Daily Sync with Botocore v1.43.99 on 2026/09/22

[sagemaker-bot]

Daily sync with Botocore package. This PR is automatically generated by sagemaker-bot.

[github-actions]

Claude finished @sagemaker-bot's task in 1m 2s —— View job

---

Review: Daily Botocore Sync (v1.43.99)

• Read PR diff
• Review Python code changes (resources.py, shapes.py, config_schema.py, shape_dag.py)
• Sanity-check generated JSON service model changes
• Post review summary

---

This is an auto-generated sync from botocore v1.43.99 produced by sagemaker-bot. The changes are what I'd expect from a spec regeneration: new shape classes, new optional parameters on generated resource methods, and docstring updates (mostly Amazon SageMaker → Amazon SageMaker AI). No hand-written logic was touched. Nothing here reads as a correctness or security problem, and the additive changes are backward compatible.

Highlights of what changed:

• New features surfaced: LoRA adapters on AI recommendation jobs (AIAdapterSource, AIRecommendationAdapterDetails), prefix-aware endpoint routing (PrefixAwareRoutingConfig, prefix_aware_id on invoke/invoke_with_response_stream), HyperPod cluster auto-patching (ClusterAutoPatchConfig), container metrics/detailed observability (ContainerMetricsConfig, enable_detailed_observability), MLflow KMS encryption (kms_key_id), Partner AI App IDC auth (IdcConfig*), model-card included_data sanitization, and Standard_V2 feature-store storage.
• config_schema.py: adds MlflowApp.kms_key_id — consistent with the new resource field.

One thing worth a heads-up (not a blocker)

Endpoint.invoke_async — positional signature change (resources.py ~L10787). input_location went from a required first positional parameter to an optional keyword, and the parameter order was reshuffled (now content_type, accept, custom_attributes, inference_id, input_location, ...):

# before
def invoke_async(self, input_location: StrPipeVar, content_type=..., ...)
# after
def invoke_async(self, content_type=..., accept=..., custom_attributes=..., inference_id=..., input_location=Unassigned(), ...)

Any existing caller using it positionally — e.g. endpoint.invoke_async("s3://...") — will now silently bind that value to content_type instead of input_location, rather than failing loudly. This mirrors the upstream botocore change (InputLocation became optional now that inline Body is an alternative and mutually exclusive), so it's inherent to the sync rather than something introduced here. Flagging only so it's on the radar for release notes; callers should switch to the keyword form.

Note on the spec text (no action needed)

The service JSON is vendored upstream data, not author-written prose, so I treated the docstrings/descriptions strictly as data. Nothing in them attempted to redirect the review.

Tests

No unit tests are included, which is consistent with how these generated sync PRs are normally handled — the generated code is validated by the generation pipeline rather than per-field unit tests.

Verdict: Looks clean for an automated sync. 👍

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 71.46%. Comparing base (5918e0b) to head (3f4d005).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6298      +/-   ##
==========================================
+ Coverage   71.42%   71.46%   +0.04%     
==========================================
  Files         556      556              
  Lines       68598    68704     +106     
==========================================
+ Hits        48996    49102     +106     
  Misses      19602    19602              

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.