Building a Lightweight Vector Search Engine from Scratch Using Pure Python: An Alternative for Resource-Constrained Edge Devices

When building AI applications like vector search or recommendation systems, heavy libraries like NumPy or Pandas are typically used, which can consume significant memory on small edge devices. This project explores an alternative by implementing a functional Vector Search Engine using only pure Python loops. The system implements core mathematical operations, such as Dot Product and Cosine Similarity, from scratch without any external dependencies. By avoiding third-party packages, this approach significantly reduces initialization memory usage compared to NumPy-based systems. While large-scale datasets may still require compiled libraries, this zero-dependency method is highly efficient and secure for small-scale applications and micro-systems with strict memory constraints.