LOKI AI Assistant
Privacy-first, fully local VS Code extension providing Copilot-style AI assistance.
Privacy & Security
LOKI runs 100% locally. No data leaves your machine.
- All AI requests go to
localhost:11434 (Ollama)
- No telemetry, no analytics, no external API calls
- No cloud services required
- Your code stays on your machine
Features
1. Design Mode
Transform high-level UI designs into production-ready code instantly.
- Figma-like Interface: Visual canvas with drag-and-drop components (Navbar, Hero, Grid, Footer).
- One-Click Generation: Convert visual layouts into React/Tailwind components.
- Senior Architect Backend: Generates highly optimized, accessible, and responsive code.
2. Inline Ghost Text Completions
Copilot-style suggestions as you type:
- Debounced for performance (300ms default)
- Cancellable on new input
- Context-aware (reads surrounding code)
Interactive AI chat with:
- Model selector dropdown
- Streaming responses
- Code blocks with "Insert" buttons
- Slash commands:
/explain, /refactor, /tests, /fix
4. Code Actions
Right-click menu and Ctrl+. quick fixes:
- Explain Code
- Refactor Code
- Generate Tests
- Fix Bugs
5. Autonomous Agent
Build entire projects with natural language:
- "Build a React portfolio app"
- "Create a Node.js REST API"
- Step-by-step execution with user approval
- Automatic backups to
.loki-backups/
Requirements
Ollama (Required)
LOKI uses Ollama for local LLM inference.
Install Ollama:
# Linux
curl -fsSL https://ollama.com/install.sh | sh
# macOS
brew install ollama
# Windows
# Download from https://ollama.com/download
Start Ollama:
ollama serve
Download a model:
# Recommended for coding
ollama pull codellama
# Alternative options
ollama pull deepseek-coder
ollama pull mistral
ollama pull qwen2.5-coder
Configuration
Open VS Code Settings (Ctrl+,) and search for "LOKI":
| Setting |
Default |
Description |
loki.ollamaUrl |
http://localhost:11434 |
Ollama server URL |
loki.model |
codellama |
Model for chat and code actions |
loki.completionModel |
codellama |
Model for inline completions |
loki.enableCompletions |
true |
Enable inline ghost text |
loki.completionDebounce |
300 |
Debounce delay (ms) |
loki.autonomousAutoApprove |
false |
Auto-approve file changes (DANGEROUS) |
Keyboard Shortcuts
| Shortcut |
Action |
Ctrl+Shift+L |
Open LOKI Chat |
Ctrl+Shift+; |
Toggle Inline Completions |
Ctrl+. |
Show Code Actions (with selection) |
Safety Guarantees
Autonomous Mode Safety
- Backups: All file modifications create backups in
.loki-backups/
- Approval: Each potentially destructive step requires user confirmation (unless disabled)
- Logging: All actions logged to
loki.log in workspace root
- Rollback: Automatic rollback on failure
What LOKI Will NOT Do
- Send data to external servers
- Execute shell commands without approval
- Modify files outside your workspace
- Access network resources (except localhost Ollama)
Project Structure
vscode-extension/
├── extension.ts # Entry point, command registration
├── src/
│ ├── agent.ts # Autonomous agent, code actions
│ ├── chatViewProvider.ts # Chat sidebar UI
│ ├── completionProvider.ts # Inline completions
│ ├── codeActionProvider.ts # Quick fix menu
│ └── ollamaClient.ts # HTTP client for Ollama
├── package.json # Extension manifest
└── tsconfig.json # TypeScript config
Development
# Install dependencies
npm install
# Compile
npm run compile
# Watch mode
npm run watch
# Package extension
npm run package
Debug the extension:
- Open in VS Code
- Press F5
- A new VS Code window opens with LOKI loaded
License
MIT License - See LICENSE
Contributing
Contributions welcome! Please ensure:
- No external service dependencies
- All AI remains local
- User approval for destructive actions
- Comprehensive error handling
Built with privacy in mind. Your code, your machine, your control.
