⚠️ Prerelease — Not suitable for production use.
This is an early-stage proof of concept. APIs, features, and behaviour will change without notice. Use at your own risk.
AskTux is a desktop application that lets you ask natural-language questions about using Linux and get step-by-step instructions from a local or remote LLM. It collects information about your system (distro, desktop environment, shell, hardware) and includes it as context so answers are tailored to your exact environment.
- Read-only assistance — AskTux never executes commands. It provides instructions only.
- Streaming responses — LLM output streams token-by-token, rendered as formatted HTML in real time via WebKitGTK.
- System-aware context — Automatically detects your distro, desktop environment, display server, shell, and hardware to give relevant answers.
- Local & remote providers — Supports Ollama (local) and any OpenAI-compatible API (OpenAI, Groq, DeepSeek, Google AI, GitHub Models, etc.) with saved provider profiles and per-provider model lists.
- External styling — Appearance is controlled by a CSS file that can be edited post-compilation.
- Customisable system prompt — Edit the prompt template in Settings to change AskTux's behaviour.
- GTK 4 with gtkmm 4.0 (C++ bindings)
- libcurl
- WebKitGTK 6.0 (
libwebkitgtk-6.0-dev) — for HTML rendering - libcmark (
libcmark-dev) — CommonMark markdown parser - SQLite 3 (
libsqlite3-dev) — configuration storage - Meson — build system
- A running Ollama instance (default) or an OpenAI-compatible API endpoint
See INSTALL.md for full dependencies, build, and install instructions. Quick start:
sudo apt install meson libgtkmm-4.0-dev libcurl4-openssl-dev \
libwebkitgtk-6.0-dev libcmark-dev libsqlite3-dev
meson setup build
meson compile -C build
./build/src/asktuxSettings are stored in ~/.config/asktux/config.db (SQLite). You can edit it
directly with any SQLite tool, or use the Settings dialog inside the app.
| ID | Provider | Base URL | Default model |
|---|---|---|---|
| 1 | Ollama | http://localhost:11434 |
llama3.1:8b |
| 2 | OpenAI | https://api.openai.com/v1 |
— |
| 3 | Groq | https://api.groq.com/openai/v1 |
— |
| 4 | DeepSeek | https://api.deepseek.com/v1 |
— |
| 5 | Google AI | https://generativelanguage.googleapis.com/v1beta/openai |
— |
| 6 | GitHub Models | https://models.inference.ai.azure.com |
— |
- Provider ID 1 (Ollama) uses a local model and does not need an API key.
- All other providers require an API key (entered in Settings).
- Each provider remembers the last-used model (
last_modelcolumn). - The model dropdown is populated from saved models and (for Ollama) live queries to the running instance.
| Key | Default | Description |
|---|---|---|
provider |
1 (Ollama) |
Currently selected provider ID |
system_prompt |
(built-in template) | Custom system prompt with {distro}, {desktop}, etc. |
./build/src/asktux --debugEnables detailed timing and trace logging (ScopedTimer, request dumps, Ollama internal breakdown). Useful for diagnosing performance issues.
The markdown output is styled via data/style.css, installed to /usr/share/asktux/style.css. Edit this file to customise fonts, colours, and layout — no recompilation required.
See ROADMAP.ai for planned improvements. Highlights include:
- Smarter model-pull logic (only pull if not already available)
- Chat history for follow-up questions
- Read-only tool use (local man pages, installed packages)
- OpenAPI-compatible endpoint support
- Distribution packages (.deb, .rpm, AppImage)
MIT