# LLM Model Switcher

A simple web-based tool to switch between different `llama.cpp` model configurations on a headless AI server, featuring a retro-futuristic terminal interface.

## How it works

The tool manages `llama.cpp` configurations by manipulating symbolic links and interacting with `systemd`:

1.  **Configurations**: It scans `/etc/llama.cpp.d` for `.conf` files. Each file should contain the environment variables for a specific model.
2.  **Active Link**: The systemd service for `llama.cpp` is expected to read its parameters from `/etc/llama.cpp.conf`.
3.  **Switching**: When you select a model in the web UI:
    *   The tool updates the symlink at `/etc/llama.cpp.conf` to point to the selected file in `/etc/llama.cpp.d`.
    *   It restarts the `llama.cpp` systemd service to apply the changes.
4.  **Monitoring**: The tool provides a real-time view of the last 100 lines of the `llama.cpp` service logs using `journalctl`.

## Features

*   **Model Management**: Easily swap between different model configurations.
*   **Service Logs**: Integrated log viewer showing the last 100 lines of the systemd service (newest first).
*   **Retro UI**: A "Cyberpunk CRT" aesthetic with scanlines, phosphor glow, and screen flicker effects.
*   **Zero Dependencies**: The backend uses standard Python libraries, and the frontend uses Preact via ESM—no build step or `npm install` required.

## Requirements

*   Python 3.7+ (uses `subprocess.run` with `capture_output`).
*   `sudo` privileges (to modify `/etc`, restart system services, and read `journalctl` logs).
*   A `systemd` service named `llama.cpp`.

## Usage

Run the script with `sudo`:

```bash
sudo python3 main.py
```

By default, the server will be available at `http://127.0.0.1:7330`.

### Command Line Options

*   `--host`: Set the listening address (e.g., `0.0.0.0` for all interfaces). If a path is provided (e.g., `/tmp/model.sock`), it will listen on a Unix socket. (Default: `127.0.0.1`)
*   `--port`: Set the TCP port. (Default: `7330`)

## Project Structure

*   `main.py`: The Python backend using `http.server` and `subprocess`.
*   `static/`: Contains the frontend assets (HTML/CSS/JS).
    *   `index.html`: Preact-based application.
    *   `style.css`: Custom "CRT Terminal" styling and animations.