[hln33]

Summary

Adds the Python handler for conformance requirement telemetry-3 (per-operation result selection). The requirement is defined in aws/aws-durable-execution-conformance-tests#183 and was raised in aws/aws-durable-execution-conformance-tests#182.

Step fetch-user returns {userId, email, token}. An OperationOverride(operation_name="fetch-user", result=select_user_id) returns {userId} alone. content.input=False and content.output=False omit the execution input and output from the record. So the exported record has no input, no output, and operationsByName.fetch-user.result == {"userId": "u-123"}. The execution result still holds all three keys, which the suite's core ExpectedResult check verifies.

Design and alternatives

Problem: Verify that telemetry exports only userId from a step result while the execution still returns userId, email, and token.

Coverage gap: The existing telemetry scenarios do not exercise per-operation result selection.

Alternatives evaluated during review:

• Add no handler. This leaves the result-selection behavior untested by the deployed conformance suite.
• Reuse the existing success handler unchanged. It exercises an increment step and includes execution input and output. It cannot produce the required result-selection record.
• Extend that handler to serve both scenarios. The scenarios require different step results and content settings. Combining them adds scenario selection to the handler and configuration.
• Return only userId from the step. This changes the execution result. It also avoids testing whether the telemetry transform selects the field correctly.

Chosen approach: Keep a dedicated handler with one step and one result transform. Disable execution input and output in its telemetry configuration. Return the complete step result unchanged. This preserves the existing scenarios without adding scenario-selection logic.

Removals: None. This adds conformance coverage using the existing SDK API.

Please confirm this approach before further implementation changes.

Changes

• src/telemetry_3_result_select.py: new handler.
• template.yaml: Telemetry3ResultSelect → TestDescription: [telemetry-3], function name ${AWS::StackName}-telemetry-3.
• src/Makefile: build-Telemetry3ResultSelect target.
• tests/test_telemetry_examples.py: module/function sets include telemetry-3; new local-runner test asserts the one-key record result and the absent input/output.
• README.md: coverage and layout rows.
• .github/workflows/telemetry-conformance-tests.yml: pinned to the fix: Change default key 'name' for logger extra field #183 merge commit on main (cf13ef2).

Verification

• hatch run test:all packages/aws-durable-execution-sdk-python-conformance-tests-telemetry/tests: 20 passed.
• hatch fmt --check on the package: clean.
• hatch run types:check: clean.
• Telemetry conformance e2e: telemetry-1, telemetry-2, and telemetry-3 passed. 3 passed, 0 failed. Run 42691679335 at 3efaa4be. Re-run after the re-pin: Run 42713583152 at c1943e80, pinned to cf13ef27, 3 passed, 0 failed.

[hln33]

/ai review

[ParidelPooya]

Looks good. The handler matches telemetry-3. _apply_result_override decodes the checkpoint before it calls the transform, so result["userId"] is right. CI ran telemetry-3 against cb3753e and it passed.

A few small things:

1. The PR body ("Draft until") and the workflow comment at telemetry-conformance-tests.yml:46 say "before leaving draft", but the PR is not a draft. Suggest rewording it to "before this PR merges".
2. README.md:23: the new row is 4 characters shorter than the other rows. The layout comment on line 31 starts at column 35, while the other lines use column 33.
3. Not from this PR: mypy reports Too few arguments [call-arg] on durable_execution(plugins=...) at telemetry_3_result_select.py:66. The sibling handlers on main have the same error. types:check does not cover this package, so CI does not catch it. A separate change could add @overloads to durable_execution.

[github-actions]

Codex AI review

No actionable findings across the six changed files and contribution rules 1–9. The updated design note resolves the prior finding. Residual risk: tests were not run, and the newly pinned external workflow was not independently inspected.

Reviewed commit c1943e806da15b0efd6ae767709981599399ece2. Workflow run

[hln33]

Thanks. Addressed in c1943e8:

1. The workflow pin now points at the fix: Change default key 'name' for logger extra field #183 merge commit cf13ef27 and the TODO is removed. The PR body no longer has a "Draft until" section.
2. README.md:23: the telemetry-3 row is padded to the table width. The layout block's comments all start at column 32, since telemetry_3_result_select.py is longer than the column the other lines used.
3. The mypy call-arg error is tracked in [Bug]: durable_execution(plugins=...) return type fails mypy with "Too few arguments" #921. The declared return type is the two-argument handler in both call forms, so the decorator-factory form fails; two @overloads fix it.

Telemetry e2e on the re-pinned head: run 42713583152, telemetry-1, -2, -3 passed.

[ParidelPooya]

Thanks. The README and the pin comment are fixed. The suite is pinned to the #183 merge commit cf13ef2, and telemetry-3 passes on it. The mypy point is from main, so it can go in a separate change. Approving.