary):

addressBarText = getChromeBrowserAddressBarText()

if addressBarText:

fullURL = addProtocolToURL(addressBarText)

return fullURL

return "크롬 브라우저의 URL을 가져오지 못했습니다."

def runNotepad(*arugmentTuple, **keywordDictionary):

try:

subprocess.Popen("notepad.exe")

return "메모장이 성공적으로 실행되었습니다."

except Exception as e:

return f"메모장 실행 중 오류가 발생했습니다 : {str(e)}"

chatOllama = ChatOllama(model = "benedict/linkbricks-llama3.1-korean:8b")

conversationBufferMemory = ConversationBufferMemory(memory_key = "chat_history", return_messages = True)

chromeURLTool = Tool(

name = "ChromeURLTool",

func = getChromeBrowserActiveTabURL,

description = "현재 크롬 브라우저의 활성 탭 URL을 가져옵니다."

)

runNotepadTool = Tool(

name = "NotepadTool",

func = runNotepad,

description = "윈도우즈 메모장을 실행합니다."

)

agentExecutor = initialize_agent(

agent = AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,

llm = chatOllama,

memory = conversationBufferMemory,

tools = [chromeURLTool,runNotepadTool],

verbose = False

)

print("채팅을 시작합니다. '종료'를 입력하면 프로그램이 종료됩니다.")

while True:

user_input = input("사용자 : ")

if user_input.lower() == '종료':

print("채팅을 종료합니다.")

break

response = agentExecutor.invoke({"input" : user_input})

print(f"AI : {response['output']}")

▶ requirements.txt


aiohappyeyeballs==2.4.0
aiohttp==3.10.5
aiosignal==1.3.1
annotated-types==0.7.0
anyio==4.4.0
attrs==24.2.0
certifi==2024.8.30
charset-normalizer==3.3.2
comtypes==1.4.6
dataclasses-json==0.6.7
frozenlist==1.4.1
greenlet==3.0.3
h11==0.14.0
httpcore==1.0.5
httpx==0.27.2
idna==3.8
jsonpatch==1.33
jsonpointer==3.0.0
langchain==0.2.15
langchain-community==0.2.15
langchain-core==0.2.37
langchain-text-splitters==0.2.2
langsmith==0.1.108
marshmallow==3.22.0
multidict==6.0.5
mypy-extensions==1.0.0
numpy==1.26.4
orjson==3.10.7
packaging==24.1
psutil==6.0.0
pydantic==2.8.2
pydantic_core==2.20.1
pywin32==306
pywinauto==0.6.8
PyYAML==6.0.2
requests==2.32.3
six==1.16.0
sniffio==1.3.1
SQLAlchemy==2.0.32
tenacity==8.5.0
typing-inspect==0.9.0
typing_extensions==4.12.2
urllib3==2.2.2
yarl==1.9.7

aiohappyeyeballs==2.4.0

aiohttp==3.10.5

aiosignal==1.3.1

annotated-types==0.7.0

anyio==4.4.0

attrs==24.2.0

certifi==2024.8.30

charset-normalizer==3.3.2

comtypes==1.4.6

dataclasses-json==0.6.7

frozenlist==1.4.1

greenlet==3.0.3

h11==0.14.0

httpcore==1.0.5

httpx==0.27.2

idna==3.8

jsonpatch==1.33

jsonpointer==3.0.0

langchain==0.2.15

langchain-community==0.2.15

langchain-core==0.2.37

langchain-text-splitters==0.2.2

langsmith==0.1.108

marshmallow==3.22.0

multidict==6.0.5

mypy-extensions==1.0.0

numpy==1.26.4

orjson==3.10.7

packaging==24.1

psutil==6.0.0

pydantic==2.8.2

pydantic_core==2.20.1

pywin32==306

pywinauto==0.6.8

PyYAML==6.0.2

requests==2.32.3

six==1.16.0

sniffio==1.3.1

SQLAlchemy==2.0.32

tenacity==8.5.0

typing-inspect==0.9.0

typing_extensions==4.12.2

urllib3==2.2.2

yarl==1.9.7

※ pip install langchain langchain-community pywin32 psutil pywinauto requests 명령을 실행했다.