otel-instrumentation-google-adk 0.2.0

OpenTelemetry instrumentation for Google ADK (Agent Development Kit)

otel-instrumentation-google-adk

OpenTelemetry instrumentation for Google ADK (Agent Development Kit).

This instrumentor patches the Google ADK runner, agent, LLM, and tool layers to produce spans that follow the OpenTelemetry GenAI semantic conventions (gen_ai.* attributes). It is designed to work inside Google Agent Engine (GAE) where other telemetry libraries may be stripped.

Installation

pip install otel-instrumentation-google-adk

Usage

from otel_instrumentation_google_adk import GoogleADKInstrumentor

GoogleADKInstrumentor().instrument(tracer_provider=tracer_provider)

What gets instrumented

• Runner.run_async — creates an invocation span with input/output values, user ID, and session ID
• BaseAgent.run_async — creates an agent_run span per agent with agent name and output
• LLM calls — enriches spans with model name, input/output messages, token usage, temperature, and tool definitions
• Tool calls — enriches spans with tool name, description, input arguments, and output

Span attributes

Spans follow the OpenTelemetry GenAI semantic conventions:

Attribute	Description
gen_ai.operation.name	invoke_agent, chat, or execute_tool
gen_ai.system	gcp.vertex_ai
gen_ai.request.model	Model name from the LLM request
gen_ai.agent.name	Agent name
gen_ai.input.messages	JSON-serialized input messages
gen_ai.output.messages	JSON-serialized output messages
gen_ai.usage.input_tokens	Input token count
gen_ai.usage.output_tokens	Output token count (candidates + thoughts)
gen_ai.tool.name	Tool name
gen_ai.request.tools	JSON-serialized tool definitions

License

Apache-2.0