feat(app): require double-tap Escape to cancel AI response#14845
Closed
praxstack wants to merge 14 commits intoanomalyco:devfrom
Closed
feat(app): require double-tap Escape to cancel AI response#14845praxstack wants to merge 14 commits intoanomalyco:devfrom
praxstack wants to merge 14 commits intoanomalyco:devfrom
Conversation
…cing When message.part.updated events coalesce within a 16ms batch window (text-end replaces text-start at the same queue index), stale message.part.delta events remained in the queue. On flush, the reducer applied the full text from the coalesced update, then appended the stale deltas on top — doubling the content. Fix: When a message.part.updated event coalesces over an earlier entry, void all stale message.part.delta events for the same messageID:partID already in the queue. Voided indices are skipped during flush. Zero overhead in normal operation (voided set stays empty when no coalescing occurs). Scoped by messageID+partID to avoid affecting unrelated parts. Fixes content repetition visible during LLM streaming in desktop UI. Content displays correctly after app restart (loaded from DB).
- Add createEffect that applies fontSize to CSS custom properties (--font-size-base, --font-size-small, --font-size-large) - Replace hardcoded fontSize: 14 in terminal with settings value - Add font size stepper control (10-24px) in Settings > Appearance
When the LLM API returns a context overflow error, trigger compaction instead of killing the session. Previously the catch block had a TODO comment and fell through to a fatal error (return 'stop'). Now: sets needsCompaction = true and breaks to post-catch cleanup, which returns 'compact'. The prompt loop then creates a compaction task, generating a summary and allowing the session to continue.
Add SessionCompaction.prune() calls before both compaction triggers: 1. Proactive overflow check (before LLM call) 2. Reactive compact result (after LLM returns 'compact') Pruning strips old tool outputs first, maximizing context reduction before the compaction summary LLM call. This prevents the summarization itself from hitting context limits.
Add a prominent context usage card at the top of the session context panel showing: - Large progress circle (32px) with token count and usage % - Color-coded progress bar (green/yellow/red by usage level) - Compact button that calls session.summarize API directly - Button highlights when usage > 70% to encourage compaction
Add wideMode toggle in Settings > Appearance that disables the centered max-width constraint on the chat timeline. When enabled, chat content expands to fill the full window width instead of being capped at 800px/1000px.
Remove document.execCommand('insertText') which creates DIV block
elements for multi-line paste. getCursorPosition doesn't count the
implicit newlines at block boundaries, causing cursor to land in the
middle of pasted text after re-render.
Now always uses addPart() which inserts via createTextFragment
(text nodes + BR tags, no block elements), ensuring correct cursor
positioning. Also removes deprecated execCommand API usage.
Single Escape during streaming now shows a toast warning instead of immediately aborting. User must press Escape again within 500ms to confirm cancellation. Prevents accidental interruption of AI responses. Ctrl+G still provides immediate cancel for power users.
Contributor
|
Closing this pull request because it has had no updates for more than 60 days. If you plan to continue working on it, feel free to reopen or open a new 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.
Issue for this PR
Closes #14844
Type of change
What does this PR do?
Single Escape during streaming shows a toast warning. Must press Escape again within 500ms to confirm cancellation. Prevents accidental interruption. Ctrl+G still provides immediate cancel for power users.
How did you verify your code works?
Checklist