PyllamaUI for VS Code 🧠🦙
A lightweight offline AI assistant inside VS Code powered by Python + Ollama.
Access local LLMs directly from your editor — no cloud, no telemetry, no internet required.
🌟 About
PyllamaUI (VS Code Edition) is a VS Code extension that lets you chat with locally hosted LLMs using Python as the backend and Ollama as the model runner.
It's fully offline, privacy-friendly, and designed for low-resource systems.
✨ Features
- 💬 Chat UI inside a VS Code panel
- 🧠 Interact with local Ollama models (
llama3, mistral, etc.)
- 📝 Full Markdown support with bold text, italics, headers, lists, and blockquotes
- 🎨 Syntax highlighting for code blocks (Python, JavaScript, JSON, Bash, and more)
- 📋 One-click copy buttons for all code blocks with proper indentation
- 🔌 Uses VS Code's WebView for integrated GUI
- 🚫 Fully offline – no internet required, all dependencies bundled
- 📁 All user data processed locally
- ⚡ Real-time markdown rendering as responses stream in
🛠️ Requirements
{{ ... }}
- VS Code 1.75+
- Python 3.10+
- ollama
✅ Model Recommendations
| Use Case |
Model Name |
Approx. Size |
Description |
| 📝 Text Chat |
tinyllama |
~600 MB |
Lightweight text model |
| 💻 Coding Help |
deepseek-coder |
~700 MB |
Designed for code generation |
| ⚡ All-in-One |
gemma:3b (aka gemma3n) |
~5.5 GB |
Great for both chat & coding (Google DeepMind) |
✅ If unsure, just install gemma3n for the best all-around experience.
To run a model:
ollama run gemma3n
ollama run tinyllama
ollama run deepseek-coder
📌 Special thanks to Ollama for making local LLMs accessible to all.
🎨 Markdown Features
PyLlamaUI now supports full markdown rendering with:
Text Formatting
- Bold text using
**text**
- Italic text using
*text*
- Headers using
# ## ###
- Blockquotes using
>
- Lists (ordered and unordered)
Code Support
- Inline code using backticks:
console.log("hello")
- Code blocks with syntax highlighting:
def example():
print("This code has syntax highlighting!")
return True
Copy Functionality
- Every code block includes a Copy button
- Preserves proper indentation and formatting
- Works with all supported languages: Python, JavaScript, JSON, Bash, and more
- Fully offline - no external dependencies required
Real-time Rendering
- Markdown is rendered as the AI response streams in
- Smooth, responsive interface
- All processing happens locally in VS Code
