내장 도구¶
이러한 내장 도구는 Google 검색이나 코드 실행기와 같이 에이전트에게 일반적인 기능을 제공하는 바로 사용 가능한 기능을 제공합니다. 예를 들어, 웹에서 정보를 검색해야 하는 에이전트는 추가 설정 없이 google_search 도구를 직접 사용할 수 있습니다.
사용 방법¶
- 가져오기: 도구 모듈에서 원하는 도구를 가져옵니다. Python에서는
agents.tools이고 Java에서는com.google.adk.tools입니다. - 구성: 필요한 매개변수가 있는 경우 제공하여 도구를 초기화합니다.
- 등록: 초기화된 도구를 에이전트의 도구 목록에 추가합니다.
에이전트에 추가되면, 에이전트는 사용자 프롬프트와 지침을 기반으로 도구 사용을 결정할 수 있습니다. 프레임워크는 에이전트가 도구를 호출할 때 도구 실행을 처리합니다. 중요: 이 페이지의 제한 사항 섹션을 확인하세요.
사용 가능한 내장 도구¶
참고: Java는 현재 Google 검색 및 코드 실행 도구만 지원합니다.
Google 검색¶
google_search 도구는 에이전트가 Google 검색을 사용하여 웹 검색을 수행할 수 있도록 합니다. google_search 도구는 Gemini 2 모델과만 호환됩니다.
google_search 도구 사용 시 추가 요구 사항
Google 검색으로 그라운딩을 사용하고 응답에서 검색 제안을 받는 경우, 프로덕션 및 애플리케이션에 검색 제안을 표시해야 합니다. Google 검색으로 그라운딩에 대한 자세한 내용은 Google AI Studio 또는 Vertex AI의 Google 검색으로 그라운딩 문서를 참조하세요. UI 코드(HTML)는 Gemini 응답에서 renderedContent로 반환되며, 정책에 따라 앱에 HTML을 표시해야 합니다.
from google.adk.agents import Agent
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.adk.tools import google_search
from google.genai import types
APP_NAME="google_search_agent"
USER_ID="user1234"
SESSION_ID="1234"
root_agent = Agent(
name="basic_search_agent",
model="gemini-2.0-flash",
description="Agent to answer questions using Google Search.",
instruction="I can answer your questions by searching the internet. Just ask me anything!",
# google_search is a pre-built tool which allows the agent to perform Google searches.
tools=[google_search]
)
# Session and Runner
session_service = InMemorySessionService()
session = session_service.create_session(app_name=APP_NAME, user_id=USER_ID, session_id=SESSION_ID)
runner = Runner(agent=root_agent, app_name=APP_NAME, session_service=session_service)
# Agent Interaction
def call_agent(query):
"""
Helper function to call the agent with a query.
"""
content = types.Content(role='user', parts=[types.Part(text=query)])
events = runner.run(user_id=USER_ID, session_id=SESSION_ID, new_message=content)
for event in events:
if event.is_final_response():
final_response = event.content.parts[0].text
print("Agent Response: ", final_response)
call_agent("what's the latest ai news?")
import com.google.adk.agents.BaseAgent;
import com.google.adk.agents.LlmAgent;
import com.google.adk.runner.Runner;
import com.google.adk.sessions.InMemorySessionService;
import com.google.adk.sessions.Session;
import com.google.adk.tools.GoogleSearchTool;
import com.google.common.collect.ImmutableList;
import com.google.genai.types.Content;
import com.google.genai.types.Part;
public class GoogleSearchAgentApp {
private static final String APP_NAME = "Google Search_agent";
private static final String USER_ID = "user1234";
private static final String SESSION_ID = "1234";
/**
* Calls the agent with the given query and prints the final response.
*
* @param runner The runner to use.
* @param query The query to send to the agent.
*/
public static void callAgent(Runner runner, String query) {
Content content =
Content.fromParts(Part.fromText(query));
InMemorySessionService sessionService = (InMemorySessionService) runner.sessionService();
Session session =
sessionService
.createSession(APP_NAME, USER_ID, /* state= */ null, SESSION_ID)
.blockingGet();
runner
.runAsync(session.userId(), session.id(), content)
.forEach(
event -> {
if (event.finalResponse()
&& event.content().isPresent()
&& event.content().get().parts().isPresent()
&& !event.content().get().parts().get().isEmpty()
&& event.content().get().parts().get().get(0).text().isPresent()) {
String finalResponse = event.content().get().parts().get().get(0).text().get();
System.out.println("Agent Response: " + finalResponse);
}
});
}
public static void main(String[] args) {
// Google Search is a pre-built tool which allows the agent to perform Google searches.
GoogleSearchTool googleSearchTool = new GoogleSearchTool();
BaseAgent rootAgent =
LlmAgent.builder()
.name("basic_search_agent")
.model("gemini-2.0-flash") // Ensure to use a Gemini 2.0 model for Google Search Tool
.description("Agent to answer questions using Google Search.")
.instruction(
"I can answer your questions by searching the internet. Just ask me anything!")
.tools(ImmutableList.of(googleSearchTool))
.build();
// Session and Runner
InMemorySessionService sessionService = new InMemorySessionService();
Runner runner = new Runner(rootAgent, APP_NAME, null, sessionService);
// Agent Interaction
callAgent(runner, "what's the latest ai news?");
}
}
코드 실행¶
built_in_code_execution 도구는 에이전트가 코드를 실행할 수 있게 하며, 특히 Gemini 2 모델을 사용할 때 그렇습니다. 이를 통해 모델은 계산, 데이터 조작 또는 작은 스크립트 실행과 같은 작업을 수행할 수 있습니다.
# Copyright 2025 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import asyncio
from google.adk.agents import LlmAgent
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.adk.code_executors import BuiltInCodeExecutor
from google.genai import types
AGENT_NAME = "calculator_agent"
APP_NAME = "calculator"
USER_ID = "user1234"
SESSION_ID = "session_code_exec_async"
GEMINI_MODEL = "gemini-2.0-flash"
# Agent Definition
code_agent = LlmAgent(
name=AGENT_NAME,
model=GEMINI_MODEL,
executor=[BuiltInCodeExecutor],
instruction="""You are a calculator agent.
When given a mathematical expression, write and execute Python code to calculate the result.
Return only the final numerical result as plain text, without markdown or code blocks.
""",
description="Executes Python code to perform calculations.",
)
# Session and Runner
session_service = InMemorySessionService()
session = session_service.create_session(
app_name=APP_NAME, user_id=USER_ID, session_id=SESSION_ID
)
runner = Runner(agent=code_agent, app_name=APP_NAME, session_service=session_service)
# Agent Interaction (Async)
async def call_agent_async(query):
content = types.Content(role="user", parts=[types.Part(text=query)])
print(f"\n--- Running Query: {query} ---")
final_response_text = "No final text response captured."
try:
# Use run_async
async for event in runner.run_async(
user_id=USER_ID, session_id=SESSION_ID, new_message=content
):
print(f"Event ID: {event.id}, Author: {event.author}")
# --- Check for specific parts FIRST ---
has_specific_part = False
if event.content and event.content.parts:
for part in event.content.parts: # Iterate through all parts
if part.executable_code:
# Access the actual code string via .code
print(
f" Debug: Agent generated code:\n```python\n{part.executable_code.code}\n```"
)
has_specific_part = True
elif part.code_execution_result:
# Access outcome and output correctly
print(
f" Debug: Code Execution Result: {part.code_execution_result.outcome} - Output:\n{part.code_execution_result.output}"
)
has_specific_part = True
# Also print any text parts found in any event for debugging
elif part.text and not part.text.isspace():
print(f" Text: '{part.text.strip()}'")
# Do not set has_specific_part=True here, as we want the final response logic below
# --- Check for final response AFTER specific parts ---
# Only consider it final if it doesn't have the specific code parts we just handled
if not has_specific_part and event.is_final_response():
if (
event.content
and event.content.parts
and event.content.parts[0].text
):
final_response_text = event.content.parts[0].text.strip()
print(f"==> Final Agent Response: {final_response_text}")
else:
print("==> Final Agent Response: [No text content in final event]")
except Exception as e:
print(f"ERROR during agent run: {e}")
print("-" * 30)
# Main async function to run the examples
async def main():
await call_agent_async("Calculate the value of (5 + 7) * 3")
await call_agent_async("What is 10 factorial?")
# Execute the main async function
try:
asyncio.run(main())
except RuntimeError as e:
# Handle specific error when running asyncio.run in an already running loop (like Jupyter/Colab)
if "cannot be called from a running event loop" in str(e):
print("\nRunning in an existing event loop (like Colab/Jupyter).")
print("Please run `await main()` in a notebook cell instead.")
# If in an interactive environment like a notebook, you might need to run:
# await main()
else:
raise e # Re-raise other runtime errors
import com.google.adk.agents.BaseAgent;
import com.google.adk.agents.LlmAgent;
import com.google.adk.runner.Runner;
import com.google.adk.sessions.InMemorySessionService;
import com.google.adk.sessions.Session;
import com.google.adk.tools.BuiltInCodeExecutionTool;
import com.google.common.collect.ImmutableList;
import com.google.genai.types.Content;
import com.google.genai.types.Part;
public class CodeExecutionAgentApp {
private static final String AGENT_NAME = "calculator_agent";
private static final String APP_NAME = "calculator";
private static final String USER_ID = "user1234";
private static final String SESSION_ID = "session_code_exec_sync";
private static final String GEMINI_MODEL = "gemini-2.0-flash";
/**
* Calls the agent with a query and prints the interaction events and final response.
*
* @param runner The runner instance for the agent.
* @param query The query to send to the agent.
*/
public static void callAgent(Runner runner, String query) {
Content content =
Content.builder().role("user").parts(ImmutableList.of(Part.fromText(query))).build();
InMemorySessionService sessionService = (InMemorySessionService) runner.sessionService();
Session session =
sessionService
.createSession(APP_NAME, USER_ID, /* state= */ null, SESSION_ID)
.blockingGet();
System.out.println("\n--- Running Query: " + query + " ---");
final String[] finalResponseText = {"No final text response captured."};
try {
runner
.runAsync(session.userId(), session.id(), content)
.forEach(
event -> {
System.out.println("Event ID: " + event.id() + ", Author: " + event.author());
boolean hasSpecificPart = false;
if (event.content().isPresent() && event.content().get().parts().isPresent()) {
for (Part part : event.content().get().parts().get()) {
if (part.executableCode().isPresent()) {
System.out.println(
" Debug: Agent generated code:\n```python\n"
+ part.executableCode().get().code()
+ "\n```");
hasSpecificPart = true;
} else if (part.codeExecutionResult().isPresent()) {
System.out.println(
" Debug: Code Execution Result: "
+ part.codeExecutionResult().get().outcome()
+ " - Output:\n"
+ part.codeExecutionResult().get().output());
hasSpecificPart = true;
} else if (part.text().isPresent() && !part.text().get().trim().isEmpty()) {
System.out.println(" Text: '" + part.text().get().trim() + "'");
}
}
}
if (!hasSpecificPart && event.finalResponse()) {
if (event.content().isPresent()
&& event.content().get().parts().isPresent()
&& !event.content().get().parts().get().isEmpty()
&& event.content().get().parts().get().get(0).text().isPresent()) {
finalResponseText[0] =
event.content().get().parts().get().get(0).text().get().trim();
System.out.println("==> Final Agent Response: " + finalResponseText[0]);
} else {
System.out.println(
"==> Final Agent Response: [No text content in final event]");
}
}
});
} catch (Exception e) {
System.err.println("ERROR during agent run: " + e.getMessage());
e.printStackTrace();
}
System.out.println("------------------------------");
}
public static void main(String[] args) {
BuiltInCodeExecutionTool codeExecutionTool = new BuiltInCodeExecutionTool();
BaseAgent codeAgent =
LlmAgent.builder()
.name(AGENT_NAME)
.model(GEMINI_MODEL)
.tools(ImmutableList.of(codeExecutionTool))
.instruction(
"""
You are a calculator agent.
When given a mathematical expression, write and execute Python code to calculate the result.
Return only the final numerical result as plain text, without markdown or code blocks.
""")
.description("Executes Python code to perform calculations.")
.build();
InMemorySessionService sessionService = new InMemorySessionService();
Runner runner = new Runner(codeAgent, APP_NAME, null, sessionService);
callAgent(runner, "Calculate the value of (5 + 7) * 3");
callAgent(runner, "What is 10 factorial?");
}
}
Vertex AI 검색¶
vertex_ai_search_tool은 Google Cloud의 Vertex AI Search를 사용하여 에이전트가 비공개로 구성된 데이터 저장소(예: 내부 문서, 회사 정책, 지식 기반)를 검색할 수 있도록 합니다. 이 내장 도구는 구성 중에 특정 데이터 저장소 ID를 제공해야 합니다.
import asyncio
from google.adk.agents import LlmAgent
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.genai import types
from google.adk.tools import VertexAiSearchTool
# Replace with your actual Vertex AI Search Datastore ID
# Format: projects/<PROJECT_ID>/locations/<LOCATION>/collections/default_collection/dataStores/<DATASTORE_ID>
# e.g., "projects/12345/locations/us-central1/collections/default_collection/dataStores/my-datastore-123"
YOUR_DATASTORE_ID = "YOUR_DATASTORE_ID_HERE"
# Constants
APP_NAME_VSEARCH = "vertex_search_app"
USER_ID_VSEARCH = "user_vsearch_1"
SESSION_ID_VSEARCH = "session_vsearch_1"
AGENT_NAME_VSEARCH = "doc_qa_agent"
GEMINI_2_FLASH = "gemini-2.0-flash"
# Tool Instantiation
# You MUST provide your datastore ID here.
vertex_search_tool = VertexAiSearchTool(data_store_id=YOUR_DATASTORE_ID)
# Agent Definition
doc_qa_agent = LlmAgent(
name=AGENT_NAME_VSEARCH,
model=GEMINI_2_FLASH, # Requires Gemini model
tools=[vertex_search_tool],
instruction=f"""You are a helpful assistant that answers questions based on information found in the document store: {YOUR_DATASTORE_ID}.
Use the search tool to find relevant information before answering.
If the answer isn't in the documents, say that you couldn't find the information.
""",
description="Answers questions using a specific Vertex AI Search datastore.",
)
# Session and Runner Setup
session_service_vsearch = InMemorySessionService()
runner_vsearch = Runner(
agent=doc_qa_agent, app_name=APP_NAME_VSEARCH, session_service=session_service_vsearch
)
session_vsearch = session_service_vsearch.create_session(
app_name=APP_NAME_VSEARCH, user_id=USER_ID_VSEARCH, session_id=SESSION_ID_VSEARCH
)
# Agent Interaction Function
async def call_vsearch_agent_async(query):
print("\n--- Running Vertex AI Search Agent ---")
print(f"Query: {query}")
if "YOUR_DATASTORE_ID_HERE" in YOUR_DATASTORE_ID:
print("Skipping execution: Please replace YOUR_DATASTORE_ID_HERE with your actual datastore ID.")
print("-" * 30)
return
content = types.Content(role='user', parts=[types.Part(text=query)])
final_response_text = "No response received."
try:
async for event in runner_vsearch.run_async(
user_id=USER_ID_VSEARCH, session_id=SESSION_ID_VSEARCH, new_message=content
):
# Like Google Search, results are often embedded in the model's response.
if event.is_final_response() and event.content and event.content.parts:
final_response_text = event.content.parts[0].text.strip()
print(f"Agent Response: {final_response_text}")
# You can inspect event.grounding_metadata for source citations
if event.grounding_metadata:
print(f" (Grounding metadata found with {len(event.grounding_metadata.grounding_attributions)} attributions)")
except Exception as e:
print(f"An error occurred: {e}")
print("Ensure your datastore ID is correct and the service account has permissions.")
print("-" * 30)
# --- Run Example ---
async def run_vsearch_example():
# Replace with a question relevant to YOUR datastore content
await call_vsearch_agent_async("Summarize the main points about the Q2 strategy document.")
await call_vsearch_agent_async("What safety procedures are mentioned for lab X?")
# Execute the example
# await run_vsearch_example()
# Running locally due to potential colab asyncio issues with multiple awaits
try:
asyncio.run(run_vsearch_example())
except RuntimeError as e:
if "cannot be called from a running event loop" in str(e):
print("Skipping execution in running event loop (like Colab/Jupyter). Run locally.")
else:
raise e
다른 도구와 함께 내장 도구 사용하기¶
다음 코드 샘플은 여러 내장 도구를 사용하거나 여러 에이전트를 사용하여 다른 도구와 함께 내장 도구를 사용하는 방법을 보여줍니다:
from google.adk.tools import agent_tool
from google.adk.agents import Agent
from google.adk.tools import google_search
from google.adk.code_executors import BuiltInCodeExecutor
search_agent = Agent(
model='gemini-2.0-flash',
name='SearchAgent',
instruction="""
당신은 Google 검색 전문가입니다.
""",
tools=[google_search],
)
coding_agent = Agent(
model='gemini-2.0-flash',
name='CodeAgent',
instruction="""
당신은 코드 실행 전문가입니다.
""",
code_executor=[BuiltInCodeExecutor],
)
root_agent = Agent(
name="RootAgent",
model="gemini-2.0-flash",
description="루트 에이전트",
tools=[agent_tool.AgentTool(agent=search_agent), agent_tool.AgentTool(agent=coding_agent)],
)
import com.google.adk.agents.BaseAgent;
import com.google.adk.agents.LlmAgent;
import com.google.adk.tools.AgentTool;
import com.google.adk.tools.BuiltInCodeExecutionTool;
import com.google.adk.tools.GoogleSearchTool;
import com.google.common.collect.ImmutableList;
public class NestedAgentApp {
private static final String MODEL_ID = "gemini-2.0-flash";
public static void main(String[] args) {
// 검색 에이전트 정의
LlmAgent searchAgent =
LlmAgent.builder()
.model(MODEL_ID)
.name("SearchAgent")
.instruction("당신은 Google 검색 전문가입니다.")
.tools(new GoogleSearchTool()) // GoogleSearchTool 인스턴스화
.build();
// 코딩 에이전트 정의
LlmAgent codingAgent =
LlmAgent.builder()
.model(MODEL_ID)
.name("CodeAgent")
.instruction("당신은 코드 실행 전문가입니다.")
.tools(new BuiltInCodeExecutionTool()) // BuiltInCodeExecutionTool 인스턴스화
.build();
// 루트 에이전트 정의, AgentTool.create()를 사용하여 검색 에이전트와 코딩 에이전트를 래핑
BaseAgent rootAgent =
LlmAgent.builder()
.name("RootAgent")
.model(MODEL_ID)
.description("루트 에이전트")
.tools(
AgentTool.create(searchAgent), // create 메서드 사용
AgentTool.create(codingAgent) // create 메서드 사용
)
.build();
// 참고: 이 샘플은 에이전트 정의만 보여줍니다.
// 이 에이전트들을 실행하려면 Runner와 SessionService와 통합해야 합니다.
// 이전 예제와 유사하게.
System.out.println("에이전트가 성공적으로 정의되었습니다:");
System.out.println(" 루트 에이전트: " + rootAgent.name());
System.out.println(" 검색 에이전트 (중첩됨): " + searchAgent.name());
System.out.println(" 코드 에이전트 (중첩됨): " + codingAgent.name());
}
}
제한 사항¶
Warning
현재 각 루트 에이전트 또는 단일 에이전트당 하나의 내장 도구만 지원됩니다. 동일한 에이전트에서 다른 유형의 도구를 사용할 수 없습니다.
예를 들어, 단일 에이전트 내에서 다른 도구와 함께 내장 도구를 사용하는 다음 접근 방식은 현재 지원되지 않습니다:
Warning
내장 도구는 하위 에이전트 내에서 사용할 수 없습니다.
예를 들어, 하위 에이전트 내에서 내장 도구를 사용하는 다음 접근 방식은 현재 지원되지 않습니다:
search_agent = Agent(
model='gemini-2.0-flash',
name='SearchAgent',
instruction="""
당신은 Google 검색 전문가입니다.
""",
tools=[google_search],
)
coding_agent = Agent(
model='gemini-2.0-flash',
name='CodeAgent',
instruction="""
당신은 코드 실행 전문가입니다.
""",
executor=[BuiltInCodeExecutor],
)
root_agent = Agent(
name="RootAgent",
model="gemini-2.0-flash",
description="루트 에이전트",
sub_agents=[
search_agent,
coding_agent
],
)
LlmAgent searchAgent =
LlmAgent.builder()
.model("gemini-2.0-flash")
.name("SearchAgent")
.instruction("당신은 Google 검색 전문가입니다.")
.tools(new GoogleSearchTool())
.build();
LlmAgent codingAgent =
LlmAgent.builder()
.model("gemini-2.0-flash")
.name("CodeAgent")
.instruction("당신은 코드 실행 전문가입니다.")
.tools(new BuiltInCodeExecutionTool())
.build();
LlmAgent rootAgent =
LlmAgent.builder()
.name("RootAgent")
.model("gemini-2.0-flash")
.description("루트 에이전트")
.subAgents(searchAgent, codingAgent) // 지원되지 않음, 하위 에이전트가 내장 도구를 사용하기 때문.
.build();