Part 4/9:
Lightweight and Affordable: Enabling individual users to deploy their own Raven nodes on consumer hardware like Raspberry Pis and USB speakers for as little as $40-$50.
Flexible and Scalable: Using microservices to compartmentalize functionalities like embedding, speech recognition, and text-to-speech, which can be expanded or replaced as needed.
Open Source and Cloud-Ready: While nodes can run locally on commodity hardware, some services—such as large language models—will utilize cloud providers like OpenAI or open-source alternatives like NLP Cloud with NeoX.
Technical Architecture and Microservices
David shared a prototype of the initial microservice: an embedding engine based on Google's Universal Sentence Encoder v5. This lightweight service: