I’ve been thinking about this a lot because I remember spending weeks on vector store tuning last year, and I’m really curious what the visual builder approach actually eliminates versus what it just hides.
When you use Latenode’s visual builder to wire up a RAG pipeline, you’re connecting data sources to retrieval components to LLMs. But vector stores are still part of the architecture, right? So does the visual approach mean the platform handles vector store management for you, or does it mean you’re just not touching that layer directly but it’s still there doing its thing?
I’ve read that you can implement RAG workflows without managing vector stores yourself, and that the builder handles real-time data retrieval dynamically. But I’m genuinely unclear if that’s a technical limitation removal or just a UI convenience.
Also, does anyone know if the trade-off is that you lose customization options? Like, can you still optimize retrieval performance if you’re not managing the vector store yourself, or does that become someone else’s problem?