Lab 25: Building an Observability System with Plugins Challenge

Goal

In this lab, you will build a comprehensive observability system for an agent using the ADK's Plugin System. You will implement separate plugins for metrics collection, alerting, and performance profiling.

Step 1: Create the Agent Project

1. Create the agent project:

   uv run adk create observability_agent

   When prompted, choose the Programmatic (Python script) option.

2. Navigate into the new directory:

   cd observability_agent

Step 2: Implement a Business Logic Plugin

Exercise: Open agent.py. We will start by creating an Alerting Plugin that detects consecutive errors.

# In agent.py (Starter Code)
from google.adk import Agent
from google.adk.apps import App
from google.adk.plugins import BasePlugin
from google.adk.events import Event

# A tool that can genuinely fail, so there's a real exception for the
# plugin to observe -- not just a simulated one.
def risky_operation(should_fail: bool) -> dict:
    """Performs an operation that can be made to fail, for testing error handling."""
    if should_fail:
        raise ValueError("Simulated failure!")
    return {"status": "success"}

class AlertingPlugin(BasePlugin):
    """A plugin that prints an alert after 3 consecutive request errors."""
    def __init__(self, name: str = 'alerting_plugin'):
        super().__init__(name)
        self.error_count = 0
        self._had_error_this_turn = False

    async def on_tool_error_callback(self, *, tool, tool_args, tool_context, error):
        # TODO: Implement alerting logic here.
        # 1. Mark that this turn had an error (self._had_error_this_turn = True).
        # 2. Increment self.error_count and print an alert, escalating to a
        #    critical alert once you hit the threshold.
        # 3. Return a dict (e.g. {"status": "error", "message": str(error)})
        #    so the agent recovers gracefully instead of crashing the run --
        #    returning None here would let the exception propagate.
        pass

    async def on_event_callback(self, *, event: Event, **kwargs):
        # TODO: When event.is_final_response() is True, if this turn had NO
        # error, reset self.error_count to 0 -- a clean turn means we've
        # recovered. Don't forget to reset self._had_error_this_turn too.
        pass

# --- Create a simple Agent ---
agent = Agent(
    name="monitored_agent",
    model="gemini-3.5-flash",
    instruction="You have a risky_operation tool. If the user says 'FAIL', call it with should_fail=True. Otherwise, call it with should_fail=False.",
    tools=[risky_operation],
)

# --- Register Plugin with App ---
app = App(
    name="observability_demo",
    root_agent=agent,
    plugins=[AlertingPlugin()]
)

Step 3: Configure Enterprise Telemetry

Exercise: Now, let's enable native Cloud Trace integration. In a real production environment, this would send data to Google Cloud. Here, we will configure the hooks.

# Add this to your agent.py imports
from google.adk.telemetry.google_cloud import get_gcp_exporters
from google.adk.telemetry.setup import maybe_set_otel_providers

# TODO: 1. Use get_gcp_exporters to enable cloud tracing and metrics.
# TODO: 2. Call maybe_set_otel_providers to register the hooks.

Step 4: Run and Verify

1. Launch the Dev UI:

   uv run adk web .

2. Test the Alerting Plugin: Send prompts and verify the console output.
3. Inspect the Trace Metadata: In the Dev UI, open the Trace tab. Notice how ADK 2.0 provides detailed information about each event, which is exactly what gets sent to Cloud Trace.

Having Trouble?

If you get stuck, you can find the complete, working code in the lab-solution.md file.

Lab Summary

You have successfully built a modular observability system using the ADK's Plugin System and OpenTelemetry. You have learned to:

• Create custom plugins by inheriting from BasePlugin.
• Implement on_tool_error_callback to intercept a tool's exception, recover gracefully, and still observe what failed.
• Use on_event_callback with event.is_final_response() to detect a clean turn (no errors) and reset your own tracking state.
• Configure enterprise-grade telemetry using Cloud Trace hooks.
• Register plugins and telemetry with the App object.

Self-Reflection Questions

• What is the main advantage of using the Plugin System for observability instead of adding logging and metrics code directly inside your agent and tool functions?
• How does ADK 2.0's node_info field improve your ability to debug complex multi-agent graphs?
• When should you use a custom Plugin versus native OpenTelemetry hooks?

---

🕵️ Hidden Solution 🕵️

Looking for the solution? Here's a hint (Base64 decode me): L2RvYy1hZGstdHJhaW5pbmcvbW9kdWxlMjUtb2JzZXJ2YWJpbGl0eS9sYWItc29sdXRpb24=

The direct link is: Lab Solution