Reduce OpenAI costs in RAG pipelines by optimizing FAISS retrieval and chunking to minimize token usage.