I’m a few weeks into playing with n8n on the side. I’ve worked with EAI tools and integrations way back, and more recently with platforms like Zapier, Workato, and Zoho Flow. It’s interesting to see how each one tackles the low-code/no-code automation space.
n8n has my attention lately – thanks to its deeper developer features: code nodes, AI nodes, LangChain support, and more. (I’m sure the others are evolving too – I just haven’t revisited them recently.)
I’ve been messing around with an n8n instance on Railway, pulling in tools like a cheap OpenAI API access (looking at you, GPT-3.5 Turbo), Gemini, SerpAPI, and Qdrant – the last two inspired by some Udemy labs I’ve been exploring. Qdrant, in particular, caught my attention. I didn’t just want to use a vector database – I wanted to understand how it works.
If you’re curious too, check out Vector Database Fundamentals by Paulo Dichone on Udemy. It’s a fantastic crash course. He’s also on YouTube if you don’t have Udemy access.
Tools like ChatGPT don’t run on vector DBs directly, but vector embeddings are at the core of how they work. When you want memory, semantic search, or smarter context handling – vector databases start to matter.
Paulo uses great examples throughout – an early one about picking a campsite based on meaningful similarity really clicked for me and set the tone for the whole course.
Taking directly from his summary, this course covered:
- Foundations of Vector Databases
– What are they?
– What problem vector databases solve?
– Top 5 vector database
– Key Differences
– Challenges and use cases
– How to build vector databases from scratch - Metrics and data structure
- Vectorization with abstraction frameworks
- Hands-on use cases: full AI-based application workflow (with LLMs)
- Vector database comparisons
- How to choose a vector database
For me, what was of particular note were the sections on similarity – a very light example of the vector math that allows you to see how close two objects might be to each other in the object store.
There were some great labs in there with the source code on Paulo’s github page. All in all, while only a short course (4-5) hours – it was time well spent.
If you’re curious how all this stuff works, enjoy the rabbit hole!