PlasmaLLM

A KDE Plasma 6 widget that provides a chat interface to OpenAI-compatible LLM endpoints. Integrates system information gathering, markdown rendering, and shell command execution directly from the desktop.

PlasmaLLM is designed for quick questions and system tasks right from your panel — not as a full-featured AI chat application. Think of it as a handy assistant for looking things up, running commands, and getting help with your system without leaving the desktop.

Features

• Chat with any OpenAI-compatible API (Ollama, LM Studio, OpenAI, Anthropic, Groq, and more)
• System-aware: automatically gathers hardware and OS info for contextual responses
• Interactive command blocks: run, copy, save, or open suggested shell commands in a terminal
• Markdown rendering for assistant responses
• Auto-run and auto-share modes for agentic workflows
• Chat history auto-save
• Configurable provider presets, temperature, max tokens, and custom system prompts

Requirements

• KDE Plasma 6.0+
• Qt 6

No external dependencies — uses only Qt and KDE Plasma APIs.

Installation

From GitHub Releases (recommended)

Download the latest .plasmoid file from the Releases page, then install it:

plasmapkg2 --install PlasmaLLM-*.plasmoid

Or right-click your desktop → Add Widgets → Get New Widgets → Install from Local File.

Then restart Plasma:

plasmashell --replace &

From Source

Clone the repository and run the install script:

git clone https://github.com/joshuaeroman/plasmallm.git
cd plasmallm
./install.sh
plasmashell --replace &

For development (symlinks the package directory so changes apply on Plasma restart instead of requiring a reinstall):

./install.sh --dev

To uninstall:

./install.sh --remove

Building a .plasmoid from Source

package.sh zips the package/ directory into a versioned .plasmoid file (a standard KDE widget archive):

./package.sh
# Creates e.g. PlasmaLLM-x.y.z.plasmoid

Requires python3 (to read the version from metadata.json), zip, and GNU gettext for I18N.

Configuration

Right-click the widget and open Configure.... Settings include:

• Provider: choose a preset (Ollama, LM Studio, OpenAI, etc.) or enter a custom endpoint
• Model: select from available models via "Fetch Models" or type manually
• API Key: required for cloud providers
• Temperature / Max Tokens: control response behavior
• Auto-save chat history: saves conversations to ~/PlasmaLLM/chats/
• Auto-run commands: automatically executes shell commands from LLM responses
• Auto-share command output: sends command output back to the LLM (enables agentic loops when combined with auto-run)
• Custom system prompt: appended to the built-in prompt with highest precedence

Support

If you find this useful, consider donating to KDE — the project that makes all of this possible.

License

This project is licensed under the GNU General Public License v2.0 or later.

Vibe-coded with 🤖