Mini LLM Flow is a streamlined framework designed specifically for leveraging large language models (LLMs) in a more efficient and focused manner. By distilling the framework down to just 100 lines of code, it removes unnecessary complexities and highlights essential high-level programming paradigms. This minimalist approach centers around a nested directed graph, which facilitates task decomposition and LLM-driven processes such as branching and recursion for decision-making. The framework supports agent-based architectures, making it ideal for applications requiring dynamic task management and resource allocation. Additionally, it seamlessly incorporates Retrieval-Augmented Generation (RAG), enhancing the flexibility and functionality of LLM implementations. With Mini LLM Flow, users can easily add more complex features, providing a solid foundation for scalable and adaptable AI solutions.