OS environment

Annif is being developed on Ubuntu machines. However, it should work on any Linux distribution with some work. As for other OS:s, Annif itself is quite portable as it only requires a Python 3.9 (or newer) environment, although some of the optional extensions require native code libraries which may be challenging to install in some environments.

Annif is also available as a pre-built Docker image that can be run on all operating systems supporting Docker; see the Usage-with-Docker wiki page.

Resource requirements

• At least 8GB of RAM; 16GB or more is recommended if you are planning on hosting several models, e.g. different vocabularies and/or languages.
• Annif itself requires up to 2GB disk space with all optional features installed. Corpora and models can be large, thus 100GB disk space is recommended
• Multiple CPU cores (ideally 4+) are useful, e.g. for building fastText and Omikuji models which can take advantage of parallel processing