How I built a zero-cost RAG chatbot for my website
Precomputed embeddings, cosine similarity in a serverless function and an LLM with guardrails: the complete RAG pipeline behind "Ask Matteo", with no vector DB and no spend.
Blog.
Technical notes from the field: how the things I build actually work, decisions and lessons learned.
Tech Radar: a GenAI pipeline that works for me twice a week
GitHub Actions, zero npm dependencies and an LLM with constrained output: how the /radar page updates itself — but nothing gets published without my approval.