Creating Embeddings in KNIME with DeepSeek

KNIME Extensions
1

Hello KNIME Community,

I want to use DeepSeek to create some embeddings, but I’m unsure as to how I would go about that. I see in the FAISS Vector Store Creator node that you can use other embeddings if you specify them from somewhere:

“By default, the node embeds the selected documents using the embeddings model, but it is also possible to create the vector store from existing embeddings by specifying the corresponding embeddings column in the node dialog.”

I know that DeepSeek does not have an embedding model for use right now, but I’ve seen people create embeddings from a DeepSeek model on Hugging Face in Colab with Torch and vLLM.

Is this possible using DeepSeek in KNIME? I’ve also seen a thread where someone developed a workflow to do this using GPT4All ( workflow linked here )

I was wondering if there’s a way to accomplish this in KNIME without doing this locally. Ideally, I’d like to avoid purchasing anything for this, so would I need to run something like the GPT4All workflow I linked?

Thanks in advance!

2

@benfin I do not think you can use DeepSeek as Embedding with KNIME nodes. But you could try different embeddings like this:

image
image1611×1526 307 KB

3

@mlauber71 Thanks for your response! I believe I linked one of your previous workflows from another thread in my post haha. I’ve looked into this a bit more and I do think that GPT4All will be the best solution for this aside from what I currently have.

My current solution is using the OpenAI embedding model to create the embeddings, creating and querying my vector store, and then using those embeddings in a DeepSeek prompt (using the DeepSeek nodes, not in GPT4All).

I may try doing something with your workflow as an example. Thanks again!