🛸 OllaPilot
OllaPilot is a lightning-fast, privacy-first VSCode extension that provides GitHub Copilot-style inline code suggestions powered by Ollama. Keep your code completely local, or connect to a powerful remote server—the choice is yours.
✨ Features
- Inline Ghost Text: Seamless, real-time code completion as you type.
- Local & Remote Support: Connect to
localhost for maximum privacy or point to your beefy remote server URL for better models.
- Model Switching On-The-Fly: Easily switch between your installed Ollama models right from the VSCode Status Bar.
- Toggle Suggestions: Quickly turn suggestions On or Off via the Command Palette.
- Smart Context: Reads your surrounding code structure for smarter completions using Fill-in-the-Middle (FIM) techniques.
🚀 Getting Started
Prerequisites
Install Ollama.
Pull a model tailored for coding. Note: Base coding models work best for Fill-in-the-Middle (FIM) autocomplete. Instruct/Reasoning models (like qwen or deepseek-r1) might return chat formatting or "thinking" blocks instead of raw code.
Recommended models:
starcoder2qwen2.5-coder (Base model preferred over Instruct)deepseek-coder
ollama pull qwen2.5-coder
Installation (Development)
Currently, the extension is in development mode. To run it:
- Clone or open the repository in VSCode.
- Open terminal and run to install dependencies:
npm install
- Press
F5 to launch the Extension Development Host.
- Open any file in the new window and start typing!
⚙️ Configuration
You can configure OllaPilot through VSCode's standard Settings UI (Ctrl + , or Cmd + ,).
| Setting |
Type |
Default |
Description |
ollapilot.ollamaUrl |
string |
http://localhost:11434 |
The API URL to your local or remote Ollama instance. (e.g., https://my-cloud-ollama.com:11434) |
ollapilot.enabled |
boolean |
true |
Globally enable or disable code suggestions. |
🎮 Usage
The Status Bar
Look at the bottom-right corner of your VSCode editor. You will see the OllaPilot Indicator.
[✓] Ollama: qwen2.5-coder 🟢 Suggestions are Active.[⊘] Ollama: No Model 🔴 Suggestions are Disabled or waiting for config.
Click the Status Bar Item to open a quick-pick menu of all models currently pulled on your Ollama server. Choose a different model and OllaPilot switches instantly.
Command Palette
Press Ctrl+Shift+P (or Cmd+Shift+P on Mac) and search for:
Ollama Suggestions: Toggle On/Off: Master switch to pause suggestions.Ollama Suggestions: Select Model: Same as clicking the status bar.
🛠️ How it Works
OllaPilot utilizes VSCode's InlineCompletionItemProvider API. As you type, the extension reads the text directly before and after your cursor. It debounces the keystrokes (wait 500ms) to prevent overwhelming your local machine, and then dispatches a highly-tailored prompt to your Ollama API.
Fly through your codebase with OllaPilot. 🛸
