fix: preserve Step thinking effort preferences - #192
Merged
Merged
Conversation
…ffort
Review follow-up on the previous commit.
The stale-discovery guard compared model objects by reference, while the
session emits model_select through modelsAreEqual. When a provider is
registered or unregistered, _refreshCurrentModelFromRegistry swaps in an
equal but new model object, and the reference check then skipped a clamp
that should run. Use modelsAreEqual so the guard agrees with the session.
stepHighestEffort was referenced only by its own test, and its doc comment
("used as the default level for a model") describes the default this
branch removes. Delete it together with that test case.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Step's startup and model-selection hooks always reselected the highest supported thinking effort after the session had already applied CLI and saved preferences. For models supporting low/medium/high, both
step --thinking mediumand a saved medium default were overwritten with high.Keep the session-resolved effort when enriching model capabilities, clamp only unsupported choices, and ignore late discovery results for a model that is no longer active. The thinking selector now marks the effective saved default using per-model/global precedence and capability clamping.
With no configured preference, Step follows the SDK default (medium, clamped to model capabilities). Enter and
/effort mediumremain session-scoped; Ctrl+S still saves a default.Validation:
pnpm run build:offlinepassed.pnpm run checkpassed, with Biome run without--write.