Interview trainer

Your company's support chatbot answers questions using help-center articles. To figure out which article is relevant, both the user's question and every article get turned into embeddings — lists of numbers where similar meanings end up pointing in similar directions. So "my package never arrived" and the shipping-policy article land close together, even though they share no words. When a question comes in, the bot needs to find the articles whose embeddings are most similar to the question's. That lookup is your piece to build.

Write topK(query, docs, k) (Python: top_k). query is a list of numbers (the question's embedding). docs is a list of objects like { id: "shipping", vector: [...] }. Score each doc with cosine similarity: the dot product of the two vectors, divided by the product of their lengths. Return the k highest-scoring docs as { id, score } objects, sorted best first. You can assume all vectors are the same length and non-zero, and no two scores tie.