GPT4O Archives - icodebroker

[PYTHON/LANGCHAIN] AsyncCallbackHandler 클래스 : on_llm_start/on_llm_end 메소드를 재정의해서 커스텀 비동기 콜백 핸들러 만들기

■ AsyncCallbackHandler 클래스의 on_llm_start/on_llm_end 메소드를 재정의해서 커스텀 비동기 콜백 핸들러를 만드는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] BaseCallbackHandler 클래스 : on_llm_new_token 메소드를 재정의해서 커스텀 콜백 핸들러 만들기

■ BaseCallbackHandler 클래스의 on_llm_new_token 메소드를 재정의해서 커스텀 콜백 핸들러를 만드는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py

[PYTHON/LANGCHAIN] ChatOpenAI 클래스 : 생성자에서 callbacks 인자를 사용해 콜백 핸들러 리스트 설정하기

■ ChatOpenAI 클래스의 생성자에서 callbacks 인자를 사용해 콜백 핸들러 리스트를 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] RunnableSequence 클래스 : with_config 메소드에서 callbacks 인자를 사용해 콜백 핸들러 리스트 설정하기

■ RunnableSequence 클래스의 with_config 메소드에서 callbacks 인자를 사용해 콜백 핸들러 리스트를 설정하는 방법을 보여준다. ▶ main.py


from dotenv                   import load_dotenv
from langchain_core.callbacks import BaseCallbackHandler
from typing                   import Dict
from typing                   import Any
from typing                   import List
from langchain_core.messages  import BaseMessage
from langchain_core.outputs   import LLMResult
from langchain_core.prompts   import ChatPromptTemplate
from langchain_openai         import ChatOpenAI

load_dotenv()

class CustomCallbackHandler(BaseCallbackHandler):
    def on_chat_model_start(self, serializedDictionary : Dict[str, Any], messageListList : List[List[BaseMessage]], **keywordArgumentDictionary) -> None:
        print("on_chat_model_start")
        print()

    def on_llm_end(self, result : LLMResult, **keywordArgumentDictionary) -> None:
        print(f"on_llm_end : {result}")
        print()

    def on_chain_start(self, serializedDictionary: Dict[str, Any], inputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:
        if serializedDictionary:
            print(f"on_chain_start : {serializedDictionary.get("name")}")
        else:
            print(f"on_chain_start : {serializedDictionary}")
        print()

    def on_chain_end(self, outputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:
        print(f"on_chain_end : {outputDictionary}")
        print()

chatPromptTemplate = ChatPromptTemplate.from_template("What is 1 + {number}?")

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

runnableSequence = chatPromptTemplate | chatOpenAI

callBackHandlerList = [CustomCallbackHandler()]

runnableBinding = runnableSequence.with_config(callbacks = callBackHandlerList)

responseAIMessage = runnableBinding.invoke({"number" : "2"})

print(responseAIMessage)

"""
on_chain_start : None

on_chain_start : ChatPromptTemplate

on_chain_end : messages=[HumanMessage(content='What is 1 + 2?', additional_kwargs={}, response_metadata={})]

on_chat_model_start

on_llm_end : generations=[[ChatGeneration(text='1 + 2 equals 3.', generation_info={'finish_reason': 'stop', 'logprobs': None}, message=AIMessage(content='1 + 2 equals 3.', additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82', 'finish_reason': 'stop', 'logprobs': None}, id='run-a0d0dff3-4469-49a3-96ad-41ed409bf28d-0', usage_metadata={'input_tokens': 15, 'output_tokens': 8, 'total_tokens': 23, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}))]] llm_output={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82'} run=None type='LLMResult'

on_chain_end : content='1 + 2 equals 3.' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82', 'finish_reason': 'stop', 'logprobs': None} id='run-a0d0dff3-4469-49a3-96ad-41ed409bf28d-0' usage_metadata={'input_tokens': 15, 'output_tokens': 8, 'total_tokens': 23, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}

content='1 + 2 equals 3.' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82', 'finish_reason': 'stop', 'logprobs': None} id='run-a0d0dff3-4469-49a3-96ad-41ed409bf28d-0' usage_metadata={'input_tokens': 15, 'output_tokens': 8, 'total_tokens': 23, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}
"""

from dotenv import load_dotenv

from langchain_core.callbacks import BaseCallbackHandler

from typing import Dict

from typing import Any

from typing import List

from langchain_core.messages import BaseMessage

from langchain_core.outputs import LLMResult

from langchain_core.prompts import ChatPromptTemplate

from langchain_openai import ChatOpenAI

load_dotenv()

class CustomCallbackHandler(BaseCallbackHandler):

def on_chat_model_start(self, serializedDictionary : Dict[str, Any], messageListList : List[List[BaseMessage]], **keywordArgumentDictionary) -> None:

print("on_chat_model_start")

print()

def on_llm_end(self, result : LLMResult, **keywordArgumentDictionary) -> None:

print(f"on_llm_end : {result}")

print()

def on_chain_start(self, serializedDictionary: Dict[str, Any], inputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:

if serializedDictionary:

print(f"on_chain_start : {serializedDictionary.get("name")}")

else:

print(f"on_chain_start : {serializedDictionary}")

print()

def on_chain_end(self, outputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:

print(f"on_chain_end : {outputDictionary}")

print()

chatPromptTemplate = ChatPromptTemplate.from_template("What is 1 + {number}?")

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

runnableSequence = chatPromptTemplate | chatOpenAI

callBackHandlerList = [CustomCallbackHandler()]

runnableBinding = runnableSequence.with_config(callbacks = callBackHandlerList)

responseAIMessage = runnableBinding.invoke({"number" : "2"})

print(responseAIMessage)

"""

on_chain_start : None

on_chain_start : ChatPromptTemplate

on_chain_end : messages=[HumanMessage(content='What is 1 + 2?', additional_kwargs={}, response_metadata={})]

on_chat_model_start

on_llm_end : generations=[[ChatGeneration(text='1 + 2 equals 3.', generation_info={'finish_reason': 'stop', 'logprobs': None}, message=AIMessage(content='1 + 2 equals 3.', additional_kwargs={'refusal': None}, response_metadata={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82', 'finish_reason': 'stop', 'logprobs': None}, id='run-a0d0dff3-4469-49a3-96ad-41ed409bf28d-0', usage_metadata={'input_tokens': 15, 'output_tokens': 8, 'total_tokens': 23, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}))]] llm_output={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82'} run=None type='LLMResult'

on_chain_end : content='1 + 2 equals 3.' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82', 'finish_reason': 'stop', 'logprobs': None} id='run-a0d0dff3-4469-49a3-96ad-41ed409bf28d-0' usage_metadata={'input_tokens': 15, 'output_tokens': 8, 'total_tokens': 23, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}

content='1 + 2 equals 3.' additional_kwargs={'refusal': None} response_metadata={'token_usage': {'completion_tokens': 8, 'prompt_tokens': 15, 'total_tokens': 23, 'completion_tokens_details': {'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0}, 'prompt_tokens_details': {'audio_tokens': 0, 'cached_tokens': 0}}, 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_831e067d82', 'finish_reason': 'stop', 'logprobs': None} id='run-a0d0dff3-4469-49a3-96ad-41ed409bf28d-0' usage_metadata={'input_tokens': 15, 'output_tokens': 8, 'total_tokens': 23, 'input_token_details': {'audio': 0, 'cache_read': 0}, 'output_token_details': {'audio': 0, 'reasoning': 0}}

"""

▶ requirements.txt


aiohappyeyeballs==2.4.3
aiohttp==3.11.6
aiosignal==1.3.1
annotated-types==0.7.0
anyio==4.6.2.post1
attrs==24.2.0
certifi==2024.8.30
charset-normalizer==3.4.0
colorama==0.4.6
distro==1.9.0
frozenlist==1.5.0
greenlet==3.1.1
h11==0.14.0
httpcore==1.0.7
httpx==0.27.2
idna==3.10
jiter==0.7.1
jsonpatch==1.33
jsonpointer==3.0.0
langchain==0.3.7
langchain-core==0.3.19
langchain-openai==0.2.9
langchain-text-splitters==0.3.2
langsmith==0.1.143
multidict==6.1.0
numpy==1.26.4
openai==1.54.5
orjson==3.10.11
packaging==24.2
propcache==0.2.0
pydantic==2.9.2
pydantic_core==2.23.4
python-dotenv==1.0.1
PyYAML==6.0.2
regex==2024.11.6
requests==2.32.3
requests-toolbelt==1.0.0
sniffio==1.3.1
SQLAlchemy==2.0.36
tenacity==9.0.0
tiktoken==0.8.0
tqdm==4.67.0
typing_extensions==4.12.2
urllib3==2.2.3
yarl==1.17.2

aiohappyeyeballs==2.4.3

aiohttp==3.11.6

aiosignal==1.3.1

annotated-types==0.7.0

anyio==4.6.2.post1

attrs==24.2.0

certifi==2024.8.30

charset-normalizer==3.4.0

colorama==0.4.6

distro==1.9.0

frozenlist==1.5.0

greenlet==3.1.1

h11==0.14.0

httpcore==1.0.7

httpx==0.27.2

idna==3.10

jiter==0.7.1

jsonpatch==1.33

jsonpointer==3.0.0

langchain==0.3.7

langchain-core==0.3.19

langchain-openai==0.2.9

langchain-text-splitters==0.3.2

langsmith==0.1.143

multidict==6.1.0

numpy==1.26.4

openai==1.54.5

orjson==3.10.11

packaging==24.2

propcache==0.2.0

pydantic==2.9.2

pydantic_core==2.23.4

python-dotenv==1.0.1

PyYAML==6.0.2

regex==2024.11.6

requests==2.32.3

requests-toolbelt==1.0.0

sniffio==1.3.1

SQLAlchemy==2.0.36

tenacity==9.0.0

tiktoken==0.8.0

tqdm==4.67.0

typing_extensions==4.12.2

urllib3==2.2.3

yarl==1.17.2

※ pip

[PYTHON/LANGCHAIN] RunnableSequence 클래스 : invoke 메소드에서 config 인자를 사용해 콜백 핸들러 설정하기

■ RunnableSequence 클래스의 invoke 메소드에서 config 인자를 사용해 콜백 핸들러를 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] BaseCallbackHandler 클래스 : 커스텀 콜백 핸들러 만들기

■ BaseCallbackHandler 클래스를 사용해 커스텀 콜백 핸들러를 만드는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                   import load_dotenv
from langchain_core.callbacks import BaseCallbackHandler
from typing                   import Dict
from typing                   import Any
from typing                   import List
from langchain_core.messages  import BaseMessage
from langchain_core.outputs   import LLMResult
from langchain_core.prompts   import ChatPromptTemplate
from langchain_openai         import ChatOpenAI

load_dotenv()

class CustomCallbackHandler(BaseCallbackHandler):
    def on_chat_model_start(self, serializedDictionary : Dict[str, Any], messageListList : List[List[BaseMessage]], **keywordArgumentDictionary) -> None:
        print("on_chat_model_start")
        print()

    def on_llm_end(self, result : LLMResult, **keywordArgumentDictionary) -> None:
        print(f"on_llm_end : {result}")
        print()

    def on_chain_start(self, serializedDictionary : Dict[str, Any], inputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:
        if serializedDictionary:
            print(f"on_chain_start : {serializedDictionary.get("name")}")
        else:
            print(f"on_chain_start : {serializedDictionary}")
        print()

    def on_chain_end(self, outputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:
        print(f"on_chain_end : {outputDictionary}")
        print()

chatPromptTemplate = ChatPromptTemplate.from_template("What is 1 + {number}?")

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

runnableSequence = chatPromptTemplate | chatOpenAI

callBackHandlerList = [CustomCallbackHandler()]

responseAIMessage  = runnableSequence.invoke({"number" : "2"}, config = {"callbacks" : callBackHandlerList})

print("-" * 100)
print(responseAIMessage)

from dotenv import load_dotenv

from langchain_core.callbacks import BaseCallbackHandler

from typing import Dict

from typing import Any

from typing import List

from langchain_core.messages import BaseMessage

from langchain_core.outputs import LLMResult

from langchain_core.prompts import ChatPromptTemplate

from langchain_openai import ChatOpenAI

load_dotenv()

class CustomCallbackHandler(BaseCallbackHandler):

def on_chat_model_start(self, serializedDictionary : Dict[str, Any], messageListList : List[List[BaseMessage]], **keywordArgumentDictionary) -> None:

print("on_chat_model_start")

print()

def on_llm_end(self, result : LLMResult, **keywordArgumentDictionary) -> None:

print(f"on_llm_end : {result}")

print()

def on_chain_start(self, serializedDictionary : Dict[str, Any], inputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:

if serializedDictionary:

print(f"on_chain_start : {serializedDictionary.get("name")}")

else:

print(f"on_chain_start : {serializedDictionary}")

print()

def on_chain_end(self, outputDictionary : Dict[str, Any], **keywordArgumentDictionary) -> None:

print(f"on_chain_end : {outputDictionary}")

print()

chatPromptTemplate = ChatPromptTemplate.from_template("What is 1 + {number}?")

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

runnableSequence = chatPromptTemplate | chatOpenAI

callBackHandlerList = [CustomCallbackHandler()]

responseAIMessage = runnableSequence.invoke({"number" : "2"}, config = {"callbacks" : callBackHandlerList})

print("-" * 100)

print(responseAIMessage)

▶

[PYTHON/LANGCHAIN] create_react_agent 함수 : state_modifier 인자를 사용해 장기 실행 에이전트의 중간 단계 제거하기

■ create_react_agent 함수의 state_modifier 인자를 사용해 장기 실행 에이전트의 중간 단계를 제거하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다.

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 trim_intermediate_steps 인자를 사용해 장기 실행 에이전트의 중간 단계 제거하기

■ AgentExecutor 클래스의 생성자에서 trim_intermediate_steps 인자를 사용해 장기 실행 에이전트의 중간 단계를 제거하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 early_stopping_method/max_iterations 인자를 사용해 반복 제한/시간 제한 중단 문자열 반환하기

■ AgentExecutor 클래스의 생성자에서 early_stopping_method/max_iterations 인자를 사용해 반복 제한/시간 제한 중단 문자열을 반환하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에

[PYTHON/LANGCHAIN] CompiledStateGraph 클래스 : asyncio 모듈의 create_task/wait_for 함수를 사용해 최대 실행 시간 설정하기

■ CompiledStateGraph 클래스에서 asyncio 패키지의 create_task/wait_for 함수를 사용해 최대 실행 시간을 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다.

[PYTHON/LANGCHAIN] CompiledStateGraph 클래스 : step_timeout 속성을 사용해 최대 실행 시간 설정하기

■ CompiledStateGraph 클래스의 step_timeout 속성을 사용해 최대 실행 시간을 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 max_execution_time 인자를 사용해 최대 실행 시간 설정하기

■ AgentExecutor 클래스의 생성자에서 max_execution_time 인자를 사용해 최대 실행 시간을 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶

[PYTHON/LANGCHAIN] CompiledStateGraph 클래스 : stream 메소드에서 재귀 제한 수 설정하기

■ CompiledStateGraph 클래스의 stream 메소드에서 재귀 제한 수를 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv               import load_dotenv
from langchain_openai     import ChatOpenAI
from langchain_core.tools import tool
from langgraph.prebuilt   import create_react_agent
from langgraph.errors     import GraphRecursionError

load_dotenv()

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

@tool
def magicFunction(input : int) -> int:
    """Applies a magic function to an input."""
    return input + 2

toolList = [magicFunction]

query = "what is the value of magicFunction(3)?"

RECURSION_LIMIT_COUNT = 2 * 3 + 1

compiledStateGraph = create_react_agent(chatOpenAI, tools = toolList)

try:
    for addableValuesDict in compiledStateGraph.stream(
        {"messages" : [("human", query)]},
        {"recursion_limit" : RECURSION_LIMIT_COUNT},
        stream_mode = "values"
    ):
        print(addableValuesDict["messages"][-1])
except GraphRecursionError:
    print({"input" : query, "output" : "Agent stopped due to max iterations."})

from dotenv import load_dotenv

from langchain_openai import ChatOpenAI

from langchain_core.tools import tool

from langgraph.prebuilt import create_react_agent

from langgraph.errors import GraphRecursionError

load_dotenv()

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

@tool

def magicFunction(input : int) -> int:

"""Applies a magic function to an input."""

return input + 2

toolList = [magicFunction]

query = "what is the value of magicFunction(3)?"

RECURSION_LIMIT_COUNT = 2 * 3 + 1

compiledStateGraph = create_react_agent(chatOpenAI, tools = toolList)

try:

for addableValuesDict in compiledStateGraph.stream(

{"messages" : [("human", query)]},

{"recursion_limit" : RECURSION_LIMIT_COUNT},

stream_mode = "values"

print(addableValuesDict["messages"][-1])

except GraphRecursionError:

print({"input" : query, "output" : "Agent stopped due to max iterations."})

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 max_iterations 인자를 사용해 사용자 지정 반복 횟수 초과 실행시 중단하기

■ AgentExecutor 클래스의 생성자에서 max_iterations 인자를 사용해 사용자 지정 반복 횟수 초과 실행시 중단하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 return_intermediate_steps 인자를 사용해 중단 단계 반환하기

■ AgentExecutor 클래스의 생성자에서 return_intermediate_steps 인자를 사용해 중단 단계를 반환하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : stream 메소드를 사용해 질의 응답하기

■ AgentExecutor 클래스의 stream 메소드를 사용해 질의 응답하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                 import load_dotenv
from langchain_openai       import ChatOpenAI
from langchain_core.tools   import tool
from langchain_core.prompts import ChatPromptTemplate
from langchain.agents       import create_tool_calling_agent
from langchain.agents       import AgentExecutor

load_dotenv()

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

chatPromptTemplate = ChatPromptTemplate.from_messages(
    [
        ("system"     , "You are a helpful assistant."),
        ("human"      , "{input}"                     ),
        ("placeholder", "{agent_scratchpad}"          )
    ]
)

@tool
def magicFunction(input : int) -> int:
    """Applies a magic function to an input."""
    return input + 2

toolList = [magicFunction]

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, prompt = chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

query = "what is the value of magic_function(3)?"

for addableDict in agentExecutor.stream({"input" : query}):
    print(addableDict)
    print("-" * 100)

"""
{'actions': [ToolAgentAction(tool='magic_function', tool_input={'input': 3}, log="\nInvoking: `magic_function` with `{'input': 3}`\n\n\n", message_log=[AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'function': {'arguments': '{"input":3}', 'name': 'magic_function'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-15ad5b62-46c6-411d-b7f6-fcee0f04d6cd', tool_calls=[{'name': 'magic_function', 'args': {'input': 3}, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magic_function', 'args': '{"input":3}', 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'index': 0, 'type': 'tool_call_chunk'}])], tool_call_id='call_7iH5SstP0NB75GGRzulZZ53J')], 'messages': [AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'function': {'arguments': '{"input":3}', 'name': 'magic_function'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-15ad5b62-46c6-411d-b7f6-fcee0f04d6cd', tool_calls=[{'name': 'magic_function', 'args': {'input': 3}, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magic_function', 'args': '{"input":3}', 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'index': 0, 'type': 'tool_call_chunk'}])]}
{'steps': [AgentStep(action=ToolAgentAction(tool='magic_function', tool_input={'input': 3}, log="\nInvoking: `magic_function` with `{'input': 3}`\n\n\n", message_log=[AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'function': {'arguments': '{"input":3}', 'name': 'magic_function'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-15ad5b62-46c6-411d-b7f6-fcee0f04d6cd', tool_calls=[{'name': 'magic_function', 'args': {'input': 3}, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magic_function', 'args': '{"input":3}', 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'index': 0, 'type': 'tool_call_chunk'}])], tool_call_id='call_7iH5SstP0NB75GGRzulZZ53J'), observation=5)], 'messages': [FunctionMessage(content='5', additional_kwargs={}, response_metadata={}, name='magic_function')]}
{'output': 'The value of `magic_function(3)` is 5.', 'messages': [AIMessage(content='The value of `magic_function(3)` is 5.', additional_kwargs={}, response_metadata={})]}
(env) king@cosmos:~/testproject$ python main.py
<class 'langchain_core.runnables.utils.AddableDict'>
(env) king@cosmos:~/testproject$ python main.py
{'actions': [ToolAgentAction(tool='magicFunction', tool_input={'input': 3}, log="\nInvoking: `magicFunction` with `{'input': 3}`\n\n\n", message_log=[AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'function': {'arguments': '{"input":3}', 'name': 'magicFunction'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-e8b2e03f-f3a8-4b68-90ab-5dcfbb35659e', tool_calls=[{'name': 'magicFunction', 'args': {'input': 3}, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magicFunction', 'args': '{"input":3}', 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'index': 0, 'type': 'tool_call_chunk'}])], tool_call_id='call_Qv8SUJFIEtg02siem3SE7Nxj')], 'messages': [AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'function': {'arguments': '{"input":3}', 'name': 'magicFunction'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-e8b2e03f-f3a8-4b68-90ab-5dcfbb35659e', tool_calls=[{'name': 'magicFunction', 'args': {'input': 3}, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magicFunction', 'args': '{"input":3}', 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'index': 0, 'type': 'tool_call_chunk'}])]}

from dotenv import load_dotenv

from langchain_openai import ChatOpenAI

from langchain_core.tools import tool

from langchain_core.prompts import ChatPromptTemplate

from langchain.agents import create_tool_calling_agent

from langchain.agents import AgentExecutor

load_dotenv()

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

chatPromptTemplate = ChatPromptTemplate.from_messages(

[

("system" , "You are a helpful assistant."),

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

("placeholder", "{agent_scratchpad}" )

]

)

@tool

def magicFunction(input : int) -> int:

"""Applies a magic function to an input."""

return input + 2

toolList = [magicFunction]

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, prompt = chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

query = "what is the value of magic_function(3)?"

for addableDict in agentExecutor.stream({"input" : query}):

print(addableDict)

print("-" * 100)

"""

{'actions': [ToolAgentAction(tool='magic_function', tool_input={'input': 3}, log="\nInvoking: `magic_function` with `{'input': 3}`\n\n\n", message_log=[AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'function': {'arguments': '{"input":3}', 'name': 'magic_function'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-15ad5b62-46c6-411d-b7f6-fcee0f04d6cd', tool_calls=[{'name': 'magic_function', 'args': {'input': 3}, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magic_function', 'args': '{"input":3}', 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'index': 0, 'type': 'tool_call_chunk'}])], tool_call_id='call_7iH5SstP0NB75GGRzulZZ53J')], 'messages': [AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'function': {'arguments': '{"input":3}', 'name': 'magic_function'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-15ad5b62-46c6-411d-b7f6-fcee0f04d6cd', tool_calls=[{'name': 'magic_function', 'args': {'input': 3}, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magic_function', 'args': '{"input":3}', 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'index': 0, 'type': 'tool_call_chunk'}])]}

{'steps': [AgentStep(action=ToolAgentAction(tool='magic_function', tool_input={'input': 3}, log="\nInvoking: `magic_function` with `{'input': 3}`\n\n\n", message_log=[AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'function': {'arguments': '{"input":3}', 'name': 'magic_function'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-15ad5b62-46c6-411d-b7f6-fcee0f04d6cd', tool_calls=[{'name': 'magic_function', 'args': {'input': 3}, 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magic_function', 'args': '{"input":3}', 'id': 'call_7iH5SstP0NB75GGRzulZZ53J', 'index': 0, 'type': 'tool_call_chunk'}])], tool_call_id='call_7iH5SstP0NB75GGRzulZZ53J'), observation=5)], 'messages': [FunctionMessage(content='5', additional_kwargs={}, response_metadata={}, name='magic_function')]}

{'output': 'The value of `magic_function(3)` is 5.', 'messages': [AIMessage(content='The value of `magic_function(3)` is 5.', additional_kwargs={}, response_metadata={})]}

(env) king@cosmos:~/testproject$ python main.py

(env) king@cosmos:~/testproject$ python main.py

{'actions': [ToolAgentAction(tool='magicFunction', tool_input={'input': 3}, log="\nInvoking: `magicFunction` with `{'input': 3}`\n\n\n", message_log=[AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'function': {'arguments': '{"input":3}', 'name': 'magicFunction'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-e8b2e03f-f3a8-4b68-90ab-5dcfbb35659e', tool_calls=[{'name': 'magicFunction', 'args': {'input': 3}, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magicFunction', 'args': '{"input":3}', 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'index': 0, 'type': 'tool_call_chunk'}])], tool_call_id='call_Qv8SUJFIEtg02siem3SE7Nxj')], 'messages': [AIMessageChunk(content='', additional_kwargs={'tool_calls': [{'index': 0, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'function': {'arguments': '{"input":3}', 'name': 'magicFunction'}, 'type': 'function'}]}, response_metadata={'finish_reason': 'tool_calls', 'model_name': 'gpt-4o-2024-08-06', 'system_fingerprint': 'fp_6b68a8204b'}, id='run-e8b2e03f-f3a8-4b68-90ab-5dcfbb35659e', tool_calls=[{'name': 'magicFunction', 'args': {'input': 3}, 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'type': 'tool_call'}], tool_call_chunks=[{'name': 'magicFunction', 'args': '{"input":3}', 'id': 'call_Qv8SUJFIEtg02siem3SE7Nxj', 'index': 0, 'type': 'tool_call_chunk'}])]}

{'steps':

[PYTHON/LANGCHAIN] create_react_agent 함수 : checkpointer 인자에서 MemorySaver 객체 설정하기

■ create_react_agent 함수의 checkpointer 인자에서 MemorySaver 객체를 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                           import load_dotenv
from langchain_openai                 import ChatOpenAI
from langchain_core.tools             import tool
from langchain_core.messages          import SystemMessage
from langgraph.checkpoint.memory      import MemorySaver # 인 메모리 체크 포인터
from langgraph.prebuilt               import create_react_agent

load_dotenv()

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

@tool
def magicFunction(input : int) -> int:
    """Applies a magic function to an input."""
    return input + 2

toolList = [magicFunction]

systemMessage = SystemMessage(content = "You are a helpful assistant. Respond only in Korean.")

memorySaver = MemorySaver()

compiledStateGraph = create_react_agent(
    chatOpenAI,
    toolList,
    state_modifier = systemMessage,
    checkpointer   = memorySaver
)

configurationDictionary = {"configurable" : {"thread_id" : "thread1"}}

addableValuesDict1 = compiledStateGraph.invoke(
    {"messages" : [("user", "Hi, I'm polly! What's the output of magic_function of 3?")]},
    configurationDictionary,
)

print(addableValuesDict1["messages"][-1].content)
print("-" * 100)

addableValuesDict2 = compiledStateGraph.invoke(
    {"messages" : [("user", "Remember my name?")]},
    configurationDictionary
)

print(addableValuesDict2["messages"][-1].content)
print("-" * 100)

addableValuesDict3 = compiledStateGraph.invoke(
    {"messages" : [("user", "what was that output again?")]},
    configurationDictionary
)

print(addableValuesDict3["messages"][-1].content)

"""
매직 함수에 입력값 3을 넣으면 출력값은 5입니다.

from dotenv import load_dotenv

from langchain_openai import ChatOpenAI

from langchain_core.tools import tool

from langchain_core.messages import SystemMessage

from langgraph.checkpoint.memory import MemorySaver # 인 메모리 체크 포인터

from langgraph.prebuilt import create_react_agent

load_dotenv()

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

@tool

def magicFunction(input : int) -> int:

"""Applies a magic function to an input."""

return input + 2

toolList = [magicFunction]

systemMessage = SystemMessage(content = "You are a helpful assistant. Respond only in Korean.")

memorySaver = MemorySaver()

compiledStateGraph = create_react_agent(

chatOpenAI,

toolList,

state_modifier = systemMessage,

checkpointer = memorySaver

)

configurationDictionary = {"configurable" : {"thread_id" : "thread1"}}

addableValuesDict1 = compiledStateGraph.invoke(

{"messages" : [("user", "Hi, I'm polly! What's the output of magic_function of 3?")]},

configurationDictionary,

)

print(addableValuesDict1["messages"][-1].content)

print("-" * 100)

addableValuesDict2 = compiledStateGraph.invoke(

{"messages" : [("user", "Remember my name?")]},

configurationDictionary

)

print(addableValuesDict2["messages"][-1].content)

print("-" * 100)

addableValuesDict3 = compiledStateGraph.invoke(

{"messages" : [("user", "what was that output again?")]},

configurationDictionary

)

print(addableValuesDict3["messages"][-1].content)

"""

매직 함수에 입력값 3을 넣으면 출력값은 5입니다.

당신의

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : invoke 메소드를 사용해 채팅하기 2

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


from dotenv                           import load_dotenv
from langchain_openai                 import ChatOpenAI
from langchain_core.tools             import tool
from langchain_core.prompts           import ChatPromptTemplate
from langchain.agents                 import create_tool_calling_agent
from langchain.agents                 import AgentExecutor
from langchain_core.chat_history      import InMemoryChatMessageHistory
from langchain_core.runnables.history import RunnableWithMessageHistory

load_dotenv()

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

@tool
def magicFunction(input : int) -> int:
    """Applies a magic function to an input."""
    return input + 2

toolList = [magicFunction]

chatPromptTemplate = ChatPromptTemplate.from_messages(
    [
        ("system"     , "You are a helpful assistant."),
        ("placeholder", "{chat_history}"              ),
        ("human"      , "{input}"                     ),
        ("placeholder", "{agent_scratchpad}"          )
    ]
)

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

inMemoryChatMessageHistory = InMemoryChatMessageHistory(session_id = "test-session")

runnableWithMessageHistory = RunnableWithMessageHistory(
    agentExecutor,
    lambda session_id : inMemoryChatMessageHistory,
    input_messages_key   = "input",
    history_messages_key = "chat_history"
)

configurationDictionary = {"configurable" : {"session_id" : "test-session"}}

responseDictionary1 = runnableWithMessageHistory.invoke({"input" : "Hi, I'm polly! What's the output of magic_function of 3?"}, configurationDictionary)

outputString1 = responseDictionary1["output"]

print(outputString1)
print("-" * 100)

responseDictionary2 = runnableWithMessageHistory.invoke({"input" : "Remember my name?"}, configurationDictionary)

outputString2 = responseDictionary2["output"]

print(outputString2)
print("-" * 100)

responseDictionary3 = runnableWithMessageHistory.invoke({"input": "what was that output again?"}, configurationDictionary)

outputString3 = responseDictionary3["output"]

print(outputString3)

"""
Hello Polly! The output of the magic function for the input 3 is 5.

from dotenv import load_dotenv

from langchain_openai import ChatOpenAI

from langchain_core.tools import tool

from langchain_core.prompts import ChatPromptTemplate

from langchain.agents import create_tool_calling_agent

from langchain.agents import AgentExecutor

from langchain_core.chat_history import InMemoryChatMessageHistory

from langchain_core.runnables.history import RunnableWithMessageHistory

load_dotenv()

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

@tool

def magicFunction(input : int) -> int:

"""Applies a magic function to an input."""

return input + 2

toolList = [magicFunction]

chatPromptTemplate = ChatPromptTemplate.from_messages(

[

("system" , "You are a helpful assistant."),

("placeholder", "{chat_history}" ),

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

("placeholder", "{agent_scratchpad}" )

]

)

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

inMemoryChatMessageHistory = InMemoryChatMessageHistory(session_id = "test-session")

runnableWithMessageHistory = RunnableWithMessageHistory(

agentExecutor,

lambda session_id : inMemoryChatMessageHistory,

input_messages_key = "input",

history_messages_key = "chat_history"

)

configurationDictionary = {"configurable" : {"session_id" : "test-session"}}

responseDictionary1 = runnableWithMessageHistory.invoke({"input" : "Hi, I'm polly! What's the output of magic_function of 3?"}, configurationDictionary)

outputString1 = responseDictionary1["output"]

print(outputString1)

print("-" * 100)

responseDictionary2 = runnableWithMessageHistory.invoke({"input" : "Remember my name?"}, configurationDictionary)

outputString2 = responseDictionary2["output"]

print(outputString2)

print("-" * 100)

responseDictionary3 = runnableWithMessageHistory.invoke({"input": "what was that output again?"}, configurationDictionary)

outputString3 = responseDictionary3["output"]

print(outputString3)

"""

Hello Polly! The output of the magic function for the input 3 is 5.

Yes, you

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : 생성자에서 input_messages_key/history_messages_key 인자를 사용해 RunnableWithMessageHistory 객체 만들기 2

■ RunnableWithMessageHistory 클래스의 생성자에서 input_messages_key/history_messages_key 인자를 사용해 RunnableWithMessageHistory 객체를 만드는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : invoke 메소드를 사용해 채팅하기 1

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


from dotenv                                     import load_dotenv
from langchain_community.tools.tavily_search    import TavilySearchResults
from langchain_community.document_loaders       import WebBaseLoader
from langchain_text_splitters                   import RecursiveCharacterTextSplitter
from langchain_openai                           import OpenAIEmbeddings
from langchain_community.vectorstores           import FAISS
from langchain.tools.retriever                  import create_retriever_tool
from langchain_openai                           import ChatOpenAI
from langchain                                  import hub
from langchain.agents                           import create_tool_calling_agent
from langchain.agents                           import AgentExecutor
from langchain_core.chat_history                import BaseChatMessageHistory
from langchain_community.chat_message_histories import ChatMessageHistory
from langchain_core.runnables.history           import RunnableWithMessageHistory

load_dotenv()

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

tavilySearchResults = TavilySearchResults(max_results = 2)

webBaseLoader = WebBaseLoader("https://docs.smith.langchain.com/overview")

documentList = webBaseLoader.load()

recursiveCharacterTextSplitter = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200)

splitDocumentList = recursiveCharacterTextSplitter.split_documents(documentList)

openAIEmbeddings = OpenAIEmbeddings()

faiss = FAISS.from_documents(splitDocumentList, openAIEmbeddings)

vectorStoreRetriever = faiss.as_retriever()

vectorStoreRetrieverTool = create_retriever_tool(
    vectorStoreRetriever,
    "langsmith_search",
    "Search for information about LangSmith. For any questions about LangSmith, you must use this tool!",
)

toolList = [tavilySearchResults, vectorStoreRetrieverTool]

chatPromptTemplate = hub.pull("hwchase17/openai-functions-agent")

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

sessionIDDictionary = {}

def getChatMessageHistory(sessionID : str) -> BaseChatMessageHistory:
    if sessionID not in sessionIDDictionary:
        sessionIDDictionary[sessionID] = ChatMessageHistory()
    return sessionIDDictionary[sessionID]

runnableWithMessageHistory = RunnableWithMessageHistory(
    agentExecutor,
    getChatMessageHistory,
    input_messages_key   = "input",
    history_messages_key = "chat_history",
)

responseDictionary1 = runnableWithMessageHistory.invoke(
    {"input" : "hi! I'm bob"},
    config = {"configurable" : {"session_id" : "<foo>"}}
)

print(responseDictionary1)
print("-" * 100)

responseDictionary2 = runnableWithMessageHistory.invoke(
    {"input" : "what's my name?"},
    config = {"configurable" : {"session_id" : "<foo>"}}
)

print(responseDictionary2)
print("-" * 100)

"""
{'input': "hi! I'm bob", 'chat_history': [], 'output': 'Hello Bob! How can I assist you today?'}

from dotenv import load_dotenv

from langchain_community.tools.tavily_search import TavilySearchResults

from langchain_community.document_loaders import WebBaseLoader

from langchain_text_splitters import RecursiveCharacterTextSplitter

from langchain_openai import OpenAIEmbeddings

from langchain_community.vectorstores import FAISS

from langchain.tools.retriever import create_retriever_tool

from langchain_openai import ChatOpenAI

from langchain import hub

from langchain.agents import create_tool_calling_agent

from langchain.agents import AgentExecutor

from langchain_core.chat_history import BaseChatMessageHistory

from langchain_community.chat_message_histories import ChatMessageHistory

from langchain_core.runnables.history import RunnableWithMessageHistory

load_dotenv()

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

tavilySearchResults = TavilySearchResults(max_results = 2)

webBaseLoader = WebBaseLoader("https://docs.smith.langchain.com/overview")

documentList = webBaseLoader.load()

recursiveCharacterTextSplitter = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200)

splitDocumentList = recursiveCharacterTextSplitter.split_documents(documentList)

openAIEmbeddings = OpenAIEmbeddings()

faiss = FAISS.from_documents(splitDocumentList, openAIEmbeddings)

vectorStoreRetriever = faiss.as_retriever()

vectorStoreRetrieverTool = create_retriever_tool(

vectorStoreRetriever,

"langsmith_search",

"Search for information about LangSmith. For any questions about LangSmith, you must use this tool!",

)

toolList = [tavilySearchResults, vectorStoreRetrieverTool]

chatPromptTemplate = hub.pull("hwchase17/openai-functions-agent")

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

sessionIDDictionary = {}

def getChatMessageHistory(sessionID : str) -> BaseChatMessageHistory:

if sessionID not in sessionIDDictionary:

sessionIDDictionary[sessionID] = ChatMessageHistory()

return sessionIDDictionary[sessionID]

runnableWithMessageHistory = RunnableWithMessageHistory(

agentExecutor,

getChatMessageHistory,

input_messages_key = "input",

history_messages_key = "chat_history",

)

responseDictionary1 = runnableWithMessageHistory.invoke(

{"input" : "hi! I'm bob"},

config = {"configurable" : {"session_id" : "<foo>"}}

)

print(responseDictionary1)

print("-" * 100)

responseDictionary2 = runnableWithMessageHistory.invoke(

{"input" : "what's my name?"},

config = {"configurable" : {"session_id" : "<foo>"}}

)

print(responseDictionary2)

print("-" * 100)

"""

{'input': "hi! I'm bob", 'chat_history': [], 'output': 'Hello Bob! How can I assist you today?'}

{'input': "what's

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : 생성자에서 input_messages_key/history_messages_key 인자를 사용해 RunnableWithMessageHistory 객체 만들기 1

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : invoke 메소드 사용시 채팅 히스토리 설정하기

■ AgentExecutor 클래스의 invoke 메소드 사용시 채팅 히스토리를 설정하는 방법을 보여준다. ※ OPENAI_API_KEY 환경 변수 값은 .env 파일에 정의한다. ▶ main.py


from dotenv                                  import load_dotenv
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_community.document_loaders    import WebBaseLoader
from langchain_text_splitters                import RecursiveCharacterTextSplitter
from langchain_openai                        import OpenAIEmbeddings
from langchain_community.vectorstores        import FAISS
from langchain.tools.retriever               import create_retriever_tool
from langchain_openai                        import ChatOpenAI
from langchain                               import hub
from langchain.agents                        import create_tool_calling_agent
from langchain.agents                        import AgentExecutor
from langchain_core.messages                 import HumanMessage
from langchain_core.messages                 import AIMessage

load_dotenv()

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

tavilySearchResults = TavilySearchResults(max_results = 2)

webBaseLoader = WebBaseLoader("https://docs.smith.langchain.com/overview")

documentList = webBaseLoader.load()

recursiveCharacterTextSplitter = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200)

splitDocumentList = recursiveCharacterTextSplitter.split_documents(documentList)

openAIEmbeddings = OpenAIEmbeddings()

faiss = FAISS.from_documents(splitDocumentList, openAIEmbeddings)

vectorStoreRetriever = faiss.as_retriever()

vectorStoreRetrieverTool = create_retriever_tool(
    vectorStoreRetriever,
    "langsmith_search",
    "Search for information about LangSmith. For any questions about LangSmith, you must use this tool!",
)

toolList = [tavilySearchResults, vectorStoreRetrieverTool]

chatPromptTemplate = hub.pull("hwchase17/openai-functions-agent")

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

responseDictionary = agentExecutor.invoke(
    {
        "chat_history" : [
            HumanMessage(content = "hi! my name is bob"),
            AIMessage(content = "Hello Bob! How can I assist you today?")
        ],
        "input" : "what's my name?"
    }
)

print(responseDictionary)

"""
{
    'chat_history' : [
        HumanMessage(content = 'hi! my name is bob', additional_kwargs = {}, response_metadata = {}),
        AIMessage(content = 'Hello Bob! How can I assist you today?', additional_kwargs = {}, response_metadata = {})
    ],
    'input'  : "what's my name?",
    'output' : 'Your name is Bob.'
}
"""

from dotenv import load_dotenv

from langchain_community.tools.tavily_search import TavilySearchResults

from langchain_community.document_loaders import WebBaseLoader

from langchain_text_splitters import RecursiveCharacterTextSplitter

from langchain_openai import OpenAIEmbeddings

from langchain_community.vectorstores import FAISS

from langchain.tools.retriever import create_retriever_tool

from langchain_openai import ChatOpenAI

from langchain import hub

from langchain.agents import create_tool_calling_agent

from langchain.agents import AgentExecutor

from langchain_core.messages import HumanMessage

from langchain_core.messages import AIMessage

load_dotenv()

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

tavilySearchResults = TavilySearchResults(max_results = 2)

webBaseLoader = WebBaseLoader("https://docs.smith.langchain.com/overview")

documentList = webBaseLoader.load()

recursiveCharacterTextSplitter = RecursiveCharacterTextSplitter(chunk_size = 1000, chunk_overlap = 200)

splitDocumentList = recursiveCharacterTextSplitter.split_documents(documentList)

openAIEmbeddings = OpenAIEmbeddings()

faiss = FAISS.from_documents(splitDocumentList, openAIEmbeddings)

vectorStoreRetriever = faiss.as_retriever()

vectorStoreRetrieverTool = create_retriever_tool(

vectorStoreRetriever,

"langsmith_search",

"Search for information about LangSmith. For any questions about LangSmith, you must use this tool!",

)

toolList = [tavilySearchResults, vectorStoreRetrieverTool]

chatPromptTemplate = hub.pull("hwchase17/openai-functions-agent")

runnableSequence = create_tool_calling_agent(chatOpenAI, toolList, chatPromptTemplate)

agentExecutor = AgentExecutor(agent = runnableSequence, tools = toolList)

responseDictionary = agentExecutor.invoke(

{

"chat_history" : [

HumanMessage(content = "hi! my name is bob"),

AIMessage(content = "Hello Bob! How can I assist you today?")

"input" : "what's my name?"

}

)

print(responseDictionary)

"""

{

'chat_history' : [

HumanMessage(content = 'hi! my name is bob', additional_kwargs = {}, response_metadata = {}),

AIMessage(content = 'Hello Bob! How can I assist you today?', additional_kwargs = {}, response_metadata = {})

'input' : "what's my name?",

'output' : 'Your name is Bob.'

}

"""

[태그:] GPT4O

[PYTHON/LANGCHAIN] AsyncCallbackHandler 클래스 : on_llm_start/on_llm_end 메소드를 재정의해서 커스텀 비동기 콜백 핸들러 만들기

[PYTHON/LANGCHAIN] BaseCallbackHandler 클래스 : on_llm_new_token 메소드를 재정의해서 커스텀 콜백 핸들러 만들기

[PYTHON/LANGCHAIN] ChatOpenAI 클래스 : 생성자에서 callbacks 인자를 사용해 콜백 핸들러 리스트 설정하기

[PYTHON/LANGCHAIN] RunnableSequence 클래스 : with_config 메소드에서 callbacks 인자를 사용해 콜백 핸들러 리스트 설정하기

[PYTHON/LANGCHAIN] RunnableSequence 클래스 : invoke 메소드에서 config 인자를 사용해 콜백 핸들러 설정하기

[PYTHON/LANGCHAIN] BaseCallbackHandler 클래스 : 커스텀 콜백 핸들러 만들기

[PYTHON/LANGCHAIN] create_react_agent 함수 : state_modifier 인자를 사용해 장기 실행 에이전트의 중간 단계 제거하기

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 trim_intermediate_steps 인자를 사용해 장기 실행 에이전트의 중간 단계 제거하기

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 early_stopping_method/max_iterations 인자를 사용해 반복 제한/시간 제한 중단 문자열 반환하기

[PYTHON/LANGCHAIN] CompiledStateGraph 클래스 : asyncio 모듈의 create_task/wait_for 함수를 사용해 최대 실행 시간 설정하기

[PYTHON/LANGCHAIN] CompiledStateGraph 클래스 : step_timeout 속성을 사용해 최대 실행 시간 설정하기

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 max_execution_time 인자를 사용해 최대 실행 시간 설정하기

[PYTHON/LANGCHAIN] CompiledStateGraph 클래스 : stream 메소드에서 재귀 제한 수 설정하기

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 max_iterations 인자를 사용해 사용자 지정 반복 횟수 초과 실행시 중단하기

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 return_intermediate_steps 인자를 사용해 중단 단계 반환하기

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : stream 메소드를 사용해 질의 응답하기

[PYTHON/LANGCHAIN] create_react_agent 함수 : checkpointer 인자에서 MemorySaver 객체 설정하기

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : invoke 메소드를 사용해 채팅하기 2

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : 생성자에서 input_messages_key/history_messages_key 인자를 사용해 RunnableWithMessageHistory 객체 만들기 2

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : invoke 메소드를 사용해 채팅하기 1

[PYTHON/LANGCHAIN] RunnableWithMessageHistory 클래스 : 생성자에서 input_messages_key/history_messages_key 인자를 사용해 RunnableWithMessageHistory 객체 만들기 1

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : invoke 메소드 사용시 채팅 히스토리 설정하기

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : OPENAI 모델과 TAVILY 도구를 사용해 질의 응답하기 2

[PYTHON/LANGCHAIN] AgentExecutor 클래스 : 생성자에서 agent/tools 인자를 사용해 AgentExecutor 객체 만들기 2

[PYTHON/LANGCHAIN] create_tool_calling_agent 함수 : 모델, 도구 및 프롬프트 템플리트를 결합한 RunnableSequence 객체 만들기 2