Google主催『AI エージェント 実践集中コース』に参加した

Google と Kaggle が提供し世界で 20 万人が登録した「5-Day Gen AI Intensive Course」をわずか 2 日間に凝縮し、これからのソフトウェア開発に求められる AI エージェントの開発スキルと、その基盤となるプロンプト エンジニアリングを実践的に学ぶ短期集中ワークショップです。
効果的なプロンプト設計や AI エージェント開発の基礎から実践までを体験、習得できます。

開催日 : 
・Day1：2025 年 6 月 7 日（土）10:00 - 19:00
・Day2：2025 年 6 月 8 日（日）10:00 - 17:00

この質問について、3人の異なる専門家が回答していると想像してください。
すべての専門家は、自分の思考の1つのステップを書き留め、
それをグループと共有します。
その後、すべての専門家は次のステップに進みます。以後同様です。
もし専門家の中に、いかなる時点で誤りに気づいた場合は、退場します。
質問は...

from google import genai

client = genai.Client(api_key=GOOGLE_API_KEY)


response = client.models.generate_content(
    model="gemini-2.0-flash",
    contents="Explain AI to me like I'm a kid.")

print(response.text)


chat = client.chats.create(model='gemini-2.0-flash', history=[])
response = chat.send_message('Hello! My name is Zlork.')

high_temp_config = types.GenerateContentConfig(temperature=2.0)


for _ in range(5):
  response = client.models.generate_content(
      model='gemini-2.0-flash',
      config=high_temp_config,
      contents='Pick a random colour... (respond in a single word)')

  if response.text:
    print(response.text, '-' * 25)

Magenta
 -------------------------
Turquoise
 -------------------------
Magenta
 -------------------------
Purple
 -------------------------
Azure.
 -------------------------

Azure
 -------------------------
Azure
 -------------------------
Azure
 -------------------------
Azure
 -------------------------
Azure
 -------------------------

import enum

class Sentiment(enum.Enum):
    POSITIVE = "positive"
    NEUTRAL = "neutral"
    NEGATIVE = "negative"


response = client.models.generate_content(
    model='gemini-2.0-flash',
    config=types.GenerateContentConfig(
        response_mime_type="text/x.enum",
        response_schema=Sentiment
    ),
    contents=zero_shot_prompt)

print(response.text)

import typing_extensions as typing

class PizzaOrder(typing.TypedDict):
    size: str
    ingredients: list[str]
    type: str


response = client.models.generate_content(
    model='gemini-2.0-flash',
    config=types.GenerateContentConfig(
        temperature=0.1,
        response_mime_type="application/json",
        response_schema=PizzaOrder,
    ),
    contents="Can I have a large dessert pizza with apple and chocolate")

print(response.text)

prompt = """When I was 4 years old, my partner was 3 times my age. Now, I
am 20 years old. How old is my partner? Return the answer directly."""

prompt = """When I was 4 years old, my partner was 3 times my age. Now,
I am 20 years old. How old is my partner? Let's think step by step."""

Here's how to solve this problem step-by-step:

Find the age difference: When you were 4, your partner was 3 * 4 = 12 years old.
Calculate the age difference: The age difference between you and your partner is 12 - 4 = 8 years.
Determine the partner's current age: Since the age difference remains constant, your partner is always 8 years older than you. So, now that you are 20, your partner is 20 + 8 = 28 years old.
Therefore, your partner is 28 years old.

model_instructions = """
Solve a question answering task with interleaving Thought, Action, Observation steps. Thought can reason about the current situation,
Observation is understanding relevant information from an Action's output and Action can be one of three types:
 (1) entity, which searches the exact entity on Wikipedia and returns the first paragraph if it exists. If not, it
     will return some similar entities to search and you can try to search the information from those topics.
 (2) keyword, which returns the next sentence containing keyword in the current context. This only does exact matches,
     so keep your searches short.
 (3) answer, which returns the answer and finishes the task.
"""

example1 = """Question
Musician and satirist Allie Goertz wrote a song about the "The Simpsons" character Milhouse, who Matt Groening named after who?

Thought 1
The question simplifies to "The Simpsons" character Milhouse is named after who. I only need to search Milhouse and find who it is named after.

Action 1
Milhouse

Observation 1
Milhouse Mussolini Van Houten is a recurring character in the Fox animated television series The Simpsons voiced by Pamela Hayden and created by Matt Groening.

# ...
"""

example2 = """..."""

question = """Question
Who was the youngest author listed on the transformers NLP paper?
"""


react_config = types.GenerateContentConfig(
    stop_sequences=["\nObservation"],
    system_instruction=model_instructions + example1 + example2,
)


react_chat = client.chats.create(
    model='gemini-2.0-flash',
    config=react_config,
)

resp = react_chat.send_message(question)
print(resp.text)

def list_tables() -> list[str]:
    """Retrieve the names of all tables in the database."""
    
    

db_tools = [list_tables, describe_table, execute_query]

instruction = """You are a helpful chatbot that can interact with an SQL database
for a computer store. You will take the users questions and turn them into SQL
queries using the tools available. Once you have the information you need, you will
answer the user's question using the data returned.

Use list_tables to see what tables are present, describe_table to understand the
schema, and execute_query to issue an SQL SELECT query."""

client = genai.Client(api_key=GOOGLE_API_KEY)


chat = client.chats.create(
    model="gemini-2.0-flash",
    config=types.GenerateContentConfig(
        system_instruction=instruction,
        tools=db_tools,
    ),
)

resp = chat.send_message("What is the cheapest product?")
print(f"\n{resp.text}")

 - DB CALL: execute_query(SELECT ProductName, Price FROM Products ORDER BY Price ASC LIMIT 1;)
 - DB CALL: describe_table(Products)
 - DB CALL: execute_query(SELECT product_name, price FROM Products ORDER BY price ASC LIMIT 1;)

The cheapest product is the Mouse, which costs $29.99.

from google.adk import Agent
from google.adk.tools import google_search

import sys
sys.path.append("..")
from callback_logging import log_query_to_model, log_model_response

root_agent = Agent(
    name="google_search_agent", 
    description="Google検索を使って質問に回答する", 
    model="gemini-2.0-flash-001", 
    instruction="あなたは専門の研究者。事実に固執する。", 
    before_model_callback=log_query_to_model, 
    after_model_callback=log_model_response, 
    tools=[google_search] 
)

adk run my_google_search_agent

from pydantic import BaseModel, Field

class CountryCapital(BaseModel):
    capital: str = Field(description="A country's capital.")


async def main():

    
    app_name = 'my_agent_app'
    user_id_1 = 'user1'

    
    root_agent = Agent(
        model=model_name,
        name="my_agent",
        instruction="Answer questions.",
        before_model_callback=log_query_to_model,
        after_model_callback=log_model_response,
        disallow_transfer_to_parent=True,
        disallow_transfer_to_peers=True,
        output_schema=CountryCapital,
    )

** User says: {'parts': [{'text': 'What is the capital of France?'}], 'role': 'user'}
** my_agent: The capital of France is Paris.

** User says: {'parts': [{'text': 'What is the capital of France?'}], 'role': 'user'}
** my_agent: {
  "capital": "Paris"
}

"""LLM Auditor for verifying & refining LLM-generated answers using the web."""


from google.adk.agents import SequentialAgent


from .sub_agents.critic import critic_agent
from .sub_agents.reviser import reviser_agent

import sys
sys.path.append("..")
from callback_logging import log_query_to_model, log_model_response

llm_auditor = SequentialAgent(
    name='llm_auditor',
    description=(
        'Evaluates LLM-generated answers, verifies actual accuracy using the'
        ' web, and refines the response to ensure alignment with real-world'
        ' knowledge.'
    ),
    sub_agents=[critic_agent, reviser_agent],
)

root_agent = llm_auditor

critic_agent = Agent(
    model='gemini-2.0-flash',
    name='critic_agent',
    instruction=prompt.CRITIC_PROMPT,
    tools=[google_search],
    after_model_callback=_render_reference,
)

attractions_planner = Agent(
    name="attractions_planner",
    model=model_name,
    description="Build a list of attractions to visit in a country.",
    instruction="""
        - Provide the user options for attractions to visit within their selected country.
        """,
    before_model_callback=log_query_to_model,
    after_model_callback=log_model_response,
    

    )

travel_brainstormer = Agent(
    name="travel_brainstormer",
    model=model_name,
    description="Help a user decide what country to visit.",
    instruction="""
        Provide a few suggestions of popular countries for travelers.
        
        Help a user identify their primary goals of travel:
        adventure, leisure, learning, shopping, or viewing art

        Identify countries that would make great destinations
        based on their priorities.
        """,
    before_model_callback=log_query_to_model,
    after_model_callback=log_model_response,
)

root_agent = Agent(
    name="steering",
    model=model_name,
    description="Start a user on a travel adventure.",
    instruction="""
        If they need help deciding, send them to
        'travel_brainstormer'.
        If they know what country they'd like to visit,
        send them to the 'attractions_planner'.
        """,
    generate_content_config=types.GenerateContentConfig(
        temperature=0,
    ),
    
    sub_agents=[travel_brainstormer, attractions_planner]
)

def save_attractions_to_state(
    tool_context: ToolContext,
    attractions: List[str]
) -> dict[str, str]:
    """Saves the list of attractions to state["attractions"].

    Args:
        attractions [str]: a list of strings to add to the list of attractions

    Returns:
        None
    """
    
    existing_attractions = tool_context.state.get("attractions", [])

    
    
    
    tool_context.state["attractions"] = existing_attractions + attractions

    
    return {"status": "success"}


attractions_planner = Agent(
    name="attractions_planner",
    model=model_name,
    description="Build a list of attractions to visit in a country.",
    instruction="""
        - Provide the user options for attractions to visit within their selected country.
        - When they reply, use your tool to save their selected attraction
        and then provide more possible attractions.
        - If they ask to view the list, provide a bulleted list of
        {{ attractions? }} and then suggest some more.
        """,
    before_model_callback=log_query_to_model,
    after_model_callback=log_model_response,
    
    tools=[save_attractions_to_state]
    )

- If they ask to view the list, provide a bulleted list of
        {{ attractions? }} and then suggest some more.

from google.adk.agents import LoopAgent

writers_room = LoopAgent(
    name="writers_room",
    description="Iterates through research and writing to improve a movie plot outline.",
    sub_agents=[
        researcher,
        screenwriter,
        critic
    ],
    max_iterations=3,
)

from google.adk.tools import exit_loop


    tools=[append_to_state, exit_loop],


    If the PLOT_OUTLINE does a good job with these questions, exit the writing loop with your 'exit_loop' tool.

from google.adk.agents import SequentialAgent

film_concept_team = SequentialAgent(
    name="film_concept_team",
    description="Write a film plot outline and save it as a text file.",
    sub_agents=[
        writers_room,
        file_writer
    ],
)