Using Docker, clone the repo and run "docker compose up". It will take a few minutes first time(the docker images can take up to 5GB), But generally in 20 seconds it'll be up, and you can go to http://localhost:3000 to try it.
About
AI inference is expensive, and running AI based systems mostly require many GPU hours and a large bill. I present BioRag, a multimodal biometric system for identity verification, which is a low-resources and a cpu based system. The technology is based on feature extraction of face and iris images, as well as speaker audio files, and use existing vector database capabilities to be able to search through milions of records in less than a second, utilizing only CPU of a laptop. Because of the multi-modality(meaning, using multiple verification methods - face, iris, voice), each of the identification methods can be less precise and use a simple CPU-inference ML model. When combining the methods together, the system will still provide state-of-the-art performance, just like much more expensive solutions which rely on GPUs.