Supported Operations

• Synchronous and asynchronous chat via the Converse API (client.converse)

• Synchronous and asynchronous streaming chat via the ConverseStream API (client.converse_stream)

• Synchronous and asynchronous model invocation via the InvokeModel API (client.invoke_model)

• Synchronous and asynchronous streaming model invocation via the InvokeModelWithResponseStream API (client.invoke_model_with_response_stream)

Installation

pip install opentelemetry-instrumentation-genai-bedrock

Usage

import boto3
from opentelemetry.instrumentation.genai.bedrock import BedrockInstrumentor

# Enable instrumentation
BedrockInstrumentor().instrument()

# Use Bedrock runtime client normally
client = boto3.client("bedrock-runtime", region_name="us-east-1")
response = client.converse(
    modelId="amazon.nova-micro-v1:0",
    messages=[{"role": "user", "content": [{"text": "Hello, Bedrock!"}]}],
)

Configuration

By default, prompts and completions are not captured. To capture message content, set the environment variable OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT to one of NO_CONTENT, SPAN_ONLY, EVENT_ONLY, or SPAN_AND_EVENT:

export OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=SPAN_ONLY

Prompts and completions can also be redirected via a completion hook by setting OTEL_INSTRUMENTATION_GENAI_COMPLETION_HOOK or by passing instrument(completion_hook=...).

References

• OpenTelemetry Project

• OpenTelemetry GenAI semantic conventions

• Amazon Bedrock Documentation

• Boto3 Documentation