Shader Linux Glsl

The hyperbolic app displays a fullscreen shader with a video texture coming from camera. It is a tiling of the hyperbolic plane. It can be controlled from a LLM through MCP protocol (the app is an MCP server).

The tiling is defined by a group of symmetries and they can be changed.

The hyperbolic plane can either be represented as a disk or as a plane:

The video device can be passed as argument when launching the app:

hyperbolic /dev/video5

The app needs to read the files fragment_shader.glsl and vertex_shader.glsl. They should be in the working directory. You should run the app from the root folder otherwise the files won't be found by the app:

build\Debug\hyperbolic

The app exposes an MCP server on port 8100 and can be controlled by a LLM. Here is an example of a chat to change the settings of the hyperbolic app:

To use the app with a LLM you need:

• A LLM server. The server.sh shows how to laumch llama-server from llama.cpp project. An Open AI API compatible server will work
• A chat client that can talk to the open AI compatible interface and to MCP servers. For instance Jan

Typical commands provided by the tool:

• change background to black
• change edge to green
• stop animation
• change geometry to plane
• change symmetry to 1
• reset settings

To build

git clone --recurse-submodules https://github.com/christophe0606/shader_linux_glsl.git

Some libraries are required for the UI

sudo apt install libglew-dev
sudo apt install libglfw3-dev
sudo apt install libgles2-mesa-dev

If you don't want to build but just use the excutable:

sudo apt install libglfw3
sudo apt install libglew2.2
sudo apt install libgles2

Then use cmake to build.

AI

Except the shader that is coming from an old experiment by the author, other parts of this demo have been generated a lot by chat GPT.