fix(web): preserve streamed text spacing#2174
Open
AcWrong02 wants to merge 1 commit into
Open
Conversation
🦋 Changeset detectedLatest commit: ff456a8 The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Related Issue
#2184
Problem
In
apps/kimi-web, streamed assistant text and thinking segments were merged with a hardcoded'\n'between consecutive chunks. Because stream chunks already carry the model's own whitespace (including paragraph breaks), this inserted newline turned token-sized chunks into one word per line — most visible for Chinese output, where tokens carry no trailing space.What changed
In
apps/kimi-web/src/composables/messagesToTurns.ts, stop inserting'\n'when appending to a trailing text/thinking block. Concatenate the chunk text directly so the model's original whitespace is preserved. The block-splitting behavior (a tool call between two text segments still splits them into separate blocks) is unchanged.Added a test in
apps/kimi-web/test/turn-logic.test.tscovering streamed text merged around empty thinking parts, asserting the result is'one two three'rather than newline-separated fragments.Checklist
gen-changesetsskill, or this PR needs no changeset.gen-docsskill, or this PR needs no doc update.