Copilot Prompt Optimizer + Graphify

A VS Code extension that dramatically reduces token usage in GitHub Copilot Chat while maintaining 100% prompt accuracy. Includes a built-in 3D interactive knowledge graph of your codebase, smart context injection, node-level exclusion controls, context-strength recommendations, and full session export for zero-loss context handoff.

Features

Token Optimization

• Smart prompt compression — removes filler words, deduplicates instructions, and compresses verbose patterns
• Zero accuracy loss — preserves semantic meaning while reducing token count by 20–60%
• Three optimization levels — Light, Moderate, and Aggressive
• Real-time token counting — see exact token savings per prompt
• Selection optimizer — right-click any selected text in the editor to optimize it in-place

3D Knowledge Graph (Built-in, Zero Setup)

• Auto-builds on activation — scans your workspace in the background; no CLI, no config files
• Interactive 3D visualization — drag to rotate, scroll to zoom, click to focus any node
• Relevant nodes highlighted — green nodes show which parts of your codebase match your query
• Node exclusion — click any node → modal with Exclude from Context / Re-include / Cancel; excluded nodes are greyed out and persisted across sessions
• Context-strength indicator — after every /graphify query you get a 🟢/🟡/🔴 strength score plus suggested nodes to add or remove
• Word-boundary + bigram scoring — accurate context selection with less false-positive noise
• Auto-updates — graph rebuilds when you save, create, or delete files

Session Export (/export)

• Full context export panel — tabbed view with a 3D session graph and a structured text document
• Context Graph tab — 3D force-graph showing sessions (purple), keywords (green), and code references (amber) with hover tooltips
• Text Export tab — complete markdown with knowledge graph, optimization history, and graph context; ready to paste into any new chat
• Copy Last 3 / Copy All — paste at the top of a new Copilot Chat to continue with zero context loss

Activity Bar Panel

• Visual history — all original vs optimized prompts side-by-side with sparklines
• Session statistics — total tokens saved, average reduction, best reduction
• Context quality score — shows how well the graph context matched each session

Status Bar Integration

• Live token savings — shows percentage and absolute token reduction
• Session stats tooltip — hover for detailed statistics
• Quick toggle — enable/disable optimization

Chat Commands

VS Code Commands (Command Palette)

Keyboard shortcut: Cmd+Shift+Alt+O (Mac) / Ctrl+Shift+Alt+O (Windows/Linux) — optimize selected text.

How to Use /export in a New Chat

1. Run @optimize /export (or Prompt Optimizer: Export Session Context from the Command Palette).
2. Click Copy All in the export panel.
3. Open a new Copilot Chat window.
4. Paste as your very first message — the AI will have full workspace context and optimization history with zero re-explanation.

How to Exclude a Node from Context

1. Run @optimize /graphify <your query> — the 3D graph opens automatically.
2. Click any node in the graph.
3. Choose Exclude from Context in the modal.
4. The node turns grey and is skipped in future queries. Click it again → Re-include to restore it.

Knowledge Graph

Supported Languages

TypeScript, JavaScript, Python, Java, Kotlin, Go, Rust, C/C++, C#, Ruby, Swift, Vue, Svelte, Markdown / MDX

What Gets Indexed

• Classes, interfaces, functions, methods, and arrow-function exports
• Import/export relationships between files
• Python classes and functions
• Java / Kotlin / Go / Rust / C# classes and methods
• Markdown headings, frontmatter titles, and keyword tags

Performance

• Max 500 files scanned per build
• Minified files, node_modules, dist, out, build, media automatically excluded
• Debounced rebuild (3 s) on file changes to avoid thrashing

Configuration

Optimization Pipeline

1. Normalize whitespace — remove redundant spaces and line breaks
2. Remove filler — strip "please", "I would like you to", polite wrappers
3. Compress patterns — "Write a function that" → fn:
4. Deduplicate — remove repeated instructions
5. Abbreviate — "JavaScript" → JS, "authentication" → auth
6. Inject graph context — prepend relevant codebase symbols as compressed context
7. Final cleanup — ensure clean, minimal output

License

MIT