Vector Search vs Agentic Search #ai #rag #search #explained #tech #semantic #vector #claudecode

Key Takeaways

Here's the difference between vector versus agentic search. AI tools like cloud code uses agentic search, but there is a lot of discussion online about which one is better. Vector search takes a bunch of text, whether it's a codebase or a bunch of documentation pages, chunks them up, and then turns those chunks into numbers. So when you ask a question like where does the database logic in this codebase live, your question also gets turned into numbers. And then this vector system will find out mathematically which vector from the large body of text is closest to your question. So vector search is meant to be really fast, but you can imagine that context gets lost when things get vectorzed. This is because when you chunk up text, then certain pieces don't know about the pieces around it. Aentic search is different. So aentic search will use ls to list files, gre to search things, and it reads through all the files. Essentially acts more like a human in the way that it searches through something. This approach is slower because it has to go through more volume to find the answer, but tends to be a little bit more accurate because agentic search can see the full