⚡ Copilot Pulse — Intelligent Model Router & Auto-Escalation Engine for VS Code

Copilot Pulse is an autonomous AI router and auto-escalation engine for Visual Studio Code. It analyzes every prompt in real-time, calculates a multidimensional complexity score (0–10), and routes the request to the most cost-effective and capable AI model tier. If a lighter model gets stuck or encounters errors, Pulse automatically escalates to a stronger frontier model.

One Chat Agent. Every Tier. Zero Manual Model-Switching. Up to 85% Token & Credit Savings.

🚀 Key Features

• 🎯 Dynamic 3-Tier Smart Routing

  • Light Tier (Fast & Cheap): GPT-4o-mini, Claude 3.5 Haiku, Gemini Flash — for syntax lookups, docstrings, boilerplate, and basic edits.
  • Medium Tier (Balanced): Claude 3.5 Sonnet, GPT-4o — for refactoring, test generation, and multi-file debugging.
  • Heavy Tier (Frontier Reasoning): Claude 3.7 Sonnet (Thinking), o1 / o3-mini, GPT-4.5 — for full architecture design, algorithms, race conditions, and complex system planning.

• ⬆️ Autonomous Auto-Escalation

  • Automatically detects repeated tool loops, context window pressure, or execution errors on cheaper models.
  • Retries on a stronger frontier model without interrupting your workflow or losing conversation context.

• 📊 Real-Time Interactive Savings Dashboard

  • Live breakdown of tokens saved, cost efficiency (50%–85% savings), model distribution, and prompt complexity score breakdowns.

• 🌐 Cloud Intelligence & Pro Subscription (Lemon Squeezy)

  • Free Tier: 50 smart-routed queries per day.
  • Pro Tier: Unlimited real-time cloud routing, instant Lemon Squeezy license validation, and premium multi-model escalation.

🛠️ How to Use Copilot Pulse

Method 1: Chat Participant @copilotpulse (Recommended)

1. Open the GitHub Copilot Chat panel in VS Code (Ctrl + Alt + I on Windows/Linux or Cmd + Shift + I on Mac).
2. Type @copilotpulse at the start of your message:

   @copilotpulse refactor this function to handle async errors and write unit tests
   

3. Sticky Session: Once you invoke @copilotpulse, all follow-up prompts in the same thread automatically route through Pulse without retyping the tag.

Method 2: Open the Interactive Dashboard

1. Click the Zap (⚡) icon on the VS Code Activity Bar (left sidebar) OR press Cmd + Shift + P / Ctrl + Shift + P and search:

   Copilot Pulse: Open Dashboard
   

2. In the dashboard, you can:
   • Monitor real-time routing decisions and complexity scores.
   • View your estimated cost and token savings.
   • Activate your Copilot Pulse Pro license key.
   • Click Upgrade to Pro to unlock unlimited queries via Lemon Squeezy.

Method 3: Bring Your Own Keys (BYOK Mode)

If you prefer using your own OpenAI or OpenRouter API keys directly:

1. Open VS Code Settings (Cmd + , or Ctrl + ,).
2. Search for Copilot Pulse and set:
   • copilotPulse.apiKey: Your OpenRouter or OpenAI API key (sk-or-... or sk-...).
   • copilotPulse.baseUrl: https://openrouter.ai/api/v1 (or your custom OpenAI-compatible endpoint).

⚙️ Configuration Reference (settings.json)

You can customize Copilot Pulse via the VS Code Settings UI or directly in .vscode/settings.json:

{
  // Cloud Routing & Licensing Endpoint
  "copilotPulse.cloudEndpoint": "https://bold-tooth-3b12.anand-madhav.workers.dev",
  "copilotPulse.useCloudRouting": true,
  
  // Pro Subscription License Key (from Lemon Squeezy)
  "copilotPulse.licenseKey": "LS-xxxx-xxxx-xxxx",
  
  // Model Tier Customization
  "copilotPulse.cheapModel": "openai/gpt-4o-mini",
  "copilotPulse.strongModel": "openai/gpt-4o",
  
  // Routing Thresholds (0.0 to 10.0 scale)
  "copilotPulse.threshold": 6.5,
  "copilotPulse.lowThreshold": 4.0,
  
  // Auto-Escalation Controls
  "copilotPulse.escalation.enabled": true,
  "copilotPulse.escalation.maxEscalationsPerConversation": 2,
  "copilotPulse.maxAgentTurns": 25,
  
  // Upgrade Checkout URL
  "copilotPulse.upgradeUrl": "https://copilotpulse.lemonsqueezy.com/checkout/buy/8a364ccb-07cf-4f76-ad74-c0e6613bb913"
}

⌨️ Command Palette Shortcuts

🔒 Privacy & Security

• Zero Prompt Leakage: Your prompts are evaluated for complexity metadata and sent securely to your configured AI model provider.
• No Telemetry Logging: No proprietary source code is logged or stored.

📄 License & Publisher

• Publisher: Copilot Pulse
• License: MIT License
• Support & Issues: GitHub Issues