This seems like a really nice framework for feeding documentation into llms. Would be curious to use same dataset without fine-tuned llm for RAG over same data

This seems like a really nice framework for feeding documentation into llms. Would be curious to use same dataset without fine-tuned llm for RAG over same data. Then one could immediately make use of building the dataset and then measure gains from training.