Google ADK

Overview

Google ADK (Agent Development Kit) is Google’s open-source framework for building, evaluating, and deploying AI agents. Confident AI allows you to trace and evaluate Google ADK agents in just a few lines of code.

The integration works via OpenTelemetry: instrument_google_adk() wraps the community-maintained openinference-instrumentation-google-adk package, which instruments every ADK agent, model call, and tool invocation as an OTel span. Confident AI’s OpenInference span interceptor then translates these into Confident AI spans and ships them in real time.

Tracing Quickstart

$
export CONFIDENT_OTEL_URL="https://eu.otel.confident-ai.com"

$
pip install -U deepeval google-adk openinference-instrumentation-google-adk opentelemetry-sdk opentelemetry-exporter-otlp-proto-http

$
export CONFIDENT_API_KEY="<your-confident-api-key>"

1
from google.adk.agents import Agent
2
from google.adk.runners import Runner
3
from google.adk.sessions import InMemorySessionService
4
from deepeval.integrations.google_adk import instrument_google_adk
5
6
instrument_google_adk()
7
8
root_agent = Agent(
9
    name="my_agent",
10
    model="gemini-2.0-flash",
11
    description="A helpful assistant.",
12
    instruction="Answer questions concisely.",
13
)
14
15
session_service = InMemorySessionService()
16
runner = Runner(agent=root_agent, app_name="my_app", session_service=session_service)

$
python main.py

Advanced Usage

Logging threads

Threads group related traces together and are useful for chat apps, agents, or any multi-turn interactions. You can learn more about threads here. Pass the thread_id to instrument_google_adk.

1
from google.adk.agents import Agent
2
from deepeval.integrations.google_adk import instrument_google_adk
3
4
instrument_google_adk(
5
    thread_id="session_abc123",
6
    user_id="user_1",
7
)
8
9
root_agent = Agent(
10
    name="my_agent",
11
    model="gemini-2.0-flash",
12
    description="A helpful assistant.",
13
    instruction="Answer questions concisely.",
14
)

Trace attributes

Other trace-level attributes can be passed to instrument_google_adk. All parameters are optional and apply to every trace produced while the instrumentation is active.

1
from deepeval.integrations.google_adk import instrument_google_adk
2
3
instrument_google_adk(
4
    name="Name of Trace",
5
    tags=["Tag 1", "Tag 2"],
6
    metadata={"Key": "Value"},
7
    user_id="user_1",
8
    thread_id="session_abc123",
9
    environment="production",
10
)

Logging prompts

If you are managing prompts on Confident AI and wish to log them, use next_llm_span to associate a Prompt with the next LLM span before invoking your agent.

1
from google.adk.agents import Agent
2
from google.adk.runners import Runner
3
from google.adk.sessions import InMemorySessionService
4
from deepeval.prompt import Prompt
5
from deepeval.tracing import next_llm_span
6
from deepeval.integrations.google_adk import instrument_google_adk
7
8
instrument_google_adk(environment="production")
9
10
prompt = Prompt(alias="<prompt-alias>")
11
prompt.pull(version="00.00.01")
12
13
root_agent = Agent(
14
    name="my_agent",
15
    model="gemini-2.0-flash",
16
    description="A helpful assistant.",
17
    instruction=prompt.interpolate(),
18
)
19
20
session_service = InMemorySessionService()
21
runner = Runner(agent=root_agent, app_name="my_app", session_service=session_service)
22
23
with next_llm_span(prompt=prompt):
24
    # run your agent here
25
    pass

Evals Usage

Online evals

You can run online evals on your Google ADK agent, which will run evaluations on all incoming traces on Confident AI’s servers. This approach is recommended if your agent is in production.

1
from google.adk.agents import Agent
2
from deepeval.integrations.google_adk import instrument_google_adk
3
4
instrument_google_adk(
5
    metric_collection="my_metric_collection",
6
)
7
8
root_agent = Agent(
9
    name="my_agent",
10
    model="gemini-2.0-flash",
11
    description="A helpful assistant.",
12
    instruction="Answer questions concisely.",
13
)

You can view eval results on Confident AI by clicking on the link printed in the console.