DeepTab

Bring-your-own-key AI code completion for VS Code. Use your own API keys with DeepSeek, OpenAI, or Gemini for private, zero-cost-while-idle inline completions.

Features

• Ghost-text inline completions — like Copilot, but with your own keys
• Multi-provider — DeepSeek, OpenAI (GPT-4o, o1, o3), and Gemini
• Privacy-first — your code goes directly to the model provider you choose, no intermediate server
• Configurable — model, temperature, max tokens, custom endpoints

Setup

1. Get an API key from your preferred provider:
   • DeepSeek
   • OpenAI
   • Google AI
2. Open VS Code settings and set deeptab.apiKey, or create a .env file with DEEPTAB_API_KEY=your-key
3. Optionally change deeptab.model (default: deepseek-chat)

Configuration

Setting	Default	Description
deeptab.apiKey	—	API key for your model provider
deeptab.model	deepseek-chat	Model ID (auto-detects provider)
deeptab.endpoint	—	Custom API endpoint override
deeptab.temperature	0.2	Model temperature (0–2)
deeptab.maxTokens	256	Max tokens per completion

Model Routing

DeepTab detects the provider from your model name:

• Starts with gpt-, o1, o3 → OpenAI API
• Starts with gemini → Google Gemini API
• Everything else → DeepSeek API

Requirements

• VS Code 1.85.0 or later

License

MIT