suyashpradhan/blue-green-re-embed-simulator

This project is a learning exercise that shows how a technique called blue green deployment can be applied to the process of re-generating embeddings, which are the numeric representations that machine learning models use to understand and compare pieces of text or data. Instead of teaching the concept only in theory, the project builds a small interactive simulator in the browser so you can see the idea in action. Blue green deployment normally means running two identical versions of a service at once, an old one and a new one, and switching traffic over to the new version only once it is ready, so users never experience downtime. Here that same idea is applied to embeddings: the simulator runs the re-indexing process on separate instances in parallel and then switches over, rather than replacing embeddings in place while the system is live. The simulator is built with Next.js and React for the web interface, styled with Tailwind CSS, and animated with Framer Motion. It uses Transformers.js to generate embeddings directly in the browser using local inference, so it does not need to call an external AI API to work. The whole project is written in TypeScript. Features include a real time visualization of the re-indexing and traffic switching process, and interactive controls so a learner can experiment with different deployment scenarios themselves. The author describes it as part of a personal learning journey into deployment strategies for data pipelines, safe embedding model updates, and running machine learning inference on the client side rather than on a server. This is a small, educational side project best suited to developers curious about deployment patterns or client side machine learning, and it currently has 1 star.