yaarakas/AI-Powered-Vector-Search-Engine ? reverse-engineered prompt

Build me a C# .NET app that does semantic search over a set of documents using AI generated embeddings.

I want it to take a user query, turn that into a vector, compare it against stored document vectors, and return the most relevant matches. Then have it generate a short natural language explanation or answer based on the results, so it feels like a smart search experience instead of just keyword matching.

Please include file based data loading and retrieval, use REST APIs for embedding generation and response generation, and measure how long the search takes so I can see performance. Keep it practical and easy to run locally, with clear handling for structured files like JSON. If you need to look up current API or .NET docs online, go ahead.