A general-purpose low-latency application to serve cognitive tasks, built with egui.
The most reliable way to install CogTask is by installing Cargo through rustup and compiling the binaries locally.
Install Cargo:
bash
$ curl https://sh.rustup.rs -sSf | sh
Build binaries (choose one):
- Stable binaries from crates.io:
bash
$ cargo install cog-task@1.0.0-beta [--features=...]
- Nightly binaries from github:
bash
$ cargo install --git https://github.com/menoua/cog-task [--features=...]
By default (no features), this package should compile and run out-of-the-box on a reasonably recent macOS or Linux distribution. Some types of actions however depend on features that can be enabled during installation. These features are not enabled by default because they rely on system libraries that might not be installed on the OS out-of-the-box.
Currently, there are 4 main features that can be enabled:
1. audio -- enables the Audio action via the ALSA sound library.
2. gstreamer -- enables the Stream and Video actions via the gstreamer backend.
3. ffmpeg (incomplete) -- enables the Stream and Video actions via the ffmpeg backend.
4. full -- a shorthand to enable the previous three features.
Examples:
- Stable binaries with full support:
bash
$ cargo install cog-task@1.0.0-beta --features=full
- Nightly binaries with audio and gstreamer support:
bash
$ cargo install --git https://github.com/menoua/cog-task --features=audio,gstreamer
Some features depend on certain libraries that might not come preinstalled on your OS. In these cases, before building the binaries with said features enabled, you need to first install the requirements:
On linux, requires installing pkg-config and ALSA, e.g.:
bash
$ sudo apt install pkg-config libasound2-dev
On macOS, requires installing gstreamer, e.g.:
bash
$ brew install gstreamer gst-plugins-base gst-plugins-good gst-plugins-bad gst-plugins-ugly gst-libav gst-rtsp-server
On linux, requires installing libgstreamer, e.g.:
bash
$ sudo apt install libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev libgstreamer-plugins-bad1.0-dev gstreamer1.0-plugins-base gstreamer1.0-plugins-good gstreamer1.0-plugins-bad gstreamer1.0-plugins-ugly gstreamer1.0-libav gstreamer1.0-tools gstreamer1.0-alsa gstreamer1.0-pulseaudio libavfilter-dev libavdevice-dev
On linux, requires installing pkg-config and libavutil, e.g.:
bash
$ sudo apt install pkg-config libavfilter-dev libavdevice-dev
- NOTE: Although it is not a requirement at compile time, to be able to use the ffmpeg backend, you need to have the ffmpeg library installed on the system during runtime.
This crate installs two binaries -- cog-launcher and cog-server.
cog-launcher: A launcher that provides a graphical interface to find and load tasks from disk.
cog-server /path/to/task: Used to run a specific task by providing the path to its directory. cog-launcher runs this binary when starting a task, so make sure both binaries are in the same directory.
For example, to run the Basic task in this repo, you would do the following:
bash
$ git clone https://github.com/menoua/cog-task
$ cog-server cog-task/example/basic
Alternatively, you can run:
bash
$ cog-launcher
Then use the leftmost control icon to load the example/basic/ directory. Or, you can use the second button to open the parent example/ directory which contains all the example tasks within. The former, directly runs cog-server on the chosen task. The latter, displays a list of all tasks located in the chosen directory, which can be started by clicking the corresponding button.
Version 0.2.0 has gone through a massive overhaul, transitioning from the GUI framework of iced to egui. The transition was done to solve a screen update skipping issue (which it did). There have been other pros and cons too:
egui.egui's Glow backend supports image/SVG.view and update calls allowed redesigning block architecture (the dependency graph) into an action tree. This change makes it very difficult to design a buggy task, and significantly simplifies task definition style. It slightly limits the task design flexibility, but it's worth it. This change also comes with an increased overhead since update/view calls traverse the entire active subset of the tree, instead of jumping to the end nodes. However, the tree overhead is generally low compared to action-specific overheads, so that's not a huge deal.