LLMSの応用：Langchainの基礎

!pip install -U langchain
!pip install langchain-community
!pip install langchain-core
!pip install langsmith
!pip install openai
!pip install python-dotenv
!pip install -U langchain-openai
!pip install -U langchain-google-genai pillow
!pip install google-generativeai
!pip install google-search-results numexpr wikipedia langchain-experimental

!pip install duckduckgo-search

from dotenv import load_dotenv
import os
import getpass
from langchain_openai import OpenAI, ChatOpenAI     

load_dotenv(dotenv_path='../.env')
if "OPENAI_API_KEY" not in os.environ:
    os.environ["OPENAI_API_KEY"] = getpass.getpass("Enter your OpenAI API key: ")

llm = OpenAI(
    model="gpt-3.5-turbo-instruct",
    temperature=0.9,
    max_tokens=512
)

llm.invoke("東北大学を紹介してください：")

from langchain_openai.chat_models import ChatOpenAI
from langchain_core.messages import HumanMessage

model = ChatOpenAI()
prompt = [HumanMessage("東北大学を紹介してください")]

model.invoke(prompt)

from langchain_core.messages import HumanMessage, SystemMessage
from langchain_openai.chat_models import ChatOpenAI

model = ChatOpenAI()         
system_msg = SystemMessage(
    '''You are a helpful assistant that responds to questions with three exclamation marks.'''
)# あなたは優秀なアシスタントです。質問に対して3つの感嘆符を付けて回答してください。

human_msg = HumanMessage("東北大学を紹介してください")
model.invoke([system_msg, human_msg])

if "GOOGLE_API_KEY" not in os.environ:
    os.environ["GOOGLE_API_KEY"] = getpass.getpass("Enter your GOOGLE API key: ")

import google.generativeai as genai
models = [m for m in genai.list_models()]
models

import google.generativeai as genai

llm = genai.GenerativeModel('models/gemini-2.5-flash')

response = llm.generate_content("東北大学を紹介してください：")

response.text

from langchain_core.prompts import PromptTemplate

prompt = PromptTemplate(
    input_variables=["language","text"],
    template="次の日本語のテキストを{language}に翻訳してください：{text}",
)

print(prompt.format(language="英語", text="東北大学は日本の東北地方にある大学です。"))

from langchain_google_genai import ChatGoogleGenerativeAI

llm = ChatGoogleGenerativeAI(model="models/gemini-2.5-flash", google_api_key=os.getenv("GOOGLE_API_KEY"))

chain = prompt | llm
chain.invoke(
    {"language":"英語", "text":"東北大学は日本の東北地方にある大学です。"}
)

chain.invoke({"language":"中国語", "text":"東北大学は日本の東北地方にある大学です。"})

from langchain_core.prompts import PromptTemplate, FewShotPromptTemplate

examples = [
    {"word": "楽しい", "antonym": "悲しい"},
    {"word": "高い", "antonym": "低い"},
]

example_formatter_template = """
Word: {word}
Antonym: {antonym}\n
"""

example_prompt = PromptTemplate(
    input_variables=["word", "antonym"],
    template=example_formatter_template,
)

few_shot_prompt = FewShotPromptTemplate(
    examples=examples,
    example_prompt=example_prompt,
    prefix="Give the antonym of every input",
    suffix="Word: {input}\nAntonym:",
    input_variables=["input"],
    example_separator="\n",
)

print(few_shot_prompt.format(input="大きい"))

llm = ChatGoogleGenerativeAI(model="models/gemini-2.5-flash", google_api_key=os.getenv("GOOGLE_API_KEY"))

from langchain_core.output_parsers import StrOutputParser

chain = few_shot_prompt | llm | StrOutputParser()

var = few_shot_prompt.input_variables[0] 
print(chain.invoke({var: "大きい"}))

#!pip install datasets
from datasets import load_dataset
dataset = load_dataset("imdb")

df_train_sample = dataset["train"].to_pandas().sample(5, random_state = 123)
df_test_sample = dataset["test"].to_pandas().sample(5, random_state = 123)

df_train_sample

from langchain_openai import ChatOpenAI
from langchain.agents import create_agent
from langchain_community.agent_toolkits.load_tools import load_tools

# 1. チャットモデルの初期化
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0, verbose=True)

# 2. ツールの読み込み
tools = load_tools(["llm-math", "wikipedia"], llm=llm)

# 3. エージェントの作成
agent = create_agent(
    llm,
    tools,
    system_prompt="""You are a helpful assistant with access to math and wikipedia tools.
Use tools when needed to answer questions accurately. Think step-by-step internally before answering."""
)

agent

# エージェントを使わず、本来のLLMsに回答を出させる場合
llm.invoke("現在の日本の総理大臣は誰ですか？")

result = agent.invoke(
    {"messages": [{"role": "user", "content": "現在の日本の総理大臣は誰ですか？"}]}
)

result

225.555*222.555

llm.invoke("225.555*222.555=")

result = agent.invoke(
    {"messages": [{"role": "user", "content": "225.555*222.555="}]}
)

result

from langsmith import Client
from langchain_openai import ChatOpenAI

client = Client()

# 自分のプロンプト（例）
prompt = client.pull_prompt("hardkothari/blog-generator")
model = ChatOpenAI(model="gpt-4o-mini")
chain = prompt | model

prompt

chain.invoke({'target_audience': "初心者",
              "text": "Langchain hubとは何ですか？"})