[aws-application-signals-bot]

Automated update of OpenTelemetry dependencies.

Build Status: ❌ failure

Updated versions:

• OpenTelemetry Python: 1.44.0
• OpenTelemetry Contrib: 0.65b0
• opentelemetry-sdk-extension-aws: 2.1.0
• opentelemetry-propagator-aws-xray: 1.0.2
• opentelemetry-instrumentation-openai-agents-v2: 0.1.0

Upstream releases with breaking changes:
Note: the mechanism to detect upstream breaking changes is not perfect. Be sure to check all new releases and understand if any additional changes need to be addressed.

opentelemetry-python:

• Version 1.44.0/0.65b0

opentelemetry-python-contrib:

• Version 1.44.0/0.65b0

Description of changes:

• Drops the opentelemetry-instrumentation-elasticsearch runtime dependency (from pyproject.toml, the lite-mode instrumentor map, and the nightly bot's CONTRIB_DEPS allowlist), removed entirely upstream now that all supported elasticsearch versions ship native OTel instrumentation:
  Drop elasticsearch instrumentation open-telemetry/opentelemetry-python-contrib#4759

• Migrates the three Events API call sites (llo_handler.py, debugger/_snapshot_otlp_emitter.py, aws_opentelemetry_configurator._init_logging) to emit LogRecord with the event_name field via the Logs API, after upstream removed the deprecated Events API/SDK (opentelemetry._events / opentelemetry.sdk._events):
  Remove Events API/SDK open-telemetry/opentelemetry-python#5293
  To keep the OTLP wire shape byte-identical to v0.18.0 for downstream consumers, the migration explicitly preserves the two things the old Event class auto-injected: the event.name attribute (which the GenAI E2E CWLogValidator asserts on) and severity_number=INFO. Verified by encoding both versions' output through the OTLP proto encoder and diffing — the only deltas are the additive top-level eventName field and SDK-generated resource metadata (telemetry.sdk.version, service.instance.id).

• Implements force_flush on CompactConsoleLogRecordExporter, required after upstream added it as an abstract method on the LogRecordExporter ABC (instantiation otherwise fails with TypeError):
  opentelemetry-sdk: add 'force_flush' method to LogRecordExporter ABC open-telemetry/opentelemetry-python#5294

[github-actions]

Claude finished @aws-application-signals-bot's task in 1m 58s —— View job

---

Claude Code PR Review

• Check for prior Claude reviews on this PR (none found)
• Review full PR diff
• Verify dependency version consistency
• Investigate build failure
• Post findings

Summary

Found one substantive issue and posted it as an inline comment on aws-opentelemetry-distro/pyproject.toml:79:

• opentelemetry-instrumentation-threading was missed during the bump — still pinned to == 0.63b1 while all 40+ sibling contrib packages were updated to == 0.64b0. Since opentelemetry-instrumentation itself is pinned to == 0.64b0, and instrumentation-threading 0.63b1 requires opentelemetry-instrumentation ~= 0.63b1, pip cannot resolve the dependency graph. This is almost certainly the cause of the failed build called out in the PR description.

The rest of the changes (mock-collector, contract-tests, lambda-layer tests, tox.ini upstream release branches) are internally consistent.
· Branch

[github-actions]

Claude finished @ezhang6811's task in 1m 37s —— View job

---

Claude Code PR Review

• Check for prior Claude reviews on this PR (1 inline comment found on aws-opentelemetry-distro/pyproject.toml:79)
• Review full PR diff
• Verify dependency version consistency
• Post findings

Summary

No new findings to post inline. The prior review's issue is still unresolved on this branch, so I'm not re-posting a duplicate.

Status of prior finding (still open):

• aws-opentelemetry-distro/pyproject.toml:79 — opentelemetry-instrumentation-threading == 0.63b1 was not bumped. Since this branch has since moved from 0.64b0 → 0.65b0 (via the merge with main + commit 5045f59), the correct target is now 0.65b0, not 0.64b0 as the prior inline suggestion block indicates. With opentelemetry-instrumentation == 0.65b0 (line 41) pinned alongside instrumentation-threading == 0.63b1 (which requires opentelemetry-instrumentation ~= 0.63b1), pip still can't resolve the graph — this remains the likely cause of the failing build.

  Fix this →

Rest of the diff: the mock-collector, contract-tests, lambda-layer tests, and tox.ini upstream-release-branch bumps are all internally consistent at 1.44.0/0.65b0.
· Branch

[github-actions]

Claude finished @liustve's task in 50s —— View job

---

Claude Code PR Review

• Check for prior Claude reviews on this PR (1 inline comment found on aws-opentelemetry-distro/pyproject.toml:79)
• Review full PR diff
• Verify dependency version consistency
• Post findings

Summary

No new findings to post inline. The prior review's issue is still unresolved on this branch, so I'm not re-posting a duplicate.

Status of prior finding (still open):

• aws-opentelemetry-distro/pyproject.toml:79 — opentelemetry-instrumentation-threading == 0.63b1 was not bumped to == 0.65b0. Since opentelemetry-instrumentation == 0.65b0 (line 41) is pinned alongside instrumentation-threading == 0.63b1 (which requires opentelemetry-instrumentation ~= 0.63b1), pip still can't resolve the graph — this remains the likely cause of the failing build.

  Fix this →

Rest of the diff: the mock-collector, contract-tests, lambda-layer tests, and tox.ini upstream-release-branch bumps are all internally consistent at 1.44.0 / 0.65b0.
· Branch

[github-actions]

Claude finished @liustve's task in 1m 10s —— View job

---

Claude Code PR Review

• Check for prior Claude reviews on this PR (1 inline comment found; author already resolved it in 318c036)
• Review full PR diff
• Verify dependency version consistency
• Verify the threading fix (commit 318c036)
• Post findings

Summary

No new findings to post inline. The one prior finding has been resolved on this branch.

Status of prior finding (resolved):

• aws-opentelemetry-distro/pyproject.toml:78 — opentelemetry-instrumentation-threading is now pinned to == 0.65b0 (bumped in commit 318c036), matching the rest of the contrib packages. The pip resolver conflict should be cleared.

Verified consistency across the diff:

• All 1.42.1 → 1.44.0 bumps applied uniformly (aws-opentelemetry-distro, mock-collector, contract-tests).
• All 0.63b1 → 0.65b0 bumps applied uniformly (aws-opentelemetry-distro, lambda-layer test requirements).
• opentelemetry-instrumentation-elasticsearch fully dropped in three places (removed upstream in 0.65b0):
  • aws-opentelemetry-distro/pyproject.toml
  • aws-opentelemetry-distro/src/amazon/opentelemetry/distro/opentelemetry_lite_sdk.py
  • scripts/update_dependencies.py
• tox.ini upstream release branches bumped to release/v1.44.x-0.65bx.
• CHANGELOG entry accurately describes both the version bump and the elasticsearch drop rationale.

Minor observation (not worth inline noise): lambda-layer/src/otel-instrument:140 still lists elasticsearch in the default OTEL_PYTHON_DISABLED_INSTRUMENTATIONS. Now that the instrumentation package is no longer shipped, that entry is a no-op — safe to leave, cosmetic-only if you decide to clean it up in a follow-up.

CI is currently in progress on the four required checks — no blocking issues from the diff itself.
· Branch

[liustve]

Pushed 318c036 to fix the failing build. There were two distinct dependency problems on this branch, not one:

1. `opentelemetry-instrumentation-threading` left at `0.63b1` (the issue the prior Claude reviews flagged). It pins `opentelemetry-instrumentation==0.63b1` exactly, conflicting with the `==0.65b0` pin. → Bumped to `0.65b0`.

2. `opentelemetry-instrumentation-elasticsearch == 0.65b0` does not exist on PyPI (this was the actual error in the current CI run — No matching distribution found for opentelemetry-instrumentation-elasticsearch==0.65b0, max published is `0.64b0`). Upstream dropped the package in the 1.44.0/0.65b0 contrib release "since all supported elasticsearch versions have a native OTel instrumentation" (CHANGELOG). → Removed it from:

• `aws-opentelemetry-distro/pyproject.toml`
• the lite-mode instrumentor map in `opentelemetry_lite_sdk.py`
• `CONTRIB_DEPS` in `scripts/update_dependencies.py` (so the nightly bot does not re-add a nonexistent pin next cycle)

Verified with a `pip install --dry-run` of the distro: the full graph now resolves cleanly at `1.44.0`/`0.65b0` with no conflicts. The one contract-test failure (`3.12 / psycopg2`) in the last run was an unrelated transient Docker registry timeout, not a dependency issue.

Also added a CHANGELOG entry for the bump + elasticsearch drop.

[liustve]

Pushed 1e1a411. Once the install/resolution was fixed (318c036), the contract tests surfaced additional upstream breaking changes in OTel 1.44.0/0.65b0 that the version pins alone did not cover:

1. Events API/SDK removed — opentelemetry._events and opentelemetry.sdk._events were deleted upstream (opentelemetry-python#4652 / 1.44.0 changelog: "Removed deprecated Events API/SDK. Use LogRecord with the event_name field set instead."). This was the ModuleNotFoundError: No module named 'opentelemetry._events' failing every contract test. Migrated all three call sites to emit a LogRecord with event_name via the Logs API:

• llo_handler.py (consolidated Gen AI LLO event)
• debugger/_snapshot_otlp_emitter.py (DI snapshots)
• aws_opentelemetry_configurator._init_logging (dropped the now-defunct EventLoggerProvider / set_event_logger_provider setup)

2. LogExporter.force_flush is now a required abstract method — CompactConsoleLogRecordExporter only implemented export/shutdown, so it could no longer be instantiated (TypeError: Can't instantiate abstract class ... without an implementation for abstract method 'force_flush'). Added force_flush.

Updated the affected unit tests to the LogRecord shape — event_name replaces the old Event.name, and the removed Event class no longer injects an event.name key into attributes. Verified locally against 1.44.0/0.65b0: full distro unit suite passes (2189 passed / 51 skipped), and isort/black/flake8 are clean.

[github-actions]

Claude finished @liustve's task in 3m 3s —— View job

---

Claude Code PR Review

• Read prior Claude reviews on this PR (1 inline; already resolved by author in 318c036)
• Review full PR diff (focus on new commits 318c036 and 1e1a411)
• Verify Events API → LogRecord migration and force_flush fix
• Post findings

Summary

No new findings to post inline. The one prior inline review has been resolved on this branch, and the two additional breaking-change fixes in 1e1a411 look correct.

Verified for commit 1e1a411 (upstream breaking-change adaptation):

• Events API removal — All three production call sites (llo_handler.py, debugger/_snapshot_otlp_emitter.py, aws_opentelemetry_configurator._init_logging) now use LogRecord(event_name=...) via the Logs API, matching upstream's 1.44.0 removal of opentelemetry._events / opentelemetry.sdk._events. No lingering imports of the removed modules, and no stray EventLoggerProvider / get_event_logger / set_event_logger_provider usages remain in the shipped code.
• event.name attribute preservation — _snapshot_otlp_emitter.py:117 still sets "event.name": _EVENT_NAME in the log attributes alongside event_name=_EVENT_NAME on the LogRecord. The old Event class auto-injected this key; the new LogRecord does not, so keeping it explicit preserves the on-wire shape for downstream consumers that read event.name from attributes. The corresponding unit test (test_snapshot_otlp_emitter.py:127) still asserts this, so behavior is preserved intentionally.
• CompactConsoleLogRecordExporter.force_flush — Added to satisfy the newly-abstract method on LogExporter / LogRecordExporter. Implementation is a minimal self._out.flush(); return True, matching the pattern of shutdown(). Direct unit-test coverage for this new method is not present, but it's exercised transitively by any LoggerProvider.force_flush() path — low-risk given how thin the method is.

Verified for commit 318c036 (previously validated): threading bump to 0.65b0, elasticsearch fully dropped from pyproject.toml, opentelemetry_lite_sdk.py, and scripts/update_dependencies.py.

Rest of the diff (mock-collector, contract-tests, lambda-layer tests, tox.ini upstream release branches) remains internally consistent at 1.44.0 / 0.65b0.

Minor observation (unchanged since prior review, not worth re-flagging inline): lambda-layer/src/otel-instrument:140 still lists elasticsearch in the default OTEL_PYTHON_DISABLED_INSTRUMENTATIONS. It's a no-op now that the package isn't shipped — safe cosmetic cleanup for a follow-up if desired.
· Branch

[github-actions]

Claude finished @liustve's task in 3m 50s —— View job

---

Claude Code PR Review

• Read prior Claude reviews on this PR (1 inline; already resolved by author in 318c036)
• Review full PR diff (focus on new commits 318c036, ffa8c72, 1e1a411, 7199413)
• Dedupe against prior findings — no new issues

Summary

No new findings to post inline. The one prior inline review has been resolved on this branch, and the additional breaking-change fixes in 1e1a411 look correct on re-review.

Verified for the latest changes:

• Events API removal — All production call sites (llo_handler.py, debugger/_snapshot_otlp_emitter.py, aws_opentelemetry_configurator._init_logging) now use LogRecord(event_name=...) via the Logs API, matching upstream's 1.44.0 removal of opentelemetry._events / opentelemetry.sdk._events. greping under pr-head/aws-opentelemetry-distro/ for EventLoggerProvider, get_event_logger, opentelemetry._events, and opentelemetry.sdk._events returns zero hits, so no lingering imports.
• CompactConsoleLogRecordExporter.force_flush — Added at aws-opentelemetry-distro/src/amazon/opentelemetry/distro/exporter/console/logs/compact_console_log_exporter.py:85-87. Minimal self._out.flush(); return True — appropriate for a synchronous stdout writer. timeout_millis is unused, which is fine here since flushing a local stream is effectively instantaneous.
• Behavioral note on event.name (deliberate, not flagged inline): _snapshot_otlp_emitter.py:117 and serviceevents/exporter/otlp_emitter.py:258 still explicitly set attributes["event.name"] = ... alongside event_name= on the LogRecord (preserving the wire shape the old Event class used to auto-inject). llo_handler._emit_llo_attributes does not do this, and the updated tests deliberately replace assertIn("event.name", attributes) with assertEqual(event_name, ...). If any downstream consumer of Gen AI LLO events reads event.name from attributes rather than the top-level event_name field, they'll need to be updated — worth confirming with whoever owns the downstream side, but not a blocker on the diff itself.
• Version consistency — 1.42.1→1.44.0 and 0.63b1→0.65b0 bumps are uniform across aws-opentelemetry-distro/pyproject.toml, contract-tests/*, lambda-layer/src/tests/requirements.txt, mock-collector, and tox.ini. Threading is at 0.65b0; elasticsearch is fully dropped from pyproject.toml, opentelemetry_lite_sdk.py, and scripts/update_dependencies.py.
• CHANGELOG (7199413) — Simplified entry looks fine given the details are captured in the commit messages and PR body.

Minor observation (unchanged since prior review, not worth re-flagging inline): lambda-layer/src/otel-instrument:140 still lists elasticsearch in the default OTEL_PYTHON_DISABLED_INSTRUMENTATIONS. It's a no-op now that the package isn't shipped — safe cosmetic cleanup for a follow-up if desired.

CI is currently in progress on the four required checks — no blocking issues from the diff itself.
· Branch

[liustve]

Follow-up in 5f45e5d — preserving the event.name wire attribute on the Gen AI LLO events.

The removed Events API's Event class auto-injected the event name into the LogRecord's attributes as event.name (verified against 1.42.1: the old Event produced attributes={..., "event.name": <scope>} with a None top-level event_name). The DI snapshot emitter and the serviceevents emitter already set attributes["event.name"] explicitly, but llo_handler._emit_llo_attributes relied on the Event class to add it — so the initial migration silently dropped it from emitted Gen AI LLO events.

Fixed by setting event.name in the attributes dict in _emit_llo_attributes, so the on-the-wire shape is identical to before (attributes carry event.name, plus the new top-level event_name field the Logs API populates). The three emitters are now consistent, the test assertions revert to their originals (assertIn("event.name", attributes)), and no downstream consumer that reads attributes["event.name"] is affected.

[github-actions]

Claude finished @liustve's task in 1m 33s —— View job

---

Claude Code PR Review

• Read prior Claude reviews on this PR (1 inline; already resolved by author in 318c036)
• Review latest commit 5f45e5d (Gen AI LLO event.name preservation)
• Verify test assertions match the wire-shape claim
• Post findings

Summary

No new inline findings. Prior inline was already resolved by the author; the additional fixes in 1e1a411 and now 5f45e5d all look correct.

Verified for commit 5f45e5d (Gen AI LLO event.name preservation):

• aws-opentelemetry-distro/src/amazon/opentelemetry/distro/llo_handler.py:540 — event_attributes = {"event.name": span.instrumentation_scope.name} now explicitly seeds event.name in the LogRecord attributes dict, matching what the removed Event class used to auto-inject. This brings llo_handler in line with the two other emitters (_snapshot_otlp_emitter.py:117 and serviceevents/exporter/otlp_emitter.py:258) that were already setting attributes["event.name"] explicitly. The three emitters are now consistent.
• Test assertions in test_llo_handler_events.py now assert assertIn("event.name", emitted_event.attributes) on all three code paths (with/without session.id, with other attributes), which correctly verifies the wire shape is preserved. The top-level event_name field on the LogRecord is populated by the Logs API separately.
• No new imports needed. No lingering EventLoggerProvider / get_event_logger / opentelemetry._events references remain under aws-opentelemetry-distro/src/ — grep confirms zero hits.

Minor cosmetic observations (not worth re-flagging inline):

• llo_handler.py:548 — attributes=event_attributes if event_attributes else None is now effectively dead-branch: event_attributes is always truthy (it always has event.name). Harmless — safe to simplify to attributes=event_attributes on a future cleanup.
• Test comments now say # Event class always adds event.name (e.g. test_llo_handler_events.py:639, 669, 694). The old Event class is gone; the attribute is now added explicitly in _emit_llo_attributes. Comment is stale but low-value to churn on.
• lambda-layer/src/otel-instrument:140 still lists elasticsearch in the default OTEL_PYTHON_DISABLED_INSTRUMENTATIONS — no-op now that the package isn't shipped (noted in prior reviews).

Rest of the PR (unchanged since prior review): version bumps 1.42.1→1.44.0 and 0.63b1→0.65b0 are uniform across aws-opentelemetry-distro/pyproject.toml, contract-tests/*, lambda-layer/src/tests/requirements.txt, mock-collector, and tox.ini. threading is at 0.65b0; elasticsearch is fully dropped from pyproject.toml, opentelemetry_lite_sdk.py, and scripts/update_dependencies.py.

CI is currently in progress on the four required checks — no blocking issues from the diff itself.
· Branch

[github-actions]

Claude Code PR Review

• Read prior Claude reviews on this PR
• Review full PR diff (focus on latest commits)
• Dedupe against prior findings
• Post findings

View job run · Branch

[github-actions]

Claude finished @liustve's task in 1m 20s —— View job

---

Claude Code PR Review

• Read prior Claude reviews on this PR (1 inline; already resolved by author in 318c036)
• Review full PR diff (focus on latest commits 5f45e5d, 8a9eb62, 33ce98d)
• Dedupe against prior findings — no new issues
• Post findings

Summary

No new findings to post inline. Prior inline was resolved by the author; the two newest commits (8a9eb62, 33ce98d) are minimal, targeted fixes that look correct.

Verified for 8a9eb62 and 33ce98d (severity_number=SeverityNumber.INFO on Gen AI LLO + DI snapshot LogRecords):

• The removed Event class defaulted its severity_number to SeverityNumber.INFO, so anything downstream keying off severity_number on the emitted log would have seen a change (from INFO to UNSPECIFIED, the LogRecord default) once the migration in 1e1a411 landed. Setting it explicitly on both emitters restores the pre-migration wire shape — consistent with the earlier event.name preservation fix in 5f45e5d.
• The serviceevents emitter (serviceevents/exporter/otlp_emitter.py:274) intentionally keeps SeverityNumber.UNSPECIFIED — it was never built on Event, so it doesn't share the "match old Event default" rationale. No inconsistency to flag.
• No test coverage was added asserting severity_number == INFO, but the two commits are two-line changes that are self-evident from the wire-shape rationale in the commit messages — low risk.

Cosmetic observations (unchanged since prior reviews; not worth re-flagging inline):

• llo_handler.py:548 — attributes=event_attributes if event_attributes else None is now a dead branch (event_attributes always contains at least "event.name"). Safe to simplify on a future cleanup.
• lambda-layer/src/otel-instrument:140 still lists elasticsearch in the default OTEL_PYTHON_DISABLED_INSTRUMENTATIONS — a no-op now that the instrumentation isn't shipped.
• test_llo_handler_events.py:639/669/694 comments still say # Event class always adds event.name; the Event class is gone and the attribute is now set explicitly in _emit_llo_attributes. Low-value stale comment.

Rest of the PR (verified consistent): 1.42.1→1.44.0 and 0.63b1→0.65b0 bumps applied uniformly across aws-opentelemetry-distro/pyproject.toml, contract-tests/*, lambda-layer/src/tests/requirements.txt, mock-collector, and tox.ini. threading at 0.65b0; elasticsearch fully dropped from pyproject.toml, opentelemetry_lite_sdk.py, and scripts/update_dependencies.py.
· Branch

[github-actions]

Claude finished @liustve's task in 1m 47s —— View job

---

Claude Code PR Review

• Read prior Claude reviews on this PR (1 inline; resolved by author in 318c036)
• Review full PR diff (focus on latest commit e38161a)
• Dedupe against prior findings — no new inline findings
• Post findings

Summary

No new inline findings. Latest commit e38161a cleans up the three cosmetic observations flagged in prior reviews:

• llo_handler.py:549 — simplified attributes=event_attributes if event_attributes else None → attributes=event_attributes, since event_attributes is now always seeded with event.name. Dead branch removed.
• lambda-layer/src/otel-instrument:140 — dropped the now-defunct elasticsearch entry from the default OTEL_PYTHON_DISABLED_INSTRUMENTATIONS list. No-op behavior preserved (the package isn't shipped), and the env var no longer implies a package that no longer exists.
• test_llo_handler_events.py:639, 669 — stale # Event class always adds event.name comments updated to # event.name is always set in attributes, matching the new explicit-set behavior.

Minor leftover (not worth re-flagging inline): test_llo_handler_events.py:694 still has (plus event.name from Event class) in an inline comment — e38161a updated the sibling test comments but missed this one. Purely cosmetic.

Rest of the PR (unchanged since prior reviews, all still verified):

• Version bumps 1.42.1→1.44.0 and 0.63b1→0.65b0 uniform across aws-opentelemetry-distro/pyproject.toml, contract-tests/*, lambda-layer/src/tests/requirements.txt, mock-collector, and tox.ini. threading at 0.65b0.
• elasticsearch instrumentation fully dropped from pyproject.toml, opentelemetry_lite_sdk.py, scripts/update_dependencies.py, and now the Lambda layer's default-disabled list.
• Events API → LogRecord(event_name=...) migration complete across all three emitters (llo_handler.py, _snapshot_otlp_emitter.py, _init_logging). event.name attribute and severity_number=INFO preserved on the wire for downstream compatibility. No lingering opentelemetry._events / EventLoggerProvider / get_event_logger references (verified via grep).
• CompactConsoleLogRecordExporter.force_flush added to satisfy the newly-abstract LogRecordExporter method.
• CHANGELOG entry present.
  · Branch