Brute Force Vector Similarity Search

Implement a brute-force nearest neighbor search over a collection of embedding vectors.

Given a list of stored vectors, a query vector, and k, return the indices of the k most similar vectors by cosine similarity.

This is the retrieval step in a naive RAG (Retrieval-Augmented Generation) pipeline: given a user query embedding, find the most relevant document embeddings.

Input: vectors=[[0.1,0.9],[0.8,0.2],[0.5,0.5]], query=[0.2,0.8], k=2 Output: [0, 2] (indices sorted by similarity)