Loading…
19
The AI newsletter guy said vector databases were overkill for my project, he was dead wrong
I run a small repair shop in Cleveland and wanted a chatbot to answer customer questions about our 200+ service codes. This newsletter author kept saying start with keyword search and add vectors later if needed. So I did that with a simple Python setup. Six weeks in, every other question was some variation of 'my washing machine makes a thumping sound and leaks', keyword search kept returning results for the wrong machine type or missing the combo of symptoms. I finally spent two days adding embeddings with a free model and suddenly the answers made sense. That advice cost me way more time than just doing it right the first time. Has anyone else found that generic AI advice fails hard when your data has specific jargon or overlapping terms?
0 comments
Log in to join the discussion
Log In0 Comments
No comments yet
Be the first to share your thoughts on this discussion.