opencog

OpenCog Docker Containers

This directory contains docker configurations for some of OpenCog's projects. Some of the notable images include:

• opencog/atomspace -- Core AtomSpace only.
• opencog/atomspace-py -- AtomSpace plus many Python tools.
• opencog/learn -- Learning subsystem.
• opencog/lang-pairs -- Tabulate and visualize word pairs from text corpora.

Older, obsolete images can be found in the archive directory. If someone sent you here, and you can't find what you were told about, then it is probably in the archive.

Docker setup

1. Install docker. On Debian/Ubuntu, apt install docker.io. Be sure to add your userid to the /etc/group file, like so: sudo usermod -aG docker $USER. Log out and log back in, so that this new membership can become active. You can also say newgrp docker at an existing shell prompt..

2. Alternately, follow the docker.com instructions for setting up docker. They will want you to create an account. But this is not needed. If you do create an account, your probably need to say sudo docker login.

3. Pull opencog images from dockerhub by running ./docker-build.sh -a These images are automatically rebuilt fairly regularly, and should provide reasonably fresh, working code. The pull should take no more than a few minutes. The image sizes range from 1.13GB for the opencog/opencog-deps image to 1.42GB for the opencog/learn image.

4. If the absolute latest is needed, then the docker images can be built locally on your machine. The build takes from 20 minutes to an hour, depending on your machine and network speeds. Use ./docker-build.sh -h to get a list of available options.

Overview

List the current set of docker images with the command docker images. The most notable include:

• opencog/opencog-deps -- A base operating system image, with most required dependencies installed.

• opencog/atomspace -- An image containing the core AtomSpace. The AtomSpace is needed by all other subsystems.

• opencog/atomspace-py -- The AtomSpace, plus a large collection of Python tools commonly used for machine learning and DL/NN work.

• opencog/learn -- An image containing the language-learning subproject.

The above four images don't actually "do anything"; they just provide baseline software installs.

Actual demos, which "do actual stuff", are listed below. These need to be built individually; they cannot be downloaded pre-made. Go to the relevant directory, and follow the instructions there.

• opencog/lang-pairs -- Word-pair counting and visualization. Demo of the very first step of the language-learning project.

Docker Cheat Sheet

• docker images
• docker rmi <image-hex>
• docker ps -a
• docker create --name foo --hostname bar -it opencog/atomspace
• docker start -i <container-name>
• docker stop <container-name>
• docker rm <container-name>
• `docker container cp :
• docker attach <container-name>

Notes

1. Both tmux and byobu are installed, so you can use either for multiple windows/panes.

Maintainer Notes

The docker images are built and published every Saturday night, using a github workflow. However, I think (not sure) that those images are cached, and thus will contain old code, if the datestamps in the Dockerfiles aren't bumped. Not sure just right now. The default is building against Ubuntu 24.04.

Bonus

There are some "bonus" containers, which might be useful:

• claude-code -- Image with AtomSpace plus Anthropic's Claude Code.
• ollama -- Image with AtomSpace plus Ollama (Caution: this image will be 11.8 GB in size; this is the sum of the atomspace (1.4 GB), llama (5GB) and qwen3.8b (5.2GB).