LoongSuite Google ADK Instrumentation

Google ADK (Agent Development Kit) Python Agent provides comprehensive observability for Google ADK applications using OpenTelemetry.

Features

• ✅ Automatic Instrumentation: Zero-code integration via loongsuite-instrument
• ✅ Manual Instrumentation: Programmatic control via GoogleAdkInstrumentor
• ✅ GenAI Semantic Conventions: Full compliance with OpenTelemetry GenAI standards
• ✅ Comprehensive Spans: invoke_agent, chat, execute_tool
• ✅ Standard Metrics: Operation duration and token usage
• ✅ Content Capture: Optional message and response content capture
• ✅ Google ADK native instrumentation Compatible: Works seamlessly with ADK native instrumentation

Quick Start

# Step 1: install LoongSuite distro
pip install loongsuite-distro

# Step 2 (Option C): install instrumentation from PyPI
pip install loongsuite-instrumentation-google-adk

# App dependencies
pip install google-adk litellm

# Configure
export DASHSCOPE_API_KEY=your-api-key

# Run with auto instrumentation
loongsuite-instrument \
  --traces_exporter console \
  --service_name my-adk-app \
  python your_app.py

For details on LoongSuite and Jaeger setup, refer to LoongSuite Documentation.

Installing Google ADK Instrumentation

# Step 1: install LoongSuite distro
pip install loongsuite-distro

# Step 2 (Option C): install this instrumentation from PyPI
pip install loongsuite-instrumentation-google-adk

# Google ADK and LLM Dependencies
pip install google-adk>=0.1.0
pip install litellm

# Demo Application Dependencies (optional, only if running examples)
pip install fastapi uvicorn pydantic

Collect Data

Here's a simple demonstration of Google ADK instrumentation. The demo uses:

• A Google ADK application that demonstrates agent interactions with multiple tools

Running the Demo

Note: The demo uses DashScope (Alibaba Cloud LLM service) by default. You need to set the DASHSCOPE_API_KEY environment variable.

Option 1: Using LoongSuite auto instrumentation

# Set your DashScope API key
export DASHSCOPE_API_KEY=your-dashscope-api-key

# Enable content capture (optional, for debugging)
export OTEL_SEMCONV_STABILITY_OPT_IN=gen_ai_latest_experimental
export OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=SPAN_ONLY

# Run with loongsuite instrumentation
loongsuite-instrument \
  --traces_exporter console \
  --service_name demo-google-adk \
  python examples/main.py

Option 2: Export to Jaeger

# Set your DashScope API key
export DASHSCOPE_API_KEY=your-dashscope-api-key

# Configure OTLP exporter
export OTEL_SEMCONV_STABILITY_OPT_IN=gen_ai_latest_experimental
export OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=SPAN_ONLY
export OTEL_TRACES_EXPORTER=otlp
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317
export OTEL_EXPORTER_OTLP_PROTOCOL=grpc

# Run the application
loongsuite-instrument \
  --service_name demo-google-adk \
  python examples/main.py

Option 3: Local otel-gui smoke scenarios

examples/otelgui_smoke.py produces real non-streaming, SSE streaming, and concurrent Google ADK calls for local trace validation.

export DASHSCOPE_API_KEY=your-dashscope-api-key
export OTEL_EXPORTER_OTLP_ENDPOINT=http://127.0.0.1:5173
export OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
export OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=http://127.0.0.1:5173/v1/traces
export OTEL_SEMCONV_STABILITY_OPT_IN=gen_ai_latest_experimental
export OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=SPAN_ONLY
export OTEL_SERVICE_NAME=loongsuite-google-adk-smoke
export GOOGLE_ADK_SMOKE_CONFIGURE_OTLP=1
export GOOGLE_ADK_SMOKE_DISABLE_NATIVE_AGENT_SPAN=1

python examples/otelgui_smoke.py --scenario all

GOOGLE_ADK_SMOKE_DISABLE_NATIVE_AGENT_SPAN=1 uses a private ADK telemetry monkey-patch during smoke validation to remove ADK's native wrapper span, making the LoongSuite GenAI span tree easier to inspect in otel-gui. Keep it limited to local smoke tests because private ADK internals may change.

When using the local loongsuite-otelgui-plugin-verify helper, select the GenAI util agent trace explicitly because Google ADK also emits an invocation trace:

python /path/to/run_loongsuite_plugin_smoke.py \
  --repo-root /path/to/loongsuite-python \
  --base-url http://127.0.0.1:5173 \
  --service-name loongsuite-google-adk-non-stream \
  --root-span-contains invoke_agent \
  --capture-message-content SPAN_ONLY \
  --expect-span-kind AGENT \
  --expect-span-kind LLM \
  --expect-span-kind TOOL \
  --expect-content \
  --env GOOGLE_ADK_SMOKE_CONFIGURE_OTLP=1 \
  --env GOOGLE_ADK_SMOKE_DISABLE_NATIVE_AGENT_SPAN=1 \
  --env OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=http://127.0.0.1:5173/v1/traces \
  --run "python examples/otelgui_smoke.py --scenario non-stream"

Expected Results

The instrumentation will generate traces showing the Google ADK operations:

Tool Execution Span Example

{
    "name": "execute_tool get_current_time",
    "context": {
        "trace_id": "xxx",
        "span_id": "xxx",
        "trace_state": "[]"
    },
    "kind": "SpanKind.INTERNAL",
    "parent_id": "xxx",
    "start_time": "2025-10-23T06:36:33.858459Z",
    "end_time": "2025-10-23T06:36:33.858779Z",
    "status": {
        "status_code": "UNSET"
    },
    "attributes": {
        "gen_ai.operation.name": "execute_tool",
        "gen_ai.span.kind": "TOOL",
        "gen_ai.tool.name": "get_current_time",
        "gen_ai.tool.description": "xxx",
        "gen_ai.tool.call.arguments": "{xxx}",
        "gen_ai.tool.call.result": "{xxx}"
    },
    "events": [],
    "links": [],
    "resource": {
        "attributes": {
            "telemetry.sdk.language": "python",
            "telemetry.sdk.name": "opentelemetry",
            "telemetry.sdk.version": "1.37.0",
            "service.name": "demo-google-adk"
        },
        "schema_url": ""
    }
}

LLM Chat Span Example

{
    "name": "chat qwen-max",
    "kind": "SpanKind.CLIENT",
    "attributes": {
        "gen_ai.operation.name": "chat",
        "gen_ai.span.kind": "LLM",
        "gen_ai.request.model": "qwen-max",
        "gen_ai.response.model": "qwen-max",
        "gen_ai.usage.input_tokens": 150,
        "gen_ai.usage.output_tokens": 45
    }
}

Agent Invocation Span Example

{
    "name": "invoke_agent ToolAgent",
    "kind": "SpanKind.CLIENT",
    "attributes": {
        "gen_ai.operation.name": "invoke_agent",
        "gen_ai.span.kind": "AGENT",
        "gen_ai.agent.name": "ToolAgent",
        "gen_ai.input.messages": "[{\"role\": \"user\", \"parts\": [{\"type\": \"text\", \"content\": \"What time is it?\"}]}]",
        "gen_ai.output.messages": "[{\"role\": \"assistant\", \"parts\": [{\"type\": \"text\", \"content\": \"The current time is 2025-11-27 14:36:33\"}]}]"
    }
}

Viewing in Jaeger

After setting up Jaeger, you can visualize the complete trace hierarchy in the Jaeger UI, showing the relationships between Runner, Agent, LLM, and Tool spans

Configuration

Environment Variables

The following environment variables can be used to configure the Google ADK instrumentation:

Variable	Description	Default
OTEL_SEMCONV_STABILITY_OPT_IN	Enable latest experimental GenAI semantic conventions	-
OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT	Capture message content in traces (NO_CONTENT, SPAN_ONLY, SPAN_AND_EVENT)	NO_CONTENT
DASHSCOPE_API_KEY	DashScope API key (required for demo)	-

Programmatic Configuration

You can also configure the instrumentation programmatically:

from opentelemetry.instrumentation.google_adk import GoogleAdkInstrumentor

# Configure the instrumentor
instrumentor = GoogleAdkInstrumentor()

# Enable instrumentation with custom configuration
instrumentor.instrument(
    tracer_provider=your_tracer_provider,
    meter_provider=your_meter_provider
)

Supported Features

Traces

The Google ADK instrumentation automatically creates traces for:

• Agent Runs: Complete agent execution cycles
• Tool Calls: Individual tool invocations
• Model Interactions: LLM requests and responses
• Session Management: User session tracking
• Error Handling: Exception and error tracking

Metrics

The instrumentation follows the OpenTelemetry GenAI Semantic Conventions for Metrics and provides the following standard client metrics:

1. gen_ai.client.operation.duration (Histogram)

Records the duration of GenAI operations in seconds.

Instrument Type: Histogram
Unit: s (seconds)
Status: Development

Required Attributes:

• gen_ai.operation.name: Operation being performed (e.g., chat, invoke_agent, execute_tool)
• gen_ai.provider.name: Provider name (e.g., google_adk)

Conditionally Required Attributes:

• error.type: Error type (only if operation ended in error)
• gen_ai.request.model: Model name (if available)

Recommended Attributes:

• gen_ai.response.model: Response model name
• server.address: Server address
• server.port: Server port

Example Values:

• LLM operation: gen_ai.operation.name="chat", gen_ai.request.model="gemini-pro", duration=1.5s
• Agent operation: gen_ai.operation.name="invoke_agent", gen_ai.request.model="math_tutor", duration=2.3s
• Tool operation: gen_ai.operation.name="execute_tool", gen_ai.request.model="calculator", duration=0.5s

2. gen_ai.client.token.usage (Histogram)

Records the number of tokens used in GenAI operations.

Instrument Type: Histogram
Unit: {token}
Status: Development

Required Attributes:

• gen_ai.operation.name: Operation being performed
• gen_ai.provider.name: Provider name
• gen_ai.token.type: Token type (input or output)

Conditionally Required Attributes:

• gen_ai.request.model: Model name (if available)

Recommended Attributes:

• gen_ai.response.model: Response model name
• server.address: Server address
• server.port: Server port

Example Values:

• Input tokens: gen_ai.token.type="input", gen_ai.request.model="gemini-pro", count=100
• Output tokens: gen_ai.token.type="output", gen_ai.request.model="gemini-pro", count=50

Note: These metrics use Histogram instrument type (not Counter) and follow the standard OpenTelemetry GenAI semantic conventions. All other metrics (like genai.agent.runs.count, etc.) are non-standard and have been removed to ensure compliance with the latest OTel specifications.

Semantic Conventions

This instrumentation follows the OpenTelemetry GenAI semantic conventions:

• GenAI Spans
• GenAI Agent Spans
• GenAI Metrics

Troubleshooting

Common Issues

1. Module Import Error: If you encounter No module named 'google.adk.runners', ensure that google-adk is properly installed:

   pip install google-adk>=0.1.0
   

2. DashScope API Error: If you see authentication errors, verify your API key is correctly set:

   export DASHSCOPE_API_KEY=your-api-key
   # Verify it's set
   echo $DASHSCOPE_API_KEY
   

3. Instrumentation Not Working:

   • Check that the instrumentation is enabled and the Google ADK application is using the Runner class
   • Verify you see the log message: Plugin 'opentelemetry_adk_observability' registered
   • For manual instrumentation, ensure you call GoogleAdkInstrumentor().instrument() before creating the Runner

4. Missing Traces:

   • Verify that the OpenTelemetry exporters are properly configured
   • Check the OTEL_TRACES_EXPORTER environment variable is set (e.g., console, otlp)
   • For OTLP exporter, ensure the endpoint is reachable

References

• OpenTelemetry Project
• Google ADK Documentation
• GenAI Semantic Conventions