Spectyra Proxy

Local proxy server that provides an OpenAI-compatible endpoint, routing requests through Spectyra for automatic optimization. Works with GitHub Copilot, Cursor, Claude Code, and other coding assistants.

Features

• ✅ OpenAI-compatible API - Works with any tool that uses OpenAI API
• ✅ Automatic optimization - Routes through Spectyra for 40-65% token savings
• ✅ Real-time dashboard - Web UI showing savings and stats
• ✅ Configuration management - Easy setup via web dashboard
• ✅ BYOK support - Use your own API keys
• ✅ Multi-provider support - OpenAI, Anthropic, Gemini, Grok

Installation

Install via npm (Recommended)

npm install -g spectyra-proxy

Then start with:

spectyra-proxy

Note: The npm package contains only compiled JavaScript, not source code. Your code remains private.

Option 2: Install from Source (Development)

If you need to modify the code or the npm package isn't available:

Quick Start

1. Start the proxy:

   If installed via npm:

   spectyra-proxy

2. Configure via dashboard:

   • Open http://localhost:3002 in your browser
   • Enter your Spectyra API key
   • Enter your provider API key (OpenAI, Anthropic, etc.)
   • Select provider and optimization settings
   • Click "Save Configuration"

3. Configure your coding tool:

   • Set OPENAI_API_BASE=http://localhost:3001/v1
   • Or configure in your tool's settings

Usage

With GitHub Copilot

1. Set environment variable:

export OPENAI_API_BASE=http://localhost:3001/v1

2. Restart VS Code/Copilot

With Cursor

1. Open Cursor Settings
2. Set API base URL to: http://localhost:3001/v1
3. Restart Cursor

With Claude Code

1. Configure Claude Code to use custom API endpoint
2. Set endpoint to: http://localhost:3001/v1

Dashboard

Access the dashboard at: http://localhost:3002

Features:

• Real-time savings statistics
• Recent request history
• Configuration management
• Live updates every 2 seconds

Configuration

Environment Variables

• PROXY_PORT - Proxy server port (default: 3001)
• DASHBOARD_PORT - Dashboard port (default: 3002)
• SPECTYRA_API_URL - Spectyra API URL (default: https://spectyra.up.railway.app/v1)

Configuration File

Configuration is saved to .spectyra-proxy-config.json in the proxy directory.

Fields:

• spectyraKey - Your Spectyra API key
• providerKey - Your provider API key (BYOK)
• provider - Provider name (openai, anthropic, gemini, grok)
• path - Optimization path (code, talk)
• optimizationLevel - Optimization level (0-4)

API Endpoints

Proxy Endpoint

• POST /v1/chat/completions - OpenAI-compatible chat endpoint
• POST /v1/messages - Anthropic-compatible messages endpoint

Configuration

• GET /config - Get current configuration (without keys)
• POST /config - Update configuration

Statistics

• GET /stats - Get usage statistics and savings

Health Check

• GET /health - Check proxy status

How It Works

Your Coding Tool (Copilot/Cursor/etc)
  → Local Proxy (localhost:3001)
    → Spectyra API (optimization)
      → Provider API (OpenAI/Anthropic)
        → Optimized Response
          → Your Tool

Benefits:

• Transparent optimization
• No code changes needed
• Real-time savings tracking
• Works with any OpenAI-compatible tool

Troubleshooting

Proxy not starting

• Check if ports 3001 and 3002 are available
• Try different ports via environment variables

Configuration not saving

• Check file permissions in proxy directory
• Ensure .spectyra-proxy-config.json is writable

Tool not connecting

• Verify proxy is running: curl http://localhost:3001/health
• Check tool's API base URL setting
• Ensure tool supports custom API endpoints

No savings showing

• Verify configuration is correct
• Check Spectyra API key is valid
• Check provider API key is valid
• Look at proxy console for errors

Development

# Watch mode (auto-restart on changes)
npm run dev

# Build for distribution
npm run build

License

MIT