"Uffe" is a project in our AI lab where we are exploring  long-term memory and self-awareness features in Large Language Model (LLM) agents. Uffe is a highly efficient sub-100 lines LLM agent designed for performance and compactness. Equipped with long-term memory, Uffe can retain and recall information over extended interactions. This agent is distinct in that all of its code, contained in the file uffe.py, is directly injected into the system prompt, ensuring full transparency and integration with the system's operations.

​

Uffe is designed to utilize your terminal, meaning that whatever resources and commands you have access to, Uffe can also employ. This ensures that Uffe can leverage all available tools to perform its tasks efficiently. Uffe is a real doer—once given a task, it takes initiative and completes it without needing to ask for permission.