Learn extra at:
from google.adk.brokers import Agent
from google.adk.instruments import google_search
root_agent = Agent(
identify="search_assistant",
description="An agent that solutions questions augmented with internet searches.",
mannequin="gemini-2.0-flash",
instruction="Reply questions supplied by the consumer. Examine and distinction data gathered from Google with your personal data. In case you are given an announcement that isn't a query, reply, 'Please ask me a query.'",
instruments=[google_search]
)
Every time we need to create a definite agent, we arrange a subdirectory in our essential challenge listing and provides it a reputation (on this case, searchagent). This lets us have a number of brokers in a single challenge, which may run on their very own or interoperate.
The __init__.py file marks the listing as being an agent, by importing the precise agent code. The agent.py file units up the agent itself, as described by the Agent object.
Every Agent makes use of a mannequin API to interface with (right here, it’s gemini-2.0-flash). Our preliminary instructions to the agent, which prefix every enter from the consumer, are outlined in directions. Observe that these directions might be much more detailed than what we’re offering right here. The instruments part offers further tooling that can be utilized by the agent; google_search lets the agent use Google searches to reinforce its outcomes.
