I’ve been attempting to take the Moqui services and search on them with an embedding (basically rag for Moqui services), and I found this cool tool called nomic.ai. Here’s a chart of the embeddings: inspiring-laplace | Atlas Map.
Instead of embedding just the straight up xml, I’m using an llm to summarize the service and embed that. That seems to provide better performance. You can see that the chart clumps services that are similar to each other. This is exactly what you want for an embedding.
Anyways, I thought this was cool enough to share. For the (admittedly somewhat messy) code for doing this, you can see updates here: GitHub - acetousk/getXml