-
Notifications
You must be signed in to change notification settings - Fork 2.6k
Fix: Resolve Memory Leak in ChatView Virtual Scrolling Implementation #6697
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
… periodic cleanup of cache entries for removed messages, and limits visible message processing to the most recent 500. Improves type safety by adding explicit types to various callbacks and state setters, and adjusts Virtuoso viewport and event handler types for better performance and maintainability.
|
Not sure why its failing tests as it doesnt fail on my end, investigating.. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you for your contribution! I've reviewed the memory leak fix and have some suggestions for improvement. The core approach of limiting the viewport and implementing cleanup is solid, but there are opportunities to make the implementation more robust and maintainable.
| max: 250, | ||
| ttl: 1000 * 60 * 15, // 15 minutes TTL for long-running tasks | ||
| max: 100, | ||
| ttl: 1000 * 60 * 5, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could we add inline comments explaining why these specific values were chosen? For example:
| const visibleMessages = useMemo(() => { | ||
| const newVisibleMessages = modifiedMessages.filter((message) => { | ||
| const currentMessageCount = modifiedMessages.length | ||
| const startIndex = Math.max(0, currentMessageCount - 500) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Consider extracting these magic numbers as named constants at the top of the file for better maintainability:
| // increasing top by 3_000 to prevent jumping around when user collapses a row | ||
| increaseViewportBy={{ top: 3_000, bottom: Number.MAX_SAFE_INTEGER }} // hack to make sure the last message is always rendered to get truly perfect scroll to bottom animation when new messages are added (Number.MAX_SAFE_INTEGER is safe for arithmetic operations, which is all virtuoso uses this value for in src/sizeRangeSystem.ts) | ||
| data={groupedMessages} // messages is the raw format returned by extension, modifiedMessages is the manipulated structure that combines certain messages of related type, and visibleMessages is the filtered structure that removes messages that should not be rendered | ||
| increaseViewportBy={{ top: 3_000, bottom: 1000 }} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Good fix! Changing from to prevents the infinite rendering issue. Could we add a comment explaining this critical change?
| }, []) | ||
|
|
||
| useEffect(() => { | ||
| const cleanupInterval = setInterval(() => { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Consider adding explicit type annotations for better code maintainability:
Improves the cleanup logic for the everVisibleMessagesTsRef cache by ensuring that only messages not present in current or recent viewport messages are deleted. Also updates event typing for suggestion clicks and minor refactoring for clarity.
Moved the cleanup interval and debounced focus effect hooks to after the visibleMessages memoization. This change improves code organization and ensures that hooks are declared after all dependent variables are defined.
…ith limited viewport - Replace scrollTo with MAX_SAFE_INTEGER with scrollToIndex - Ensures scroll-to-bottom works correctly with viewport limited to 1000px - Add groupedMessages.length to dependency arrays
daniel-lxs
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you @xyOz-dev!
- Use dynamic increaseViewportBy bottom value based on scroll position - When at bottom, use larger value (10,000px) to maintain scroll lock - When scrolled up, use smaller value (1,000px) for memory efficiency - Add followOutput='smooth' for better scroll behavior - Fixes scroll lock issues introduced in PR #6697
The scrollToIndex method causes jitter during streaming because it tries to calculate and jump to specific message indices while content is rapidly changing. Using scrollTo with Number.MAX_SAFE_INTEGER simply scrolls to the bottom without tracking specific indices, eliminating the jitter while keeping all memory optimizations from PR #6697.
* main: Changeset version bump (RooCodeInc#6826) Update contributors list (RooCodeInc#6636) chore: add changeset for v3.25.10 (RooCodeInc#6825) feat: add GPT-5 model support (RooCodeInc#6819) fix: add missing MCP error translation keys (RooCodeInc#6821) fix: use CDATA sections in XML examples to prevent parser errors (RooCodeInc#4852) (RooCodeInc#6811) Changeset version bump (RooCodeInc#6810) chore: add changeset for v3.25.9 (RooCodeInc#6809) Fix rounding of max tokens (RooCodeInc#6808) feat: add GLM-4.5 and OpenAI gpt-oss models to Fireworks provider (RooCodeInc#6784) feat: focus chat input when clicking plus button in extension menu (RooCodeInc#6689) Support linking to @roo-code/cloud in Roo-Code repo (RooCodeInc#6799) Bring back "Use @roo-code/cloud from npm" (RooCodeInc#6795) Changeset version bump (RooCodeInc#6790) Release v3.25.8 (RooCodeInc#6789) fix: prevent unnecessary MCP server refresh on settings save (RooCodeInc#6772) (RooCodeInc#6779) fix: Replace scrollToIndex with scrollTo to fix scroll jitter (RooCodeInc#6780) Clamp default model max tokens to 20% of context window (RooCodeInc#6761) fix: handle current directory path "." correctly in codebase_search tool (RooCodeInc#6517) fix: recover from error state when Qdrant becomes available (RooCodeInc#6661) Revert "Use @roo-code/cloud from npm" (RooCodeInc#6742) fix: prevent disabled MCP servers from starting processes and show correct status (RooCodeInc#6084) feat: reduce Gemini 2.5 Pro minimum thinking budget to 128 (RooCodeInc#6588) fix: trim whitespace from OpenAI base URL to fix model detection (RooCodeInc#6560) fix: improve handling of net::ERR_ABORTED errors in URL fetching (RooCodeInc#6635) Add swift files to fallback list (RooCodeInc#6724) Fix: Resolve Memory Leak in ChatView Virtual Scrolling Implementation (RooCodeInc#6697) Update CHANGELOG.md Changeset version bump (RooCodeInc#6738) Stop making types private (RooCodeInc#6737) Delete bad changeset (RooCodeInc#6736) Revert "Changesets config tweak (RooCodeInc#6733)" (RooCodeInc#6735) feat: add OpenAI GPT OSS model to Cerebras providers (RooCodeInc#6734) Changesets config tweak (RooCodeInc#6733) v3.25.7 (RooCodeInc#6730) feat: add GPT-OSS 120b and 20b models to Groq provider (RooCodeInc#6732) feat: clean up task list in HistoryPreview and History components (RooCodeInc#6687) feat: add support for Claude Opus 4.1 (claude-opus-4-1-20250805) (RooCodeInc#6728) Revert "Extension bridge (RooCodeInc#6677)" (RooCodeInc#6729) Redesigned Task Header (RooCodeInc#6561) feat: code indexing support multiple folder similar with task history (RooCodeInc#6204) fix: prevent MCP server creation when setting is disabled (RooCodeInc#6613) Add the fireworks AI provider (RooCodeInc#6652) Extension bridge (RooCodeInc#6677) fix: prevent empty mode names from being saved (fixes RooCodeInc#5766) (RooCodeInc#5767) Bump @roo-code/types to v1.44.0 (RooCodeInc#6675) feat(ui): Make mode selection dropdowns responsive (RooCodeInc#6422) Add Z AI provider (RooCodeInc#6657) use assistantMessageParser class instead of parseAssistantMessage (RooCodeInc#5341) style: update highlightLayer style and align to textarea (RooCodeInc#6648) Remove 'Initial Checkpoint' terminology, use 'Checkpoint' consistently (RooCodeInc#6643) Use @roo-code/cloud from npm (RooCodeInc#6611) Bump @roo-code/types to v1.43.0 (RooCodeInc#6640) Changing checkpoint timing and ensuring checkpoints work (RooCodeInc#6359) feat: conditionally include reminder section based on todo list config (RooCodeInc#6411) Fix the UI for approving chained commands (RooCodeInc#6623) Bump @roo-code/types to v1.42.0 (RooCodeInc#6610) Task and TaskProvider event emitter cleanup + a few new events (RooCodeInc#6606) Changeset version bump (RooCodeInc#6579) Update contributors list (RooCodeInc#6506) Release v3.25.6 (RooCodeInc#6578) feat: set horizon-beta model max tokens to 32k for OpenRouter (RooCodeInc#6577) chore: bump @roo-code/types to v1.41.0 (RooCodeInc#6568) Cloud: support syncing provider profiles from the cloud (RooCodeInc#6540) Changeset version bump (RooCodeInc#6565) Release v3.25.5 (RooCodeInc#6564) Add Qwen 3 Coder from Cerebras (RooCodeInc#6562) fix: Fix VB.NET indexing by implementing fallback chunking system (RooCodeInc#6552) More tolerant search/replace match (RooCodeInc#6537) Clean up the auto-approve UI (RooCodeInc#6538) fix: linter not applied to locales/*/README.md (RooCodeInc#6477) Revert "Migrate evals database when deploying roo-code-website" (RooCodeInc#6525) Phase 1 website updates (RooCodeInc#6085) Cloud service cleanup callbacks / move to events (RooCodeInc#6519) Add Cerebras as a provider (RooCodeInc#6392) fix: handle Qdrant deletion errors gracefully to prevent indexing interruption (RooCodeInc#6296) Add auto-approved cost limits (RooCodeInc#6484) fix: LM Studio model context length (RooCodeInc#5075) (RooCodeInc#6183) fix: restore message sending when clicking save button (RooCodeInc#6487) Handle more variations of chaining and subshell command validation (RooCodeInc#6486) fix: improve Claude Code ENOENT error handling with installation guidance (RooCodeInc#5867) Skip interpolation for non-existent slash commands (RooCodeInc#6475) Changeset version bump (RooCodeInc#6474) Release v3.25.4 (RooCodeInc#6473) Revert experiments with nightly marketplace config (RooCodeInc#6472) feat: set horizon-alpha model max tokens to 32k for OpenRouter (RooCodeInc#6470) fix: Remove misleading task resumption message (RooCodeInc#5851) chore(deps): update dependency lint-staged to v16.1.2 (RooCodeInc#4965) fix(deps): update dependency recharts to v2.15.4 (RooCodeInc#4971) Fix: Kill button for execute_command tool (RooCodeInc#6457) feat: add zai-org/GLM-4.5-FP8 model to Chutes AI provider (RooCodeInc#6441) Update contributors list (RooCodeInc#6360) chore(deps): update dependency @changesets/cli to v2.29.5 (RooCodeInc#4936) Migrate evals database when deploying roo-code-website (RooCodeInc#6146) Miscellaneous cleanup (RooCodeInc#6453) feat: add symlink support for AGENTS.md file loading (RooCodeInc#6326) feat: diagnose nightly freeze by disabling contributes (RooCodeInc#6450) fix: resolve navigator global error by updating mammoth and bluebird dependencies (RooCodeInc#6363) Add pattern to support Databricks /invocations endpoints (RooCodeInc#6317) feat: enhance token counting by extracting text from messages using VSCode LM API (RooCodeInc#6424) feat: Add Task History Context to Prompt Enhancement (RooCodeInc#6343) Support new LLM provider: Doubao (RooCodeInc#6345) fix: Use separate changelog for nightly builds to prevent marketplace freezing (RooCodeInc#6449) feat: auto-refresh marketplace data when organization settings change (RooCodeInc#6446) feat: add SambaNova provider integration (RooCodeInc#6188) Changeset version bump (RooCodeInc#6438) chore: add changeset for v3.25.3 (RooCodeInc#6435) docs: clarify apply_diff tool descriptions to emphasize surgical edits (RooCodeInc#6278) fix(chat): Prevent input clearing when clicking chat buttons (RooCodeInc#6222) Remove event types mention from PR reviewer rules (RooCodeInc#6428) Update the auto-translate prompt (RooCodeInc#6430) Hide Gemini checkboxes on the welcome view (RooCodeInc#6415) Remove "(prev Roo Cline)" from extension title in all languages (RooCodeInc#6426) feat: add translation check action to pull_request.opened event (RooCodeInc#6393) Allow queueing images (RooCodeInc#6414) Add docs link for slash commands (RooCodeInc#6409) Update PR reviewer rules and mode configuration (RooCodeInc#6391) feat: increase Claude Code default max output tokens to 16k (RooCodeInc#6312) Changeset version bump (RooCodeInc#6390) Release v3.25.2 (RooCodeInc#6389) Fix zap hover (RooCodeInc#6388) fix: show diff view before approval when PREVENT_FOCUS_DISRUPTION is disabled (RooCodeInc#6386) Bump types to 1.40.0 (RooCodeInc#6387) Cloud: add organization MCP controls (RooCodeInc#6378) Changeset version bump (RooCodeInc#6385) Release v3.25.1 (RooCodeInc#6384) feat: add zai-org/GLM-4.5-Air model to Chutes AI provider (RooCodeInc#6377) Improvements to subshell validation (RooCodeInc#6379) docs: update docs extractor mode configuration and rules (RooCodeInc#6373) Changeset version bump (RooCodeInc#6362) Update contributors list (RooCodeInc#6302) fix: exclude rules-{slug} folder from mode export paths (RooCodeInc#6186) chore: add changeset for v3.25.0 (RooCodeInc#6358) feat: make PR reviewer rules generic (RooCodeInc#6357) Remove duplicated assignment (RooCodeInc#6355) fix(environment): Filter out non-text tab inputs (RooCodeInc#6350) Add .roomotes.yml (RooCodeInc#6346) Better command highlighting (RooCodeInc#6336) Command argument hints and change release engineer to a command (RooCodeInc#6333) Ensure form-data >= 4.0.4 (RooCodeInc#6332) feat(tools): add image support to read_file tool (RooCodeInc#5172) feat: Add experimental setting to prevent editor focus disruption (RooCodeInc#6214) feat: make task mode sticky to task (RooCodeInc#6177) feat: add ESC key handling for modes, API provider, and indexing settings popovers (RooCodeInc#6175) Parse mentions from all user inputs (RooCodeInc#6331) Support inserting mentions after a slash command (RooCodeInc#6327) feat: Update PR reviewer mode to use todo lists and GitHub CLI (RooCodeInc#6328) Restore PR reviewer (RooCodeInc#6324) Fix slash command highlighting (RooCodeInc#6325) fix: empty README.vscode.md to test marketplace freezing issue (RooCodeInc#6315) Add support for slash command frontmatter descriptions (RooCodeInc#6314) feat(mode-writer): add validation and cohesion checking for mode creation (RooCodeInc#6313) fix: normalize Windows paths to forward slashes in mode export (RooCodeInc#6308) Improve issue-writer mode initialization workflow (RooCodeInc#6311) Add a UI for managing slash commands (RooCodeInc#6286) feat: add prompt caching support for LiteLLM (RooCodeInc#5791) (RooCodeInc#6074) fix: remove demo GIF from VS Code README to prevent marketplace freezing (RooCodeInc#6305) fix: hide Test ErrorBoundary button in production mode (RooCodeInc#6216) Fix keyboard shortcuts for non-QWERTY layouts (RooCodeInc#6162) Revert "fix: optimize README for VS Code marketplace to prevent UI freezing" (RooCodeInc#6303) fix: optimize README for VS Code marketplace to prevent UI freezing (RooCodeInc#6275) Feat: Adding Gemini tools - URL Context and Grounding with Google Search (RooCodeInc#5959) feat: Add support for message queueing (RooCodeInc#6167) Update issue writer rules and workflow configurations (RooCodeInc#6230) feat: sync API config selector style with mode selector from PR RooCodeInc#6140 (RooCodeInc#6148) feat: Add search functionality to mode selector popup and reorganize layout (RooCodeInc#6140) Support for custom slash commands (RooCodeInc#6263) fix: prevent scrollbar flickering in chat view during content streaming (RooCodeInc#6266) test: update list-files test for fixed hidden files bug (RooCodeInc#6261) fix: add text wrapping to command patterns in Manage Command Permissions (RooCodeInc#6255) # fix: list_files recursive mode now works for dot directories (RooCodeInc#5176) fix: restore working settings link in command permissions tooltip (RooCodeInc#6253) feat: add markdown table rendering support (RooCodeInc#6252) Changeset version bump (RooCodeInc#6239) Kick off release (RooCodeInc#6238) Update contributors list (RooCodeInc#6110) Delete .changeset/kind-horses-sniff.md chore: add changeset for v3.24.0 (RooCodeInc#6237) Clean up some of the hugging face provider settings (RooCodeInc#6236) refactor: consolidate HuggingFace models API into providers/fetchers (RooCodeInc#6228) fix: add error message when no workspace folder is open for code indexing (RooCodeInc#6227) Hugging Face provider: add more details (RooCodeInc#6190) fix(chat): Cancel auto-approve timer when editing follow-up suggestion (RooCodeInc#6226) fix: prevent duplicate command patterns by trimming full command (RooCodeInc#6224) fix: restore list styles for markdown lists in chat interface (RooCodeInc#6095) feat: Add terminal command permissions UI to chat interface (RooCodeInc#5480) (RooCodeInc#5798) fix: respect maxReadFileLine setting for file mentions to prevent context exhaustion (RooCodeInc#6073) Add support for bedrock api keys (RooCodeInc#6132) Fix Ollama API URL normalization by removing trailing slashes (RooCodeInc#6079) fix: allow auto-approve checkbox to be toggled at any time (RooCodeInc#6061) feat: add efficiency warning for single SEARCH/REPLACE blocks in apply_diff (RooCodeInc#6055) Add type for allowing org members to view all tasks (RooCodeInc#6193) Expose default modes in @roo-code/types (RooCodeInc#6184) feat: add Google Analytics tag to marketing website (RooCodeInc#6179) fix: Fix Hugging Face provider setup not transitioning from welcome view (RooCodeInc#6173) feat: add confirmation dialog and proper cleanup for marketplace mode removal (RooCodeInc#6136) feat: add support for Agent Rules standard via AGENTS.md (RooCodeInc#5966) (RooCodeInc#5969) Update next.config.ts Fix/website logo theme persistence (RooCodeInc#6040) basic hugging face provider (RooCodeInc#6134) feat: Add settings to control diagnostic messages (RooCodeInc#5524) (RooCodeInc#5582) Changeset version bump (RooCodeInc#6131) chore: add changeset for v3.23.19 (RooCodeInc#6130) Roo Code Cloud Waitlist CTAs (RooCodeInc#6104) Smarter auto-deny (RooCodeInc#6123) Split commands on newlines (RooCodeInc#6121) Changeset version bump (RooCodeInc#6111) Update contributors list (RooCodeInc#5699) chore: add changeset for v3.23.18 (RooCodeInc#6109) debug: Add ErrorBoundary component for better error handling (RooCodeInc#5085) fix: resolve 'Bad substitution' error in command parsing (RooCodeInc#5743) Fix todo list toggle not working (RooCodeInc#6103) Use SIGKILL for command execution timeouts in the "execa" variant (RooCodeInc#6071) Changeset version bump (RooCodeInc#6092) Release v3.23.17 (RooCodeInc#6091) feat: add merge-resolver mode for intelligent conflict resolution (RooCodeInc#6090) fix: add Git installation check for checkpoints feature (RooCodeInc#3109) (RooCodeInc#5920) feat: add Qwen/Qwen3-235B-A22B-Instruct-2507 model to Chutes AI provider (RooCodeInc#6052) feat: add llama-4-maverick model to Vertex AI provider (RooCodeInc#5808) (RooCodeInc#6023) docs: clarify when to use update_todo_list tool (RooCodeInc#5926) fix: add case sensitivity mention to suggested fixes in apply_diff error message (RooCodeInc#6076) Add jump icon for newly created files (RooCodeInc#5738) Fix evals; broken by RooCodeInc#5865 (RooCodeInc#6065) Bugfix: Cloud: be more specific about session error codes (RooCodeInc#6051) feat: add moonshot provider (RooCodeInc#6046) Add todo list tool enable checkbox to provider advanced settings (RooCodeInc#6032) fix: add bedrock to ANTHROPIC_STYLE_PROVIDERS and restore vertex Claude model checking (RooCodeInc#6019) Update the max_tokens fallback logic in the sliding window (RooCodeInc#5993) fix: sort symlinked rules files by symlink names, not target names (RooCodeInc#5903) fix: properly distinguish between user cancellations and API failures (RooCodeInc#6025) fix: resolve global mode export not including rules files (RooCodeInc#5834) (RooCodeInc#5837) fix: add run parameter to vitest command in rules (RooCodeInc#5991) feat: auto-omit MCP content when no servers are configured (RooCodeInc#5889) fix: move context condensing prompt to Prompts section (RooCodeInc#4924) (RooCodeInc#5279) feat: add configurable timeout for evals (5-10 min) (RooCodeInc#5865) fix: enable export, share, and copy buttons during API operations (RooCodeInc#5324) (RooCodeInc#5849) fix: add character limit to prevent terminal output context explosion (RooCodeInc#5777) feat: Add Mistral embedding provider (RooCodeInc#5932) (RooCodeInc#5946) Changeset version bump (RooCodeInc#5953) Release v3.23.16 (RooCodeInc#5952) feat: enhance release engineer mode to include issue numbers and reporters in changelog (RooCodeInc#5940) feat: add batch limiting to code indexer (RooCodeInc#5891) feat: add global rate limiting for OpenAI-compatible embeddings (RooCodeInc#5854) feat: mark non-English translation files as linguist-generated (RooCodeInc#5943) Fix Docker port conflicts for evals services (RooCodeInc#5909) Changeset version bump (RooCodeInc#5935) chore: add changeset for v3.23.15 patch release (RooCodeInc#5934) Added YouTube to website footer (RooCodeInc#5929) Add command timeout allowlist with IPC support (RooCodeInc#5910) feat: add configurable delay for Go diagnostics to prevent premature error reporting (RooCodeInc#5863) feat: Add Issue Investigator mode and enhance Issue Writer mode (RooCodeInc#5913) Prevent completion with open todos (RooCodeInc#5716) feat: move marketplace icon from overflow menu to top navigation (RooCodeInc#5864) feat: Add description and whenToUse fields to custom modes in .roomodes (RooCodeInc#5862) fix: detect Claude models by name for API protocol selection (RooCodeInc#5840) Fix/issue fixer pr template (RooCodeInc#5839) Changeset version bump (RooCodeInc#5836) Log API-initiated tasks to a tmp directory (RooCodeInc#5833) refactor: remove orchestrator modes and update GitHub CLI usage (RooCodeInc#5832) Changeset version bump (RooCodeInc#5829) v3.23.13 (RooCodeInc#5827) Allow command execution timeout to be set via IPC task execution (RooCodeInc#5825) feat: Update Ollama UI to use text inputs (RooCodeInc#5818) Message edit/delete overhaul (RooCodeInc#5538)
- Increased message limit threshold from 500 to 750 messages - Only apply message limiting for very long conversations (>1000 messages) - Balanced viewport rendering (1500px top/bottom instead of 3000px/1000px) - Increased LRU cache capacity from 100 to 200 entries with longer TTL - Expanded viewport tracking window from 100 to 200 messages These changes prevent aggressive message filtering that was causing scroll position jumps while maintaining memory optimizations from PR #6697. Fixes #7026
Fix: Resolve Memory Leak in ChatView Virtual Scrolling Implementation
Root Cause
The memory leak was caused by React Virtuoso rendering ALL messages below the current viewport due to
increaseViewportBy={{ bottom: Number.MAX_SAFE_INTEGER }}. This prevented React from unmounting components and releasing memory, causing linear memory growth with conversation length.Changes Made
1. Fix infinite viewport rendering (line 1872)
increaseViewportByfrom{ top: 3_000, bottom: Number.MAX_SAFE_INTEGER }to{ top: 3_000, bottom: 1000 }2. Optimize LRU cache configuration (lines 183-186)
3. Implement viewport-based message filtering (lines 926-928, 982-985)
4. Add periodic cache cleanup (lines 467-480)
Impact
Note: Default VSCode + RooCode Scrolling up and down messages
roocode.default.broken.mp4
Note: VSCode Debugger + RooCode Scrolling up and down messages so memory is 3gig by default due to double vscode and the debugger
roocode.fix.mp4
Note: Videos are sped up 4x
Developed & Implemented by The Saturn Pipeline
Important
Fixes memory leak in
ChatViewby optimizing virtual scrolling and cache management.increaseViewportByinChatView.tsxfrom{ top: 3_000, bottom: Number.MAX_SAFE_INTEGER }to{ top: 3_000, bottom: 1000 }to limit rendering to 1000px below the viewport.modifiedMessagesto process only the most recent 500 messages and track the last 100 in the viewport.This description was created by
for 17b0cc0. You can customize this summary. It will automatically update as commits are pushed.