@rfeigel is it possible that you might have to download the whole workflow group
To actually create a vector store you will have to use an embedding model that you can either provide as a .gguf file to the GPT4All Embeddings Connector:
Or you can let the node download one for you. The embedding model is one to prepare the FAISS Vector Store. To actually use the store you will need an LLM from the GPT4All collection. The example used in the Learnathon used ChatGPT in this place.
EDIT: In case someone wonders if you encounter a problem with a model download behind a proxy or you deliberately want to work offline (to be sure…) there are currently some quirks: