I love this concept! I have always believed that the old methodologies used in NLP and statistics can be better and faster than new LLM technologies like embeddings, depending on the scenario. Will the code be open-sourced someday? I'm thrilled to learn from it.