Native instrumentation with OpenTelemetry by rodrigobr-msft · Pull Request #329 · microsoft/Agents-for-python

rodrigobr-msft · 2026-02-24T17:38:37Z

No description provided.

…into users/robrandao/otel-sample

Copilot

Pull request overview

This pull request introduces native OpenTelemetry instrumentation to the Microsoft Agents SDK, enabling comprehensive observability through traces, metrics, and logs. The changes add telemetry hooks at key points in the agent lifecycle including turn processing, adapter operations, and storage interactions, while also providing a test sample demonstrating the integration.

Changes:

Added OpenTelemetry integration to core SDK with AgentTelemetry class for instrumentation
Integrated telemetry tracking into agent turn processing, adapter operations, and storage methods
Created test sample demonstrating OpenTelemetry configuration and usage with aiohttp
Enhanced testing framework to support Activity template aliasing (from/from_property compatibility)

Reviewed changes

Copilot reviewed 21 out of 24 changed files in this pull request and generated 27 comments.

Show a summary per file

File	Description
test_samples/otel/src/telemetry.py	OpenTelemetry configuration helper for test sample
test_samples/otel/src/start_server.py	Sample aiohttp server with telemetry integration
test_samples/otel/src/requirements.txt	Dependencies for OpenTelemetry sample
test_samples/otel/src/main.py	Entry point configuring telemetry before agent initialization
test_samples/otel/src/env.TEMPLATE	Environment configuration template with OTLP settings
test_samples/otel/src/agent_metric.py	Sample-specific metrics wrapper (has import name mismatch)
test_samples/otel/src/agent.py	Sample agent with manual telemetry instrumentation
libraries/microsoft-agents-hosting-core/setup.py	Added OpenTelemetry dependencies
libraries/microsoft-agents-hosting-core/microsoft_agents/hosting/core/observability/_agent_telemetry.py	Core telemetry infrastructure with context managers for instrumentation
libraries/microsoft-agents-hosting-core/microsoft_agents/hosting/core/observability/init.py	Exports AgentTelemetry and agent_telemetry singleton
libraries/microsoft-agents-hosting-core/microsoft_agents/hosting/core/app/agent_application.py	Wraps turn processing with telemetry
libraries/microsoft-agents-hosting-core/microsoft_agents/hosting/core/storage/storage.py	Adds telemetry to base storage operations
libraries/microsoft-agents-hosting-core/microsoft_agents/hosting/core/storage/memory_storage.py	Adds telemetry to in-memory storage implementation
libraries/microsoft-agents-storage-cosmos/microsoft_agents/storage/cosmos/cosmos_db_storage.py	Minor whitespace change
libraries/microsoft-agents-hosting-fastapi/microsoft_agents/hosting/fastapi/cloud_adapter.py	Wraps adapter processing with telemetry
libraries/microsoft-agents-hosting-aiohttp/microsoft_agents/hosting/aiohttp/cloud_adapter.py	Wraps adapter processing with telemetry
dev/tests/sdk/observability/test_observability.py	Tests for telemetry functionality
dev/tests/scenarios/quickstart.py	Renamed init_app to init_agent
dev/tests/scenarios/init.py	Updated to use renamed init_agent function
dev/microsoft-agents-testing/tests/core/fluent/test_model_template.py	Added tests for from/from_property aliasing
dev/microsoft-agents-testing/microsoft_agents/testing/core/fluent/utils.py	Added from/from_property aliasing logic
dev/microsoft-agents-testing/microsoft_agents/testing/core/fluent/model_template.py	Integrated aliasing into template classes

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-24T17:47:01Z

test_samples/otel/src/telemetry.py

+from microsoft_agents.hosting.core import TurnContext
+


Unused import: TurnContext is imported but never used in this file. Consider removing it to keep imports clean.

Suggested change

from microsoft_agents.hosting.core import TurnContext

Copilot · 2026-02-24T17:47:02Z

test_samples/otel/src/agent_metric.py

+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import SimpleSpanProcessor, ConsoleSpanExporter
+
+


Unused imports: TracerProvider, SimpleSpanProcessor, and ConsoleSpanExporter are imported but never used in this file. Remove them to keep imports clean.

Suggested change

from opentelemetry.sdk.trace import TracerProvider

from opentelemetry.sdk.trace.export import SimpleSpanProcessor, ConsoleSpanExporter

Copilot · 2026-02-24T17:47:02Z

...icrosoft-agents-hosting-core/microsoft_agents/hosting/core/observability/_agent_telemetry.py

+def _ts() -> float:
+    """Helper function to get current timestamp in milliseconds"""
+    return datetime.now(timezone.utc).timestamp() * 1000
+


Unused function: The _ts() helper function is defined but never used in this file. Consider removing it.

Suggested change

def _ts() -> float:

"""Helper function to get current timestamp in milliseconds"""

return datetime.now(timezone.utc).timestamp() * 1000

Copilot · 2026-02-24T17:47:02Z

...icrosoft-agents-hosting-core/microsoft_agents/hosting/core/observability/_agent_telemetry.py

+
+    def __init__(self, tracer: Tracer | None = None, meter: Meter | None = None):
+        if tracer is None:
+             tracer = trace.get_tracer("M365.agents", "1.0.0")


Inconsistent indentation: Line 33 has an extra space before 'tracer'. Should be a single space like line 35.

Suggested change

tracer = trace.get_tracer("M365.agents", "1.0.0")

tracer = trace.get_tracer("M365.agents", "1.0.0")

Copilot · 2026-02-24T17:47:03Z

...icrosoft-agents-hosting-core/microsoft_agents/hosting/core/observability/_agent_telemetry.py

+        with self.tracer.start_as_current_span(span_name) as span:
+            attributes = self._extract_attributes_from_context(context)
+            span.set_attributes(attributes)
+            # span.add_event(f"{span_name} started", attributes)


Commented-out code: Line 97 contains a commented-out span.add_event() call. Either remove it if it's not needed or uncomment it if it should be active.

Suggested change

# span.add_event(f"{span_name} started", attributes)

Copilot · 2026-02-24T17:47:07Z

test_samples/otel/src/start_server.py

+    async def entry_point(req: Request) -> Response:
+
+        logger.info("Request received at /api/messages endpoint.")
+        text = await req.text()


Unused variable: The 'text' variable on line 23 is assigned but never used. Remove it if it's not needed.

Suggested change

text = await req.text()

Copilot · 2026-02-24T17:47:07Z

test_samples/otel/src/start_server.py

+    APP["adapter"] = agent_application.adapter
+
+    try:
+        run_app(APP, host="localhost", port=environ.get("PORT", 3978))


Type mismatch: environ.get("PORT", 3978) returns a string or int, but the port parameter expects an int. The default value 3978 is an int, but if PORT is set in the environment it will be a string. Convert to int: port=int(environ.get("PORT", 3978)).

Suggested change

run_app(APP, host="localhost", port=environ.get("PORT", 3978))

run_app(APP, host="localhost", port=int(environ.get("PORT", 3978)))

Copilot · 2026-02-24T17:47:07Z

test_samples/otel/src/agent_metric.py

+        self.tracer = trace.get_tracer("A365.AgentFramework")
+        self.meter = metrics.get_meter("A365.AgentFramework", "1.0.0")


Inconsistent naming: The test sample uses "A365.AgentFramework" for the tracer/meter name (line 28-29), while the core library uses "M365.agents" (see libraries/microsoft-agents-hosting-core/microsoft_agents/hosting/core/observability/_agent_telemetry.py line 33). Consider using consistent naming across the codebase for better observability and to avoid confusion.

Suggested change

self.tracer = trace.get_tracer("A365.AgentFramework")

self.meter = metrics.get_meter("A365.AgentFramework", "1.0.0")

self.tracer = trace.get_tracer("M365.agents")

self.meter = metrics.get_meter("M365.agents", "1.0.0")

Copilot · 2026-02-24T17:47:08Z

...icrosoft-agents-hosting-core/microsoft_agents/hosting/core/observability/_agent_telemetry.py

+                        success_callback(span, duration)
+                else:
+
+                    if failure_callback:
+                        failure_callback(span, exception)


Missing error handling for callbacks: If success_callback or failure_callback raise an exception in lines 137-142, it will propagate and potentially mask the original exception. Consider wrapping callback invocations in try-except blocks to ensure telemetry errors don't break application logic.

Suggested change

success_callback(span, duration)

else:

if failure_callback:

failure_callback(span, exception)

try:

success_callback(span, duration)

except Exception as callback_exc:

# Ensure telemetry callback failures do not affect application logic

span.record_exception(callback_exc)

else:

if failure_callback:

try:

failure_callback(span, exception)

except Exception as callback_exc:

# Ensure telemetry callback failures do not mask the original exception

span.record_exception(callback_exc)

Copilot · 2026-02-24T17:47:08Z

test_samples/otel/src/env.TEMPLATE

+OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_REQUEST=".*"
+OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE=".*"
+
+OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_CLIENT_REQUEST=".*"
+OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_CLIENT_RESPONSE=".*"


Using OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_* with value ".*" causes OpenTelemetry to capture and export all HTTP request and response headers, including Authorization, Cookie, and other secret-bearing headers. Anyone with access to the telemetry backend or pipeline could then retrieve credentials or other sensitive data from these exported headers. Limit these variables to a specific allowlist of non-sensitive headers and explicitly exclude authentication and other secret-related headers from capture.

Suggested change

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_REQUEST=".*"

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE=".*"

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_CLIENT_REQUEST=".*"

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_CLIENT_RESPONSE=".*"

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_REQUEST="user-agent,x-request-id,x-correlation-id"

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_SERVER_RESPONSE="content-type,content-length,x-request-id,x-correlation-id"

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_CLIENT_REQUEST="user-agent,x-request-id,x-correlation-id"

OTEL_INSTRUMENTATION_HTTP_CAPTURE_HEADERS_CLIENT_RESPONSE="content-type,content-length,x-request-id,x-correlation-id"

rodrigobr-msft added 9 commits December 5, 2025 11:10

Creating otel boilerplate sample

168bad8

Merge branch 'main' of https://github.com/microsoft/Agents-for-python …

62e9eb1

…into users/robrandao/otel-sample

Basis for otel support

b4d8e60

Improving design

8cc8757

Using telemetry hooks in storage

d17a161

Adding telemetry hooks to adapters

8d2266a

Setting up OTEL testing

e306fa8

Fix to ActivityTemplate

d0acb1d

Fixed field resolution when provided from and from_property in templates

d7fad12

Copilot AI review requested due to automatic review settings February 24, 2026 17:38

Copilot started reviewing on behalf of rodrigobr-msft February 24, 2026 17:39 View session

Copilot AI reviewed Feb 24, 2026

View reviewed changes

Another commit

d75395f

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Native instrumentation with OpenTelemetry#329

Native instrumentation with OpenTelemetry#329
rodrigobr-msft wants to merge 10 commits intomainfrom
users/robrandao/otel

rodrigobr-msft commented Feb 24, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		from opentelemetry.sdk.trace import TracerProvider
		from opentelemetry.sdk.trace.export import SimpleSpanProcessor, ConsoleSpanExporter

	def _ts() -> float:
	"""Helper function to get current timestamp in milliseconds"""
	return datetime.now(timezone.utc).timestamp() * 1000

	tracer = trace.get_tracer("M365.agents", "1.0.0")
	tracer = trace.get_tracer("M365.agents", "1.0.0")

	run_app(APP, host="localhost", port=environ.get("PORT", 3978))
	run_app(APP, host="localhost", port=int(environ.get("PORT", 3978)))

		self.tracer = trace.get_tracer("A365.AgentFramework")
		self.meter = metrics.get_meter("A365.AgentFramework", "1.0.0")

Conversation

rodrigobr-msft commented Feb 24, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants