mirror of
https://github.com/microsoft/PowerToys.git
synced 2026-09-02 04:01:25 +02:00
93aeae9aa1
## Summary of the Pull Request Adds a regression test guarding against re-introduction of a hardcoded `ReasoningEffort` value in `SemanticKernelPasteProvider.CreateExecutionSettings()`. This value previously broke every OpenAI/Azure OpenAI custom paste action for models that don't support `reasoning_effort: minimal`, causing HTTP 400 errors (see #49838). The underlying fix already exists on `main` — `CreateExecutionSettings()` no longer sets `ReasoningEffort`, and the Phi Silica on-device provider has since been split into its own `PhiSilicaPasteProvider` class. This PR does not modify provider logic; it adds test coverage to prevent this specific regression from being reintroduced. ## PR Checklist - [x] Closes: #xxx _(Not applicable — this PR does not close #49838, since the underlying fix already landed separately. Filed to add regression coverage only.)_ - [x] **Communication:** I've discussed this with core contributors already. If the work hasn't been agreed, this work might be rejected _(Not yet discussed with core contributors — flagging that this is a test-only addition in response to already-observed regression risk, and happy to adjust scope/approach based on maintainer feedback.)_ - [x] **Tests:** Added/updated and all pass - [x] **Localization:** All end-user-facing strings can be localized _(N/A — no user-facing strings changed.)_ - [x] **Dev docs:** Added/updated _(N/A — internal test-only change, no dev docs affected.)_ - [x] **New binaries:** Added on the required places _(N/A — no new binaries introduced.)_ - [x] **Documentation updated:** If checked, please file a pull request on our docs repo and link it here: #xxx _(N/A — no user-facing documentation changes.)_ ## Detailed Description of the Pull Request / Additional comments Issue #49838 reported that Advanced Paste custom actions fail with HTTP 400 for OpenAI/Azure OpenAI models that don't support `reasoning_effort: minimal`, a value that was hardcoded in `SemanticKernelPasteProvider.CreateExecutionSettings()` (originally introduced in #46727 alongside the Phi Silica on-device provider). By the time this was investigated, the hardcoded value had already been removed from `main`, and the code now includes an explicit comment warning against reintroducing model-specific tuning properties in this shared method. However, there was no test enforcing that guarantee — meaning a future change could silently reintroduce the same class of bug (this is the second time this exact pattern has caused a regression; see #43766 for the first). This PR adds a unit test in `AdvancedPaste.UnitTests` that constructs an OpenAI-configured `SemanticKernelPasteProvider` and asserts that `CreateExecutionSettings()` returns an `OpenAIPromptExecutionSettings` object with `ReasoningEffort` left unset. No production code is modified. ## Validation Steps Performed - Added the new unit test to `AdvancedPaste.UnitTests`. - Ran the test locally against the current `main` branch and confirmed it passes. - Confirmed no other tests in the `AdvancedPaste.UnitTests` project were affected by this change. Co-authored-by: google-labs-jules[bot] <161369871+google-labs-jules[bot]@users.noreply.github.com>