What is the first programming that comes to mind when thinking about machine learning? Java, obviously!

I am participating in a kind of hackaton, and I am trying to build a RAG system that can extract relevant information from the du website.
Because I love Java, Spring and Neo4J I decided to make my life easier by using Java and Spring instead of Python and Neo4J for vector embeddings instead of stuff like quarkus.
The experience so far has been bumpy but I learned a lot about the internal workings of the Neo4J vector storage and I enjoyed how much boilerplate I did not have to write compared with Python.

I will demo where am with the project: at the moment I successfully crawled the du website and stored the documents in the neo4j vector store. I am trying to use llama 3.1 to automatically come up for tags for the webpages, so that I can then pre-filter the pages by tag instead of just using cosine similarity, which has been quite poor so far

I expect to show:

• how to setup Spring AI to create embeddings using ollama and llama 3.1
• how to setup neo4j so that it can be used as a vector store
• how to setup a chat model so that you can “talk” to llama 3.1

The main takeaway from this is how little code I had to write to make all this work, leveraging Spring autoconfiguration, and how cosine similarity can be not as good as expected with real world data