A prototypal AI assistant for Visual Studio Code

A customizable, extensible AI coding assistant for Visual Studio Code, developed as my Bachelor's Degree thesis project

Overview

This is a prototypal AI assistant extension for Visual Studio Code, inspired by GitHub Copilot. Built with extensibility in mind, it provides intelligent code completions, inline editing capabilities, and a full-featured chat interface.

✨ Features

🔮 Inline Code Completions

Context-aware code suggestions as you type. Configurable idle delay, built-in rate limiting, and support for a dedicated completion model.

✏️ Inline Code Editing

A series of commands that allows to transform the selected code in the current editor (or the whole document if no selection is made).

Command	Description	Default Shortcut
Edit	Apply custom edits to selected code	Ctrl+Shift+E
Refactor	Intelligently refactor selected code	Ctrl+Shift+R
Generate Docs	Add documentation comments to code	—

💬 Chat Panel

Sidebar panel for conversational interactions. Supports file attachments for context, optional web search grounding, conversation history, and markdown rendering with syntax highlighting.

🚀 Getting Started

Prerequisites

• Visual Studio Code v1.74.0 or higher
• A Google AI API key (Get one here)

Installation

1. Clone this repository
2. Run npm install to install dependencies
3. Press F5 to launch the extension in a new Extension Development Host window

Configuration

Configure the extension via VS Code Settings (Ctrl+,) under "AI Assistant". The minimum required settings are the AI provider (default is "google"), the API key and the base (default) model to use for the various extension features.

Note: Currently only Google AI models are supported, but the extension's architecture is designed to easily add other AI providers if needed.

Example minimal configuration:

{
  "ai-assistant.provider": "google",
  "ai-assistant.apiKey": "YOUR_API_KEY",
  "ai-assistant.baseModel": "gemini-2.0-flash", // this is an example
}

Additional Settings

Setting	Type	Default	Description
provider	string	"google"	AI service provider
apiKey	string	—	API key for the AI services
baseModel	string	—	Default model for AI operations
chatModel	string	—	Model for chat (falls back to baseModel)
temperature	number	—	Controls response randomness (0 = deterministic)
maxOutputTokens	number	—	Maximum length of generated responses
topK	number	—	Limits token selection to top K candidates
topP	number	—	Nucleus sampling threshold
enableInlineCompletions	boolean	true	Enable/disable inline suggestions
inlineCompletions.model	string	—	Model for inline completions (falls back to baseModel)
inlineCompletions.timeBetweenRequests	number	2000	Minimum ms between completion requests
inlineCompletions.idleDelay	number	600	Ms to wait after typing before requesting

🎨 Customization

Custom Instructions

Personalize the assistant's behavior by creating an instructions file in your workspace:

your-project/
├── .ai/
│   └── agentInstructions.md    ← Your custom instructions
├── src/
└── ...

The contents of .ai/agentInstructions.md will be included in all AI interactions, allowing you to:

• Define coding style preferences
• Specify project-specific conventions
• Set response format guidelines
• Include domain-specific context

Example:

# Project Guidelines

- Use TypeScript strict mode conventions
- Prefer functional programming patterns
- Always include JSDoc comments for public APIs
- Follow the existing naming conventions in the codebase

🛠️ Building the Extension

# Install dependencies
npm install

# Compile TypeScript
npm run compile

# Watch for changes
npm run watch

# Run linter
npm run lint

📸 Screenshots