Query the ancient Greek corpus using natural language. Passages are retrieved by embedding similarity, then reranked with a cross-encoder for final ordering.