3 ms·
From what I can tell, the AI part resides at gregory-ms, not in the github repo. If the platform is, indeed, a standalone NLP system, then that's marvelous. If,
by gardenfelder 4y ago
From what I can tell, the AI part resides at gregory-ms, not in the github repo. If the platform is, indeed, a standalone NLP system, then that's marvelous. If, instead, it relies on API calls to a server, what will those cost? What drives financial sustainability of an otherwise fantastic project?
- brunoamaral 4y agoGardenfelder, thank you for asking. Gregory is meant to be installed as a server that fetches the new articles and runs pipelines that make sure the information is consistent. There are two machine learning algorithms that can be trained once you have a good dataset. There is also an API that can be used to feed a website or newsletter, for example. Right now, the caveat is that each field of research requires it’s own instance of the software.
- Llamamoe 4y agoSo I imagine that deploying this is nontrivial and not just easily casually done on low-end consumer hardware?
- brunoamaral 4y agoIt can be run locally, and uses docker. There is a setup.py script that guides you through the install process and asks for some environment variables. The ideal setting is to install on a server with 4Gb of RAM.