[miladh]

Stacked on #2873

Description

The forecast and impute endpoints accept at most 32 series and 5 MiB per request, counted as sent (after gzip, unless it's disabled), and a cohort must reach the model in one request. split_into_requests (in utils/_forecasting.py) packs series into requests within both limits, without splitting a cohort, and keeps their input order within each request, so the API's error locations (items[i]) match the input. The caller passes a function that measures a request body as sent; #2875 measures it with the SDK's own request serialization. A cohort with more than 32 series, or a cohort or series over 5 MiB on its own, raises ValueError before anything is sent.

This is internal and not user-facing yet; #2875 uses it.

Checklist:

• Tests added/updated: test_utils/test_forecasting.py, including seeded randomized checks of both limits and of cohorts staying together.
• Documentation updated: not applicable (private helper).
• The PR title follows the Conventional Commit spec.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 93.18182% with 3 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (ai-ts-forecasting/01-data-classes@ea1105b). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
tests/tests_unit/test_utils/test_forecasting.py	88.46%	3 Missing ⚠️

Additional details and impacted files

@@                         Coverage Diff                          @@
##             ai-ts-forecasting/01-data-classes    #2874   +/-   ##
====================================================================
  Coverage                                     ?   93.52%           
====================================================================
  Files                                        ?      530           
  Lines                                        ?    55704           
  Branches                                     ?        0           
====================================================================
  Hits                                         ?    52097           
  Misses                                       ?     3607           
  Partials                                     ?        0           

Files with missing lines	Coverage Δ
cognite/client/utils/_forecasting.py	100.00% <100.00%> (ø)
tests/tests_unit/test_utils/test_forecasting.py	88.46% <88.46%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[gemini-code-assist]

Code Review

This pull request introduces a utility function split_into_requests to batch InputTimeSeries into requests of at most 256 without splitting cohorts, along with corresponding unit tests. The review feedback points out two issues: handling of empty input (which currently returns [[]] instead of []) and potential key collisions when grouping series without a cohort using integer indices. Actionable suggestions are provided to use object() as a fallback key to avoid collisions and to add a unit test verifying empty input behavior.