OPENAI Archives - 15 중 3 번째 페이지

[PYTHON/LANGCHAIN] StructuredTool 클래스 : invoke 메소드에서 AIMessage 객체의 tool_calls 속성값을 사용해 ToolMessage 객체 구하기

■ StructuredTool 클래스의 invoke 메소드에서 AIMessage 객체의 tool_calls 속성 값을 사용해 ToolMessage 객체를 구하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env

[PYTHON/LANGCHAIN] RunnableBindable 클래스 : invoke 메소드를 사용해 모델에서 도구 호출하기

■ RunnableBindable 클래스의 invoke 메소드를 사용해 모델에서 도구를 호출하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


import random

from dotenv               import load_dotenv
from langchain_core.tools import tool
from typing               import Tuple
from typing               import List
from langchain_openai     import ChatOpenAI

load_dotenv()

@tool(response_format = "content_and_artifact")
def generateRandomIntegerValues(minimumValue : int, maximumValue : int, count : int) -> Tuple[str, List[int]]:
    """Generate size random ints in the range [minimumValue, maximumValue]."""
    array = [random.randint(minimumValue, maximumValue) for _ in range(count)]
    content = f"Successfully generated array of {count} random ints in [{minimumValue}, {maximumValue}]."
    return content, array

chatOpenAI = ChatOpenAI(model = "gpt-4o-mini")

runnableBindable = chatOpenAI.bind_tools([generateRandomIntegerValues])

responseAIMessage = runnableBindable.invoke("generate 6 positive ints less than 25")

print(responseAIMessage.tool_calls)

"""
[{'name': 'generateRandomIntegerValues', 'args': {'minimumValue': 1, 'maximumValue': 24, 'count': 6}, 'id': 'call_F9ZZFBIsd6HXz30bb0yZTWgl', 'type': 'tool_call'}]
"""

import random

from dotenv import load_dotenv

from langchain_core.tools import tool

from typing import Tuple

from typing import List

from langchain_openai import ChatOpenAI

load_dotenv()

@tool(response_format = "content_and_artifact")

def generateRandomIntegerValues(minimumValue : int, maximumValue : int, count : int) -> Tuple[str, List[int]]:

"""Generate size random ints in the range [minimumValue, maximumValue]."""

array = [random.randint(minimumValue, maximumValue) for _ in range(count)]

content = f"Successfully generated array of {count} random ints in [{minimumValue}, {maximumValue}]."

return content, array

chatOpenAI = ChatOpenAI(model = "gpt-4o-mini")

runnableBindable = chatOpenAI.bind_tools([generateRandomIntegerValues])

responseAIMessage = runnableBindable.invoke("generate 6 positive ints less than 25")

print(responseAIMessage.tool_calls)

"""

[{'name': 'generateRandomIntegerValues', 'args': {'minimumValue': 1, 'maximumValue': 24, 'count': 6}, 'id': 'call_F9ZZFBIsd6HXz30bb0yZTWgl', 'type': 'tool_call'}]

"""

[PYTHON/LANGCHAIN] ChatOpenAI 클래스 : bind_tools 메소드에서 parallel_tool_calls 인자를 사용해 병렬 도구 호출 비활성화하기

■ ChatOpenAI 클래스의 bind_tools 메소드에서 parallel_tool_calls 인자를 사용해 병렬 도구 호출을 비활성화하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다.

[PYTHON/LANGCHAIN] chatOpenAI 클래스 : bind_tools 메소드에서 tool_choice 인자를 사용해 도구 호출 강제하기

■ chatOpenAI 클래스의 bind_tools 메소드에서 tool_choice 인자를 사용해 도구 호출을 강제하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] 도구 실행 오류시 재실행하기

■ 도구 실행 오류시 재실행하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                   import load_dotenv
from langchain_core.tools     import tool
from langchain_core.messages  import ToolCall
from langchain_core.messages  import AIMessage
from langchain_core.runnables import RunnableConfig
from langchain_core.runnables import Runnable
from langchain_core.messages  import ToolMessage
from langchain_core.messages  import HumanMessage
from langchain_core.prompts   import ChatPromptTemplate
from langchain_openai         import ChatOpenAI

load_dotenv()

@tool
def complexTool(integerValue : int, floatValue : float, dictionaryValue : dict) -> int:
    """Do something complex with a complex tool."""
    return integerValue * floatValue

class CustomToolException(Exception):
    """Custom LangChain tool exception."""

    def __init__(self, toolCall : ToolCall, exception : Exception) -> None:
        super().__init__()
        self.tool_call = toolCall
        self.exception = exception

def complexToolWithCustomException(aiMessage : AIMessage, config : RunnableConfig) -> Runnable:
    try:
        return complexTool.invoke(aiMessage.tool_calls[0]["args"], config = config)
    except Exception as exception:
        raise CustomToolException(aiMessage.tool_calls[0], exception)

def getInputDictionary(inputDictionary : dict) -> dict:
    exception = inputDictionary.pop("exception")

    messageList = [
        AIMessage(content = "", tool_calls = [exception.tool_call]),
        ToolMessage(tool_call_id = exception.tool_call["id"], content = str(exception.exception)),
        HumanMessage(content = "The last tool call raised an exception. Try calling the tool again with corrected arguments. Do not repeat mistakes.")
    ]

    inputDictionary["last_output"] = messageList

    return inputDictionary

# 우리는 프롬프트에 last_output MessagesPlaceholder를 추가한다.
# 이 메시지는 전달되지 않으면 프롬프트에 전혀 영향을 미치지 않지만 필요한 경우 임의의 메시지 목록을 프롬프트에 삽입할 수 있는 옵션을 제공한다.
# 이를 재시도 시 사용하여 오류 메시지를 삽입한다.
chatPromptTemplate = ChatPromptTemplate.from_messages(
    [
        ("human"      , "{input}"      ),
        ("placeholder", "{last_output}")
    ]
)

chatOpenAI = ChatOpenAI(model="gpt-4o-mini")

runnableBindable = chatOpenAI.bind_tools([complexTool])

runnableSequence = chatPromptTemplate | runnableBindable | complexToolWithCustomException

selfCorrectingRunnableSequence = runnableSequence.with_fallbacks([getInputDictionary | runnableSequence], exception_key = "exception")

responseValue = selfCorrectingRunnableSequence.invoke({"input" : "use complex tool. the args are 5, 2.1, empty dictionary. don't forget dictionaryValue"})

print(responseValue)

from dotenv import load_dotenv

from langchain_core.tools import tool

from langchain_core.messages import ToolCall

from langchain_core.messages import AIMessage

from langchain_core.runnables import RunnableConfig

from langchain_core.runnables import Runnable

from langchain_core.messages import ToolMessage

from langchain_core.messages import HumanMessage

from langchain_core.prompts import ChatPromptTemplate

from langchain_openai import ChatOpenAI

load_dotenv()

@tool

def complexTool(integerValue : int, floatValue : float, dictionaryValue : dict) -> int:

"""Do something complex with a complex tool."""

return integerValue * floatValue

class CustomToolException(Exception):

"""Custom LangChain tool exception."""

def __init__(self, toolCall : ToolCall, exception : Exception) -> None:

super().__init__()

self.tool_call = toolCall

self.exception = exception

def complexToolWithCustomException(aiMessage : AIMessage, config : RunnableConfig) -> Runnable:

try:

return complexTool.invoke(aiMessage.tool_calls[0]["args"], config = config)

except Exception as exception:

raise CustomToolException(aiMessage.tool_calls[0], exception)

def getInputDictionary(inputDictionary : dict) -> dict:

exception = inputDictionary.pop("exception")

messageList = [

AIMessage(content = "", tool_calls = [exception.tool_call]),

ToolMessage(tool_call_id = exception.tool_call["id"], content = str(exception.exception)),

HumanMessage(content = "The last tool call raised an exception. Try calling the tool again with corrected arguments. Do not repeat mistakes.")

]

inputDictionary["last_output"] = messageList

return inputDictionary

# 우리는 프롬프트에 last_output MessagesPlaceholder를 추가한다.

# 이 메시지는 전달되지 않으면 프롬프트에 전혀 영향을 미치지 않지만 필요한 경우 임의의 메시지 목록을 프롬프트에 삽입할 수 있는 옵션을 제공한다.

# 이를 재시도 시 사용하여 오류 메시지를 삽입한다.

chatPromptTemplate = ChatPromptTemplate.from_messages(

[

("human" , "{input}" ),

("placeholder", "{last_output}")

]

)

chatOpenAI = ChatOpenAI(model="gpt-4o-mini")

runnableBindable = chatOpenAI.bind_tools([complexTool])

runnableSequence = chatPromptTemplate | runnableBindable | complexToolWithCustomException

selfCorrectingRunnableSequence = runnableSequence.with_fallbacks([getInputDictionary | runnableSequence], exception_key = "exception")

responseValue = selfCorrectingRunnableSequence.invoke({"input" : "use complex tool. the args are 5, 2.1, empty dictionary. don't forget dictionaryValue"})

print(responseValue)

▶ requirements.txt


annotated-types==0.7.0
anyio==4.6.0
certifi==2024.8.30
charset-normalizer==3.3.2
colorama==0.4.6
distro==1.9.0
h11==0.14.0
httpcore==1.0.5
httpx==0.27.2
idna==3.10
jiter==0.5.0
jsonpatch==1.33
jsonpointer==3.0.0
langchain-core==0.3.5
langchain-openai==0.2.0
langsmith==0.1.128
openai==1.47.1
orjson==3.10.7
packaging==24.1
pydantic==2.9.2
pydantic_core==2.23.4
python-dotenv==1.0.1
PyYAML==6.0.2
regex==2024.9.11
requests==2.32.3
sniffio==1.3.1
tenacity==8.5.0
tiktoken==0.7.0
tqdm==4.66.5
typing_extensions==4.12.2
urllib3==2.2.3

annotated-types==0.7.0

anyio==4.6.0

certifi==2024.8.30

charset-normalizer==3.3.2

colorama==0.4.6

distro==1.9.0

h11==0.14.0

httpcore==1.0.5

httpx==0.27.2

idna==3.10

jiter==0.5.0

jsonpatch==1.33

jsonpointer==3.0.0

langchain-core==0.3.5

langchain-openai==0.2.0

langsmith==0.1.128

openai==1.47.1

orjson==3.10.7

packaging==24.1

pydantic==2.9.2

pydantic_core==2.23.4

python-dotenv==1.0.1

PyYAML==6.0.2

regex==2024.9.11

requests==2.32.3

sniffio==1.3.1

tenacity==8.5.0

tiktoken==0.7.0

tqdm==4.66.5

typing_extensions==4.12.2

urllib3==2.2.3

※

[PYTHON/LANGCHAIN] RunnableSequence 클래스 : with_fallbacks 메소드를 사용해 도구 오류시 대체 모델 사용하기

■ RunnableSequence 클래스의 with_fallbacks 메소드를 사용해 도구 오류시 대체 모델을 사용하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] try … except 구문 : 도구 사용시 오류 처리하기

■ try … except 구문을 사용해 도구 사용시 오류를 처리하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py

[PYTHON/LANGCHAIN] RunnableSequence 클래스 : batch 메소드에서 config 인자를 사용해 동시 작업 수 설정하기

■ RunnableSequence 클래스의 batch 메소드에서 config 인자를 사용해 동시 작업 수를 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다.

[PYTHON/LANGCHAIN] PromptTemplate 클래스 : from_template 정적 메소드를 사용해 PromptTemplate 객체 만들기 4

■ PromptTemplate 클래스의 from_template 정적 메소드를 사용해 PromptTemplate 객체를 만드는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py

[PYTHON/LANGCHAIN] RunnableSequence 클래스 : invoke 메소드를 사용해 LCEL 체인 실행하기

■ RunnableSequence 클래스의 invoke 메소드를 사용해 LCEL 체인을 실행하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                 import load_dotenv
from langchain_core.prompts import PromptTemplate
from langchain_openai       import ChatOpenAI

load_dotenv()

templateString = "{country}의 수도는 어디입니까?"

promptTemplate = PromptTemplate.from_template(templateString)

chatOpenAI = ChatOpenAI(
    model_name  = "gpt-4o-mini", # 모델명
    max_tokens  = 2048,          # 최대 토큰 수
    temperature = 0.1,           # 창의성 (0.0 ~ 2.0)
)

runnableSequence = promptTemplate | chatOpenAI

responseAIMessage = runnableSequence.invoke({"country" : "미국"})

print(responseAIMessage.content)

"""
미국의 수도는 워싱턴 D.C.입니다.
"""

from dotenv import load_dotenv

from langchain_core.prompts import PromptTemplate

from langchain_openai import ChatOpenAI

load_dotenv()

templateString = "{country}의 수도는 어디입니까?"

promptTemplate = PromptTemplate.from_template(templateString)

chatOpenAI = ChatOpenAI(

model_name = "gpt-4o-mini", # 모델명

max_tokens = 2048, # 최대 토큰 수

temperature = 0.1, # 창의성 (0.0 ~ 2.0)

)

runnableSequence = promptTemplate | chatOpenAI

responseAIMessage = runnableSequence.invoke({"country" : "미국"})

print(responseAIMessage.content)

"""

미국의 수도는 워싱턴 D.C.입니다.

"""

[PYTHON/LANGCHAIN] ChatOpenAI 클래스 : bind 메소드에서 logprobs 인자를 사용해 주어진 토큰 확률 로그 값 받기

■ ChatOpenAI 클래스의 bind 메소드에서 logprobs 인자를 사용해 주어진 토큰 확률 로그 값을 받는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env

[PYTHON/LANGCHAIN] ChatOpenAI 클래스 : invoke 메소드를 사용해 질문하기

■ ChatOpenAI 클래스의 invoke 메소드를 사용해 질문하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv           import load_dotenv
from langchain_openai import ChatOpenAI

load_dotenv()

chatOpenAI = ChatOpenAI(
    model_name  = "gpt-4o-mini", # 모델명
    temperature = 0.1            # 창의성 (0.0 ~ 2.0)
)

requestString = "대한민국의 수도는 어디입니까?"

responseAIMessage = chatOpenAI.invoke(requestString)

print(responseAIMessage.content)

from dotenv import load_dotenv

from langchain_openai import ChatOpenAI

load_dotenv()

chatOpenAI = ChatOpenAI(

model_name = "gpt-4o-mini", # 모델명

temperature = 0.1 # 창의성 (0.0 ~ 2.0)

)

requestString = "대한민국의 수도는 어디입니까?"

responseAIMessage = chatOpenAI.invoke(requestString)

print(responseAIMessage.content)

▶ requirements.txt

[PYTHON/LANGCHAIN] 도구 호출시 오류 처리하기

■ 도구 호출시 오류를 처리하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                   import load_dotenv
from langchain_core.tools     import tool
from langchain_core.runnables import RunnableConfig
from langchain_core.runnables import Runnable
from langchain_openai         import ChatOpenAI

load_dotenv()

@tool
def _execute(integerValue : int, floatValue : float, dictionary1 : dict) -> int:
    """Do something complex with a complex tool."""
    return integerValue * floatValue

def execute(tool_args : dict, config : RunnableConfig) -> Runnable:
    try:
        _execute.invoke(tool_args, config = config)
    except Exception as e:
        return f"Calling tool with arguments:\n\n{tool_args}\n\nraised the following error:\n\n{type(e)}: {e}"

chatOpenAI = ChatOpenAI(model = "gpt-4o-mini")

runnableBindable = chatOpenAI.bind_tools([_execute])

runnableSequence = runnableBindable | (lambda aiMessage : aiMessage.tool_calls[0]["args"]) | execute

resultValue = runnableSequence.invoke("use complex tool. the args are 5, 2.1, empty dictionary. don't forget dictionary1")

print(resultValue)

"""
Calling tool with arguments:

{'integerValue': 5, 'floatValue': 2.1}

raised the following error:

<class 'pydantic_core._pydantic_core.ValidationError'>: 1 validation error for _execute
dictionary1
  Field required [type=missing, input_value={'integerValue': 5, 'floatValue': 2.1}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.9/v/missing
"""

from dotenv import load_dotenv

from langchain_core.tools import tool

from langchain_core.runnables import RunnableConfig

from langchain_core.runnables import Runnable

from langchain_openai import ChatOpenAI

load_dotenv()

@tool

def _execute(integerValue : int, floatValue : float, dictionary1 : dict) -> int:

"""Do something complex with a complex tool."""

return integerValue * floatValue

def execute(tool_args : dict, config : RunnableConfig) -> Runnable:

try:

_execute.invoke(tool_args, config = config)

except Exception as e:

return f"Calling tool with arguments:\n\n{tool_args}\n\nraised the following error:\n\n{type(e)}: {e}"

chatOpenAI = ChatOpenAI(model = "gpt-4o-mini")

runnableBindable = chatOpenAI.bind_tools([_execute])

runnableSequence = runnableBindable | (lambda aiMessage : aiMessage.tool_calls[0]["args"]) | execute

resultValue = runnableSequence.invoke("use complex tool. the args are 5, 2.1, empty dictionary. don't forget dictionary1")

print(resultValue)

"""

Calling tool with arguments:

{'integerValue': 5, 'floatValue': 2.1}

raised the following error:

<class 'pydantic_core._pydantic_core.ValidationError'>: 1 validation error for _execute

dictionary1

Field required [type=missing, input_value={'integerValue': 5, 'floatValue': 2.1}, input_type=dict]

For further information visit https://errors.pydantic.dev/2.9/v/missing

"""

▶ requirements.txt


annotated-types==0.7.0
anyio==4.6.0
certifi==2024.8.30
charset-normalizer==3.3.2
colorama==0.4.6
distro==1.9.0
h11==0.14.0
httpcore==1.0.5
httpx==0.27.2
idna==3.10
jiter==0.5.0
jsonpatch==1.33
jsonpointer==3.0.0
langchain-core==0.3.5
langchain-openai==0.2.0
langsmith==0.1.125
openai==1.47.0
orjson==3.10.7
packaging==24.1
pydantic==2.9.2
pydantic_core==2.23.4
python-dotenv==1.0.1
PyYAML==6.0.2
regex==2024.9.11
requests==2.32.3
sniffio==1.3.1
tenacity==8.5.0
tiktoken==0.7.0
tqdm==4.66.5
typing_extensions==4.12.2
urllib3==2.2.3

annotated-types==0.7.0

anyio==4.6.0

certifi==2024.8.30

charset-normalizer==3.3.2

colorama==0.4.6

distro==1.9.0

h11==0.14.0

httpcore==1.0.5

httpx==0.27.2

idna==3.10

jiter==0.5.0

jsonpatch==1.33

jsonpointer==3.0.0

langchain-core==0.3.5

langchain-openai==0.2.0

langsmith==0.1.125

openai==1.47.0

orjson==3.10.7

packaging==24.1

pydantic==2.9.2

pydantic_core==2.23.4

python-dotenv==1.0.1

PyYAML==6.0.2

regex==2024.9.11

requests==2.32.3

sniffio==1.3.1

tenacity==8.5.0

tiktoken==0.7.0

tqdm==4.66.5

typing_extensions==4.12.2

urllib3==2.2.3

※

[PYTHON/LANGCHAIN] 도구 호출시 사용자 승인 절차 추가하기

■ 도구 호출시 사용자 승인 절차를 추가하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


import json

from dotenv                   import load_dotenv
from langchain_core.tools     import tool
from langchain_openai         import ChatOpenAI
from langchain_core.messages  import AIMessage
from typing                   import Dict
from typing                   import List

load_dotenv()

@tool
def countEmail(lastDayCount : int) -> int:
    """Dummy function to count number of e-mails. Returns 2 * last_n_days."""
    return lastDayCount * 2

@tool
def sendEmail(message : str, recipient : str) -> str:
    """Dummy function for sending an e-mail."""
    return f"Successfully sent email to {recipient}."

toolList = [countEmail, sendEmail]

chatOpenAI = ChatOpenAI(model = "gpt-4o-mini")

runnableBinding = chatOpenAI.bind_tools(toolList)

def callTool(aiMessage : AIMessage) -> List[Dict]:
    """Simple sequential tool calling helper."""
    toolMapDictionary = {tool.name : tool for tool in toolList}
    toolCallList = aiMessage.tool_calls.copy()
    for toolCall in toolCallList:
        toolCall["output"] = toolMapDictionary[toolCall["name"]].invoke(toolCall["args"])
    return toolCallList

class NotApproved(Exception):
    """Custom exception."""

def approve(aiMessage : AIMessage) -> AIMessage:
    """Responsible for passing through its input or raising an exception.

    Args:
        aiMessage : output from the chat model

    Returns:
        aiMessage : original output from the msg
    """

    toolCallListJSON = "\n\n".join(json.dumps(toolCall, indent = 2) for toolCall in aiMessage.tool_calls)
    inputString      = f"Do you approve of the following tool invocations\n\n{toolCallListJSON}\n\nAnything except 'Y'/'Yes' (case-insensitive) will be treated as a no.\n >>>"

    outputString = input(inputString)

    if outputString.lower() not in ("yes", "y"):
        raise NotApproved(f"Tool invocations not approved:\n\n{toolCallListJSON}")
    return aiMessage

runnableSequence = runnableBinding | approve | callTool

try:
    toolCallList = runnableSequence.invoke("how many emails did i get in the last 5 days?")

    print(toolCallList)
except NotApproved as e:
    print()
    print(e)

import json

from dotenv import load_dotenv

from langchain_core.tools import tool

from langchain_openai import ChatOpenAI

from langchain_core.messages import AIMessage

from typing import Dict

from typing import List

load_dotenv()

@tool

def countEmail(lastDayCount : int) -> int:

"""Dummy function to count number of e-mails. Returns 2 * last_n_days."""

return lastDayCount * 2

@tool

def sendEmail(message : str, recipient : str) -> str:

"""Dummy function for sending an e-mail."""

return f"Successfully sent email to {recipient}."

toolList = [countEmail, sendEmail]

chatOpenAI = ChatOpenAI(model = "gpt-4o-mini")

runnableBinding = chatOpenAI.bind_tools(toolList)

def callTool(aiMessage : AIMessage) -> List[Dict]:

"""Simple sequential tool calling helper."""

toolMapDictionary = {tool.name : tool for tool in toolList}

toolCallList = aiMessage.tool_calls.copy()

for toolCall in toolCallList:

toolCall["output"] = toolMapDictionary[toolCall["name"]].invoke(toolCall["args"])

return toolCallList

class NotApproved(Exception):

"""Custom exception."""

def approve(aiMessage : AIMessage) -> AIMessage:

"""Responsible for passing through its input or raising an exception.

Args:

aiMessage : output from the chat model

Returns:

aiMessage : original output from the msg

"""

toolCallListJSON = "\n\n".join(json.dumps(toolCall, indent = 2) for toolCall in aiMessage.tool_calls)

inputString = f"Do you approve of the following tool invocations\n\n{toolCallListJSON}\n\nAnything except 'Y'/'Yes' (case-insensitive) will be treated as a no.\n >>>"

outputString = input(inputString)

if outputString.lower() not in ("yes", "y"):

raise NotApproved(f"Tool invocations not approved:\n\n{toolCallListJSON}")

return aiMessage

runnableSequence = runnableBinding | approve | callTool

try:

toolCallList = runnableSequence.invoke("how many emails did i get in the last 5 days?")

print(toolCallList)

except NotApproved as e:

print()

print(e)

▶ requirements.txt

[PYTHON/LANGCHAIN] InjectedToolArg 클래스 : 도구에 런타임 값 전달하기

■ InjectedToolArg 클래스를 사용해 도구에 런타임 값을 전달하는 방법을 보여준다. ※ 런타임에만 알려진 도구에 값을 바인딩해야 할 수도 있다. ※ 예를 들어,

[PYTHON/LANGCHAIN] AIMessageChunk 클래스 : 스트림으로 수신되는 AIMessageChunk 객체에서 도구 호출 리스트 구하기

■ AIMessageChunk 클래스 관련해 스트림으로 수신되는 AIMessageChunk 객체에서 도구 호출 리스트를 구하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다.

[PYTHON/LANGCHAIN] ChatOpenAI 클래스 : bind_tools 메소드를 사용해 도구 출력을 채팅 모델에 전달하기

■ ChatOpenAI 클래스의 bind_tools 메소드를 사용해 도구 출력을 채팅 모델에 전달하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] PydanticToolsParser 클래스 : 생성자에서 tools 인자를 사용해 도구 호출 정보 반환하기

■ PydanticToolsParser 클래스의 생성자에서 tools 인자를 사용해 도구 호출 정보를 반환하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] ChatOpenAI 클래스 : bind_tools 메소드를 사용해 도구 호출 메시지 반환하기

■ ChatOpenAI 클래스dml bind_tools 메소드를 사용해 도구 호출 메시지를 반환하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py

[PYTHON/LANGCHAIN] TimeWeightedVectorStoreRetriever 클래스 : mock_now 함수를 사용해 가상 시간 설정하기

■ TimeWeightedVectorStoreRetriever 클래스에서 mock_now 함수를 사용해 가상 시간을 설정하는 방법을 보여준다. ▶ main.py


import faiss

from langchain_openai                 import OpenAIEmbeddings
from langchain_community.docstore     import InMemoryDocstore
from langchain_community.vectorstores import FAISS
from langchain.retrievers             import TimeWeightedVectorStoreRetriever
from datetime                         import datetime
from datetime                         import timedelta
from langchain_core.documents         import Document
from langchain_core.utils             import mock_now

openAIEmbeddings = OpenAIEmbeddings()

embeddingSize = 1536

indexFlatL2 = faiss.IndexFlatL2(embeddingSize)

inMemoryDocstore = InMemoryDocstore({})

faiss = FAISS(openAIEmbeddings, indexFlatL2, inMemoryDocstore, {})

timeWeightedVectorStoreRetriever = TimeWeightedVectorStoreRetriever(vectorstore = faiss, decay_rate = 0.999, k = 1)

yesterday = datetime.now() - timedelta(days = 1)

timeWeightedVectorStoreRetriever.add_documents([Document(page_content = "hello world", metadata = {"last_accessed_at" : yesterday})])
timeWeightedVectorStoreRetriever.add_documents([Document(page_content = "hello foo")])

with mock_now(datetime(2024, 9, 16, 10, 11)):
    resultDocumentList = timeWeightedVectorStoreRetriever.get_relevant_documents("hello world")
    for resultDocument in resultDocumentList:
        print(resultDocument)

import faiss

from langchain_openai import OpenAIEmbeddings

from langchain_community.docstore import InMemoryDocstore

from langchain_community.vectorstores import FAISS

from langchain.retrievers import TimeWeightedVectorStoreRetriever

from datetime import datetime

from datetime import timedelta

from langchain_core.documents import Document

from langchain_core.utils import mock_now

openAIEmbeddings = OpenAIEmbeddings()

embeddingSize = 1536

indexFlatL2 = faiss.IndexFlatL2(embeddingSize)

inMemoryDocstore = InMemoryDocstore({})

faiss = FAISS(openAIEmbeddings, indexFlatL2, inMemoryDocstore, {})

timeWeightedVectorStoreRetriever = TimeWeightedVectorStoreRetriever(vectorstore = faiss, decay_rate = 0.999, k = 1)

yesterday = datetime.now() - timedelta(days = 1)

timeWeightedVectorStoreRetriever.add_documents([Document(page_content = "hello world", metadata = {"last_accessed_at" : yesterday})])

timeWeightedVectorStoreRetriever.add_documents([Document(page_content = "hello foo")])

with mock_now(datetime(2024, 9, 16, 10, 11)):

resultDocumentList = timeWeightedVectorStoreRetriever.get_relevant_documents("hello world")

for resultDocument in resultDocumentList:

print(resultDocument)

▶ 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
colorama==0.4.6
dataclasses-json==0.6.7
distro==1.9.0
faiss-cpu==1.8.0.post1
frozenlist==1.4.1
greenlet==3.1.0
h11==0.14.0
httpcore==1.0.5
httpx==0.27.2
idna==3.10
jiter==0.5.0
jsonpatch==1.33
jsonpointer==3.0.0
langchain==0.3.0
langchain-community==0.3.0
langchain-core==0.3.0
langchain-openai==0.2.0
langchain-text-splitters==0.3.0
langsmith==0.1.121
marshmallow==3.22.0
multidict==6.1.0
mypy-extensions==1.0.0
numpy==1.26.4
openai==1.45.1
orjson==3.10.7
packaging==24.1
pydantic==2.9.1
pydantic-settings==2.5.2
pydantic_core==2.23.3
python-dotenv==1.0.1
PyYAML==6.0.2
regex==2024.9.11
requests==2.32.3
sniffio==1.3.1
SQLAlchemy==2.0.35
tenacity==8.5.0
tiktoken==0.7.0
tqdm==4.66.5
typing-inspect==0.9.0
typing_extensions==4.12.2
urllib3==2.2.3
yarl==1.11.1

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

colorama==0.4.6

dataclasses-json==0.6.7

distro==1.9.0

faiss-cpu==1.8.0.post1

frozenlist==1.4.1

greenlet==3.1.0

h11==0.14.0

httpcore==1.0.5

httpx==0.27.2

idna==3.10

jiter==0.5.0

jsonpatch==1.33

jsonpointer==3.0.0

langchain==0.3.0

langchain-community==0.3.0

langchain-core==0.3.0

langchain-openai==0.2.0

langchain-text-splitters==0.3.0

langsmith==0.1.121

marshmallow==3.22.0

multidict==6.1.0

mypy-extensions==1.0.0

numpy==1.26.4

openai==1.45.1

orjson==3.10.7

packaging==24.1

pydantic==2.9.1

pydantic-settings==2.5.2

pydantic_core==2.23.3

python-dotenv==1.0.1

PyYAML==6.0.2

regex==2024.9.11

requests==2.32.3

sniffio==1.3.1

SQLAlchemy==2.0.35

tenacity==8.5.0

tiktoken==0.7.0

tqdm==4.66.5

typing-inspect==0.9.0

typing_extensions==4.12.2

urllib3==2.2.3

yarl==1.11.1

※ pip install langchain-community langchain-openai

[PYTHON/LANGCHAIN] TimeWeightedVectorStoreRetriever 클래스 : 높은 감쇠율로 문서 검색하기

■ TimeWeightedVectorStoreRetriever 클래스를 사용해 높은 감쇠율로 문서를 검색하는 방법을 보여준다. ※ 높은 감소율(예 : 9가 여러 개)에서는 최근성 점수가 빠르게 0으로 떨어진다!

[PYTHON/LANGCHAIN] TimeWeightedVectorStoreRetriever 클래스 : 낮은 감쇠율로 문서 검색하기

■ TimeWeightedVectorStoreRetriever 클래스를 사용해 낮은 감쇠율로 문서를 검색하는 방법을 보여준다. ※ 낮은 감쇠율(여기서는 극단적으로 0에 가깝게 설정했다)은 기억이 더 오랫동안 "기억"된다는 것을

[PYTHON/LANGCHAIN] TimeWeightedVectorStoreRetriever 클래스 : 생성자에서 vectorstore/decay_rate 인자를 사용해 TimeWeightedVectorStoreRetriever 객체 만들기

■ TimeWeightedVectorStoreRetriever 클래스의 생성자에서 vectorstore/decay_rate 인자를 사용해 TimeWeightedVectorStoreRetriever 객체를 만드는 방법을 보여준다. ▶ main.py


import faiss

from langchain_openai                 import OpenAIEmbeddings
from langchain_community.docstore     import InMemoryDocstore
from langchain_community.vectorstores import FAISS
from langchain.retrievers             import TimeWeightedVectorStoreRetriever

openAIEmbeddings = OpenAIEmbeddings()

embeddingSize = 1536

indexFlatL2 = faiss.IndexFlatL2(embeddingSize)

inMemoryDocstore = InMemoryDocstore({})

faiss = FAISS(openAIEmbeddings, indexFlatL2, inMemoryDocstore, {})

timeWeightedVectorStoreRetriever = TimeWeightedVectorStoreRetriever(vectorstore = faiss, decay_rate = 0.0000000000000000000000001, k = 1)

import faiss

from langchain_openai import OpenAIEmbeddings

from langchain_community.docstore import InMemoryDocstore

from langchain_community.vectorstores import FAISS

from langchain.retrievers import TimeWeightedVectorStoreRetriever

openAIEmbeddings = OpenAIEmbeddings()

embeddingSize = 1536

indexFlatL2 = faiss.IndexFlatL2(embeddingSize)

inMemoryDocstore = InMemoryDocstore({})

faiss = FAISS(openAIEmbeddings, indexFlatL2, inMemoryDocstore, {})

timeWeightedVectorStoreRetriever = TimeWeightedVectorStoreRetriever(vectorstore = faiss, decay_rate = 0.0000000000000000000000001, k = 1)

▶ 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
colorama==0.4.6
dataclasses-json==0.6.7
distro==1.9.0
faiss-cpu==1.8.0.post1
frozenlist==1.4.1
greenlet==3.1.0
h11==0.14.0
httpcore==1.0.5
httpx==0.27.2
idna==3.10
jiter==0.5.0
jsonpatch==1.33
jsonpointer==3.0.0
langchain==0.3.0
langchain-community==0.3.0
langchain-core==0.3.0
langchain-openai==0.2.0
langchain-text-splitters==0.3.0
langsmith==0.1.121
marshmallow==3.22.0
multidict==6.1.0
mypy-extensions==1.0.0
numpy==1.26.4
openai==1.45.1
orjson==3.10.7
packaging==24.1
pydantic==2.9.1
pydantic-settings==2.5.2
pydantic_core==2.23.3
python-dotenv==1.0.1
PyYAML==6.0.2
regex==2024.9.11
requests==2.32.3
sniffio==1.3.1
SQLAlchemy==2.0.35
tenacity==8.5.0
tiktoken==0.7.0
tqdm==4.66.5
typing-inspect==0.9.0
typing_extensions==4.12.2
urllib3==2.2.3
yarl==1.11.1

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

colorama==0.4.6

dataclasses-json==0.6.7

distro==1.9.0

faiss-cpu==1.8.0.post1

frozenlist==1.4.1

greenlet==3.1.0

h11==0.14.0

httpcore==1.0.5

httpx==0.27.2

idna==3.10

jiter==0.5.0

jsonpatch==1.33

jsonpointer==3.0.0

langchain==0.3.0

langchain-community==0.3.0

langchain-core==0.3.0

langchain-openai==0.2.0

langchain-text-splitters==0.3.0

langsmith==0.1.121

marshmallow==3.22.0

multidict==6.1.0

mypy-extensions==1.0.0

numpy==1.26.4

openai==1.45.1

orjson==3.10.7

packaging==24.1

pydantic==2.9.1

pydantic-settings==2.5.2

pydantic_core==2.23.3

python-dotenv==1.0.1

PyYAML==6.0.2

regex==2024.9.11

requests==2.32.3

sniffio==1.3.1

SQLAlchemy==2.0.35

tenacity==8.5.0

tiktoken==0.7.0

tqdm==4.66.5

typing-inspect==0.9.0

typing_extensions==4.12.2

urllib3==2.2.3

yarl==1.11.1

※ pip install langchain-community

[PYTHON/LANGCHAIN] SelfQueryRetriever 클래스 : 생성자에서 query_constructor 인자를 사용해 LCEL 설정하기

■ SelfQueryRetriever 클래스의 생성자에서 query_constructor 인자를 사용해 LCEL을 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                                        import load_dotenv
from langchain_core.documents                      import Document
from langchain_openai                              import OpenAIEmbeddings
from langchain_chroma                              import Chroma
from langchain.chains.query_constructor.base       import AttributeInfo
from langchain.chains.query_constructor.base       import get_query_constructor_prompt
from langchain.chains.query_constructor.base       import StructuredQueryOutputParser
from langchain_openai                              import ChatOpenAI
from langchain.retrievers.self_query.base          import SelfQueryRetriever
from langchain_community.query_constructors.chroma import ChromaTranslator

load_dotenv()

documentList = [
    Document(
        page_content = "A bunch of scientists bring back dinosaurs and mayhem breaks loose",
        metadata     = {"year" : 1993, "rating" : 7.7, "genre" : "science fiction"}
    ),
    Document(
        page_content = "Leo DiCaprio gets lost in a dream within a dream within a dream within a ...",
        metadata     = {"year" : 2010, "director" : "Christopher Nolan", "rating" : 8.2}
    ),
    Document(
        page_content = "A psychologist / detective gets lost in a series of dreams within dreams within dreams and Inception reused the idea",
        metadata     = {"year" : 2006, "director" : "Satoshi Kon", "rating" : 8.6}
    ),
    Document(
        page_content = "A bunch of normal-sized women are supremely wholesome and some men pine after them",
        metadata     = {"year" : 2019, "director" : "Greta Gerwig", "rating" : 8.3}
    ),
    Document(
        page_content = "Toys come alive and have a blast doing so",
        metadata     = {"year" : 1995, "genre" : "animated"}
    ),
    Document(
        page_content = "Three men walk into the Zone, three men walk out of the Zone",
        metadata = {"year" : 1979, "director" : "Andrei Tarkovsky", "genre" : "thriller", "rating" : 9.9}
    )
]

openAIEmbeddings = OpenAIEmbeddings()

chroma = Chroma.from_documents(documentList, openAIEmbeddings)

attributeInfoList = [
    AttributeInfo(
        name        = "genre",
        description = "The genre of the movie. One of ['science fiction', 'comedy', 'drama', 'thriller', 'romance', 'action', 'animated']",
        type        = "string"
    ),
    AttributeInfo(
        name        = "year",
        description = "The year the movie was released",
        type        = "integer"
    ),
    AttributeInfo(
        name        = "director",
        description = "The name of the movie director",
        type        = "string"
    ),
    AttributeInfo(
        name        = "rating",
        description = "A 1-10 rating for the movie",
        type        = "float"
    )
]

documentContentDescription = "Brief summary of a movie"

fewShotPromptTemplate = get_query_constructor_prompt(
    documentContentDescription,
    attributeInfoList
)

structuredQueryOutputParser = StructuredQueryOutputParser.from_components()

chatOpenAI = ChatOpenAI(temperature = 0)

runnableSequence = fewShotPromptTemplate | chatOpenAI | structuredQueryOutputParser

selfQueryRetriever = SelfQueryRetriever(
    query_constructor           = runnableSequence,
    vectorstore                 = chroma,
    structured_query_translator = ChromaTranslator(),
)

resultDocumentList = selfQueryRetriever.invoke("What's a movie after 1990 but before 2005 that's all about toys, and preferably is animated")

for resultDocument in resultDocumentList:
    print(resultDocument)

from dotenv import load_dotenv

from langchain_core.documents import Document

from langchain_openai import OpenAIEmbeddings

from langchain_chroma import Chroma

from langchain.chains.query_constructor.base import AttributeInfo

from langchain.chains.query_constructor.base import get_query_constructor_prompt

from langchain.chains.query_constructor.base import StructuredQueryOutputParser

from langchain_openai import ChatOpenAI

from langchain.retrievers.self_query.base import SelfQueryRetriever

from langchain_community.query_constructors.chroma import ChromaTranslator

load_dotenv()

documentList = [

Document(

page_content = "A bunch of scientists bring back dinosaurs and mayhem breaks loose",

metadata = {"year" : 1993, "rating" : 7.7, "genre" : "science fiction"}

Document(

page_content = "Leo DiCaprio gets lost in a dream within a dream within a dream within a ...",

metadata = {"year" : 2010, "director" : "Christopher Nolan", "rating" : 8.2}

Document(

page_content = "A psychologist / detective gets lost in a series of dreams within dreams within dreams and Inception reused the idea",

metadata = {"year" : 2006, "director" : "Satoshi Kon", "rating" : 8.6}

Document(

page_content = "A bunch of normal-sized women are supremely wholesome and some men pine after them",

metadata = {"year" : 2019, "director" : "Greta Gerwig", "rating" : 8.3}

Document(

page_content = "Toys come alive and have a blast doing so",

metadata = {"year" : 1995, "genre" : "animated"}

Document(

page_content = "Three men walk into the Zone, three men walk out of the Zone",

metadata = {"year" : 1979, "director" : "Andrei Tarkovsky", "genre" : "thriller", "rating" : 9.9}

)

]

openAIEmbeddings = OpenAIEmbeddings()

chroma = Chroma.from_documents(documentList, openAIEmbeddings)

attributeInfoList = [

AttributeInfo(

name = "genre",

description = "The genre of the movie. One of ['science fiction', 'comedy', 'drama', 'thriller', 'romance', 'action', 'animated']",

type = "string"

AttributeInfo(

name = "year",

description = "The year the movie was released",

type = "integer"

AttributeInfo(

name = "director",

description = "The name of the movie director",

type = "string"

AttributeInfo(

name = "rating",

description = "A 1-10 rating for the movie",

type = "float"

)

]

documentContentDescription = "Brief summary of a movie"

fewShotPromptTemplate = get_query_constructor_prompt(

documentContentDescription,

attributeInfoList

)

structuredQueryOutputParser = StructuredQueryOutputParser.from_components()

chatOpenAI = ChatOpenAI(temperature = 0)

runnableSequence = fewShotPromptTemplate | chatOpenAI | structuredQueryOutputParser

selfQueryRetriever = SelfQueryRetriever(

query_constructor = runnableSequence,

vectorstore = chroma,

structured_query_translator = ChromaTranslator(),

)

resultDocumentList = selfQueryRetriever.invoke("What's a movie after 1990 but before 2005 that's all about toys, and preferably is animated")

for resultDocument in resultDocumentList:

print(resultDocument)

[PYTHON/LANGCHAIN] SelfQueryRetriever 클래스 : from_llm 정적 메소드에서 enable_limit 인자를 사용해 SelfQueryRetriever 객체 만들기

■ SelfQueryRetriever 클래스의 from_llm 정적 메소드에서 enable_limit 인자를 사용해 SelfQueryRetriever 객체를 만드는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다.