Care documentation is an essential but time-consuming part of nursing practices. We present a first prototype to support care workers by generating summaries from audio recorded during standard nursing interactions. The audio is transcribed with Automatic Speech Recognition (ASR), and a summary is generated by a Large Language Model (LLM), both running locally. For evaluation, we recorded four mock care interaction scenarios with a training manikin. We compare different local LLMs with GPT-3.5 and GPT-4. We find that most of the important topics relevant to care documentation were present in the resulting summaries.