OpenCompletion

AI-powered code completion for VS Code using Ollama or OpenAI-compatible APIs.

Features

• Inline Code Completion: Copilot-style inline suggestions as you type
• Auto-Continue Writing: Automatically continues writing when you stop typing (configurable)
• Code Rewriting: Rewrite selected code with custom instructions
• Continue Writing: Manually trigger continuation from current position
• Flexible Backend: Works with Ollama, OpenAI API, or any compatible service
• Configurable Context: Choose between file-level or project-level context
• Smart Debouncing: Optimized request handling to reduce API calls
• File Type Management: Configure which file types enable auto-continue
• Smart Prompts: Adapts prompts for code vs. document files

Requirements

• VS Code 1.75.0 or higher
• A running LLM service compatible with OpenAI API format:
  • Ollama (recommended for local usage)
  • OpenAI API
  • Any other OpenAI-compatible endpoint

Installation

From VSIX (Development)

1. Build the extension:

   npm install
   npm run compile
   

2. Install in VS Code:

   • Open VS Code
   • Press Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (Mac)
   • Run "Extensions: Install from VSIX..."
   • Select the built .vsix file

Quick Start

Using with Ollama (Local)

1. Install and start Ollama:

   ollama serve
   

2. Pull a code model:

   ollama pull codellama
   

3. Configure OpenCompletion:

   • Click the OpenCompletion icon in the status bar
   • Or open Settings (Ctrl+,) and search for "opencompletion"
   • Set:
     • API URL: http://localhost:11434/v1
     • Model Name: codellama
     • API Key: (leave empty)

Using with OpenAI API

1. Configure OpenCompletion:
   • API URL: https://api.openai.com/v1
   • Model Name: gpt-4 or gpt-3.5-turbo
   • API Key: Your OpenAI API key

Configuration

All settings are available under opencompletion.*:

Usage

Inline Completion

The extension provides ghost text (grayed-out) completions as you type:

1. Automatic Trigger: Completions appear automatically when you pause typing
2. Press Tab to accept the suggestion
3. Press Esc to dismiss the suggestion

How it works:

• As you type, OpenCompletion waits for a brief pause (default: 300ms)
• Then requests a completion from your configured LLM
• The completion appears as ghost text at your cursor
• You can continue typing to dismiss it, or press Tab to accept

Auto-Continue Writing

When enabled, OpenCompletion will automatically trigger completions after you stop typing:

1. Enable auto-continue in settings or via status bar menu
2. Configure file types (default: markdown, plaintext)
3. Start writing in a supported file type
4. Stop typing for the configured delay (default: 2 seconds)
5. Ghost text completion appears automatically
6. Press Tab to accept or Esc to dismiss

Key difference from inline completion:

• Inline completion: Waits 300ms (short pause while typing)
• Auto-continue: Waits 2000ms (clear pause indicating you're thinking)

Adding Current File Type:

• Click status bar icon → "Manage Auto-Continue File Types"
• Click "Add [current-type] to Auto-Continue"

Managing File Types:

• Click status bar icon → "Manage Auto-Continue File Types"
• Edit the comma-separated list of language IDs

Commands

Access via Command Palette (Ctrl+Shift+P or Cmd+Shift+P):

• OpenCompletion: Rewrite Selection - Rewrite selected code with custom instructions
• OpenCompletion: Continue Writing - Continue writing from current position
• OpenCompletion: Toggle Enable/Disable - Toggle the extension on/off
• OpenCompletion: Open Settings - Open extension settings

Status Bar

The status bar icon shows the extension status:

• ⚡ [model-name] - Enabled and showing current model
• 🚫 OpenCompletion - Disabled

Click the icon to open the model switcher with options to:

• Switch Models: Choose from preset models or enter custom model name
• Toggle OpenCompletion: Enable/disable the extension
• Toggle Auto-Continue: Enable/disable auto-continue writing
• Manage File Types: Add/remove current file type for auto-continue
• Open Settings: Access full configuration

Architecture

src/
├── config/
│   └── settings.ts          # Configuration management
├── api/
│   ├── llmClient.ts          # LLM API client
│   └── promptBuilder.ts      # Prompt templates
├── completion/
│   ├── contextCollector.ts   # Context gathering
│   └── inlineProvider.ts     # Inline completion provider
├── edit/
│   └── docCommand.ts         # Edit commands
├── utils/
│   ├── logger.ts             # Logging utility
│   └── debounce.ts           # Debounce utilities
└── extension.ts              # Extension entry point

Performance Tips

1. Use Local Models: Ollama provides lower latency than cloud APIs
2. Adjust Debounce: Increase debounceMs to reduce API calls
3. Limit Context: Use currentFile scope for faster responses
4. Reduce Max Lines: Lower context.maxLines for quicker processing

Troubleshooting

Viewing Logs

OpenCompletion outputs detailed logs to help diagnose issues. Logs are NOT saved to files - they appear in VS Code's Output panel.

How to view logs:

1. Quick Method:

   • Press Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (Mac)
   • Run command: "OpenCompletion: Show Status & Debug Info"
   • Click "View Logs" button

2. Direct Method:

   • Open Output panel: Ctrl+Shift+U (Windows/Linux) or Cmd+Shift+U (Mac)
   • Or: Menu → View → Output
   • Select "OpenCompletion" from the dropdown in the upper-right corner

3. When Errors Occur:

   • Error notifications have a "View Logs" button
   • Click it to jump directly to the logs

What logs show:

• 🔍 When completions are triggered
• 📡 API requests and responses
• ✅ Successful completions
• ❌ Errors with detailed messages
• ⏱️ Performance timing information

See DEBUG_GUIDE.md for detailed debugging instructions.

No Completions Appearing

Step 1: Check Status

• Run command: "OpenCompletion: Show Status & Debug Info"
• Verify: Enabled = true, Model and API URL are correct

Step 2: Check VS Code Settings

1. Open Settings (Ctrl+,)
2. Search: editor.inlineSuggest.enabled
3. Ensure it's set to true

Step 3: View Logs

1. Open Output panel → Select "OpenCompletion"
2. Try triggering a completion
3. Look for 🔍 and ❌ symbols in the logs

Common causes:

1. OpenCompletion is disabled
2. VS Code inline suggestions are disabled
3. API service not running (Ollama)
4. Model not downloaded
5. Network/firewall blocking connection
6. Conflicting AI extension (GitHub Copilot, etc.)

Slow Completions

Check performance:

1. Open logs and check timing between 📤 and 📥 messages
2. If taking >5 seconds, the model is too slow

Solutions:

1. Increase debounceMs to 500-1000ms (reduce API calls)
2. Reduce context.maxLines to 100 (less context)
3. Use a smaller/faster model
4. Switch to local Ollama instead of cloud API
5. Use currentFile scope instead of project

Network/API Errors

Error notifications will show:

• "Request timeout" → Increase timeout setting or use faster model
• "Network error" → Check if Ollama is running (ollama serve)
• "HTTP 404/500" → Verify model name and API URL

Test your API connection:

# For Ollama
curl http://localhost:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "your-model", "messages": [{"role": "user", "content": "test"}]}'

If curl fails, fix your API setup before troubleshooting the extension.

Detailed Debugging

For comprehensive troubleshooting steps, log interpretation, and advanced debugging:

→ See DEBUG_GUIDE.md

The debug guide includes:

• Complete log location instructions
• Step-by-step diagnostic procedures
• Log symbol meanings (🔍 📡 ✅ ❌)
• Common error patterns and solutions
• Testing procedures
• How to report issues with proper logs

Development

Build

npm install
npm run compile

Watch Mode

npm run watch

Debugging

1. Open the project in VS Code
2. Press F5 to launch Extension Development Host
3. Test the extension in the new window

License

MIT

Credits

Built following best practices for VS Code extensions and inspired by GitHub Copilot's UX.