Add client-side telemetry hooks with OpenTelemetry integration and ad-hoc capture#133
Draft
Add client-side telemetry hooks with OpenTelemetry integration and ad-hoc capture#133
Conversation
Adds a complete telemetry infrastructure that lets users hook into the SDK's HTTP request lifecycle, integrate with OpenTelemetry for tracing/metrics, and use Python's standard logging -- all opt-in with zero overhead when disabled. Key components: - TelemetryConfig: frozen dataclass for configuring signals and hooks - TelemetryHook: concrete base class for custom telemetry integrations - TelemetryManager / NoOpTelemetryManager: instrumentation engine with factory that returns a zero-overhead no-op when telemetry is disabled - RequestContext / ResponseContext: typed data passed to hooks - _operation_scope: ContextVar-based mechanism to propagate operation names (e.g. "records.create") from namespace methods to _request() Design improvements over the initial draft: - TelemetryHook as concrete base class (not broken @runtime_checkable Protocol) - _span removed from user-facing RequestContext (kept internal via _TrackedRequest) - No except block in trace_request() -- eliminates double error dispatch - Per-subsystem try/except in record_response() for exception safety - NoOp yields None for true zero allocation overhead - Explicit span status (OK/ERROR) on every response - Safe log_level parsing with fallback Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Resolve conflicts in operations layer: combine telemetry _operation_scope wrapping (branch) with typed return models Record/TableInfo/RelationshipInfo (main). Both features now work together - operations are instrumented AND return typed objects.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Remove unused imports in test_telemetry.py (time, Mock) - Use env vars for org URL/tenant ID in telemetry_demo.py - Normalize None header values to empty string in _odata.py - Include network exceptions (error is not None) in error count metric - Scope _operation_scope per page fetch, not across yield points - Dispatch on_request_error hook for network exceptions - Update SKILL.md docs to match on_request_error behavior Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add skipUnless(_OTEL_AVAILABLE) to 3 tests that require opentelemetry - Clarify on_request_error docstring: covers both HTTP and network errors - Fix SKILL.md: separate request vs response fields in hook data docs Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add table_name arg to create_one_to_many (referencing_entity) and create_many_to_many (entity1_logical_name) so telemetry traces and metrics can be filtered by table. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Hook-provided headers now only set headers that aren't already present, preventing hooks from silently overwriting Authorization, OData, or correlation headers. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Hook header merge now uses case-insensitive comparison to prevent hooks from bypassing SDK headers via different casing - Remove redundant 'pip install opentelemetry-sdk' from demo (already included in [telemetry] extra) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add DataverseClient import to README and SKILL.md telemetry snippets - Stop calling logger.setLevel() which mutates global logging state; use internal _log_level threshold to filter emissions instead - Update test to check _log_level instead of logger.level Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Guard against Optional[Dict] type by assigning headers = ... or {}
before iterating, preventing potential None iteration if the type
annotation is ever accurate.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds client.capture_telemetry() -- a lightweight context manager that
captures HTTP request details without requiring full telemetry hook
setup. Works even when no TelemetryConfig is set on the client.
Usage:
with client.capture_telemetry() as t:
record_id = client.records.create("account", {"name": "Contoso"})
print(t.requests[-1].service_request_id)
New types: TelemetryCapture (hook subclass), CapturedRequest (dataclass).
When telemetry is disabled (NoOp), temporarily swaps in a real manager
for the duration of the capture block, then restores NoOp on exit.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This was referenced Mar 13, 2026
Closed
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds a complete, opt-in telemetry infrastructure that lets users hook into the SDK's HTTP request lifecycle, integrate with OpenTelemetry for tracing and metrics, and use Python's standard logging -- all with zero overhead when disabled.
New Public API
TelemetryConfigTelemetryHookon_request_start,on_request_end,on_request_error,get_additional_headers)RequestContextResponseContextclient.capture_telemetry()CapturedRequestcapture_telemetry()Internal Components
TelemetryManager/NoOpTelemetryManager-- instrumentation engine with factory; returns a zero-overhead no-op when telemetry is disabled_operation_scope--ContextVar-based mechanism to propagate operation names (e.g.,records.create) from namespace methods to_request()_TrackedRequest-- internal span tracking (not exposed to hooks)Key Design Decisions
NoOpTelemetryManageryieldsNonefor true zero-allocation overhead_log_levelthreshold filters log emissions instead of callinglogger.setLevel()pip install PowerPlatform-Dataverse-Client[telemetry]; hooks work standalone without OTelOperation Instrumentation
All operation namespaces are instrumented with
_operation_scope:records(create, read, update, delete, upsert)query(sql, fetch_page)tables(get_columns, create_one_to_many, create_many_to_many)files(upload, download)Changes
src/PowerPlatform/Dataverse/core/telemetry.py-- core telemetry engine (617 lines)examples/telemetry_demo.py-- runnable demo with console/Jaeger exporterstests/unit/core/test_telemetry.py-- comprehensive test suite (758 lines)client.py-- addscapture_telemetry()method, wires telemetry managerdata/_odata.py-- instruments_request()with trace/record/hook dispatchoperations/*.py-- wraps operations with_operation_scopecore/config.py-- addstelemetryfield toDataverseConfigcommon/constants.py-- adds telemetry-related constantspyproject.toml-- adds[telemetry]optional dependency groupREADME.md-- adds telemetry & observability section