This app implements the following text generation providers using any freely-available large language model: Free prompt, Summarize, Headline, Context Write, Chat, and Reformulate.

You can connect to the watsonx.ai service provided by IBM or to a self-hosted cluster running IBM Software Hub and related services.

To avoid task processing execution delay, setup at 4 background job workers in the main server (where Nextcloud is installed). The setup process is documented here: https://docs.nextcloud.com/server/latest/admin_manual/ai/overview.html#improve-ai-task-pickup-speed

Positive:

• The provided foundation models are freely available, and thus can be ran on-premises

Negative:

• The software for training and inference of models is proprietary, limiting modifications to the API or other functionality
• Some foundation models are trained on data that is not freely available, limiting the ability to fine tune them

Learn more about the Nextcloud Ethical AI Rating in our blog.

• Support for agency features (i.e. IBM watsonx as chat with tools provider)
• Support for additional models without complete API functionality (see: https://www.ibm.com/watsonx/developer/get-started/models/)
• Support for IBM Cloud Pak for Data Platform API (for identity management on self-hosted instances)
• Support for more than 100 models deployed in IBM watsonx.ai
• Ability to select an IBM Cloud location from a dropdown list (as a workaround, enter the location's API URL manually)

There is a "Artificial intelligence" admin settings section where you can:

• Choose whether you use an IBM-hosted watsonx.ai instance or another remote service
• Set a global API key and cloud resource for the Nextcloud instance
• Configure default models and quota settings

There is a "Artificial intelligence" personal settings section to let users set their personal API key and cloud resources. Users can also see their quota information there.