# Flight505 Skill Forge Windsurf Performance Tuning > Windsurf Performance Tuning - Skill: `tomevault-io/flight505-skill-forge-windsurf-performance-tuning` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/flight505-skill-forge-windsurf-performance-tuning` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/flight505-skill-forge-windsurf-performance-tuning/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/flight505-skill-forge-windsurf-performance-tuning --- # Windsurf Performance Tuning ## Overview Optimize Windsurf's indexing engine, Cascade context building, Supercomplete latency, and overall IDE responsiveness. Most performance issues stem from indexing too many files or missing exclusion patterns. ## Prerequisites - Windsurf IDE installed - Understanding of workspace indexing - Access to Windsurf settings ## Instructions ### Step 1: Optimize Workspace Indexing The indexing engine is the #1 performance factor. Exclude everything that isn't source code: ```gitignore # .codeiumignore — aggressive exclusion for large codebases # Build artifacts node_modules/ dist/ build/ .next/ .nuxt/ .output/ .svelte-kit/ coverage/ .cache/ # Package manager pnpm-lock.yaml package-lock.json yarn.lock # Generated code *.min.js *.min.css *.bundle.js *.chunk.js **/*.map generated/ __generated__/ # Binary and media files *.png *.jpg *.gif *.svg *.ico *.mp4 *.woff *.woff2 *.ttf *.eot # Data files *.sqlite *.db *.sql.gz *.csv # Python __pycache__/ *.pyc .venv/ venv/ .tox/ *.egg-info/ # Go vendor/ # Rust target/ ``` ### Step 2: Open Focused Workspaces ``` # DON'T: Open entire monorepo root windsurf ~/monorepo/ # Indexes ALL packages = slow # DO: Open specific service directories windsurf ~/monorepo/apps/web/ # Only indexes web app windsurf ~/monorepo/packages/api/ # Only indexes API # Each Windsurf window gets its own focused AI context # Cascade performs better with fewer, relevant files ``` ### Step 3: Tune IDE Settings ```json // settings.json — performance-focused configuration { // Indexing limits "codeium.indexing.maxFileSize": 524288, // Reduce extension overhead "extensions.autoUpdate": false, "extensions.autoCheckUpdates": false, // Editor performance "editor.minimap.enabled": false, "editor.renderWhitespace": "none", "editor.bracketPairColorization.enabled": false, // File watcher optimization "files.watcherExclude": { "**/node_modules/**": true, "**/dist/**": true, "**/.git/objects/**": true, "**/build/**": true }, // Search exclusions "search.exclude": { "**/node_modules": true, "**/dist": true, "**/coverage": true, "**/*.min.js": true } } ``` ### Step 4: Optimize Cascade Usage ```markdown ## Cascade Performance Tips 1. Start fresh sessions for new tasks - Long conversations accumulate context, slowing responses - Click + in Cascade panel to start new conversation 2. Use @ mentions instead of describing files - BAD: "Look at the auth service file" - GOOD: "@src/services/auth.ts fix the token refresh logic" 3. Scope prompts narrowly - BAD: "Refactor the authentication system" - GOOD: "Extract JWT validation from src/middleware/auth.ts into src/services/jwt.ts" 4. Use Chat mode for questions, Write mode for edits - Chat mode is lighter — doesn't build full edit context - Switch modes based on intent 5. Choose the right model for the task - SWE-1 Lite: fast, lightweight questions (no credits) - SWE-1: standard coding tasks - SWE-1.5 / Claude: complex multi-file tasks (worth the wait) ``` ### Step 5: Memory Management ```markdown ## Reduce Memory Usage 1. Close unused editor tabs - Each open file adds to memory footprint - Cmd/Ctrl+K Cmd/Ctrl+U: close unused tabs 2. Limit extensions - Each extension consumes memory - Disable extensions you don't actively use - Check: Extensions > "@enabled" to review 3. Disable unused Supercomplete languages - Settings > codeium.autocomplete.languages - Disable for languages you don't write 4. Monitor process memory - Help > Open Process Explorer - Look for extensions consuming >200MB ``` ### Step 6: Benchmark Your Setup ```bash #!/bin/bash set -euo pipefail echo "=== Windsurf Performance Check ===" # Workspace size FILE_COUNT=$(find . -type f -not -path '*/node_modules/*' -not -path '*/.git/*' | wc -l) echo "Indexed files: ~$FILE_COUNT" # Config status echo "Has .codeiumignore: $([ -f .codeiumignore ] && echo 'YES' || echo 'NO (SLOW!)')" IGNORE_LINES=$(wc -l < .codeiumignore 2>/dev/null || echo 0) echo "Ignore patterns: $IGNORE_LINES" # Recommendations if [ "$FILE_COUNT" -gt 5000 ]; then echo "WARNING: >5000 files. Add more patterns to .codeiumignore" fi if [ "$IGNORE_LINES" -lt 5 ]; then echo "WARNING: Few ignore patterns. Add build artifacts, binaries, lock files" fi ``` ## Error Handling | Issue | Cause | Solution | |-------|-------|----------| | Indexing never completes | Too many files | Add `.codeiumignore`, open subdirectory | | Cascade timeout | Complex prompt + large context | Narrow scope, start fresh session | | High memory usage | Too many extensions/tabs | Close tabs, disable unused extensions | | Supercomplete lag | Large file open | Split large files, add to maxFileSize limit | | IDE freezes | Extension conflict | Safe Mode: `windsurf --disable-extensions` | ## Examples ### Quick Performance Fix ```bash # Create minimal .codeiumignore if missing [ -f .codeiumignore ] || cat > .codeiumignore << 'EOF' node_modules/ dist/ build/ .next/ coverage/ *.min.js *.map EOF ``` ### Reset Indexing ``` Command Palette (Cmd/Ctrl+Shift+P): "Codeium: Reset Indexing" ``` ## Resources - [Windsurf Context Awareness](https://docs.windsurf.com/context-awareness/overview) - [Windsurf Ignore](https://docs.windsurf.com/context-awareness/windsurf-ignore) - [Autocomplete Tips](https://docs.windsurf.com/autocomplete/tips) ## Next Steps For cost optimization, see `windsurf-cost-tuning`. --- > Source: [flight505/skill-forge](https://github.com/flight505/skill-forge) — distributed by [TomeVault](https://tomevault.io).