VMVirginia Mwegainvirginiamwegahashnodedev.hashnode.dev·6h ago · 9 min readWriting API docs an AI agent can actually consumeI found the gap the embarrassing way: one of my own agents couldn't call one of my own APIs. FamNest runs a small agent graph — a router hands a parent's message to a retriever, the retriever grounds 00
VMVirginia Mwegainvirginiamwegahashnodedev.hashnode.dev·1d ago · 5 min readMulti-tenant vector search with pgvector + Row Level SecurityMy app is an AI wellness coach for parents. Every user's data is about the most private thing they have: how they're actually coping. Their check-ins, their bad nights, the things they'd never say out00
VMVirginia Mwegainvirginiamwegahashnodedev.hashnode.dev·3d ago · 5 min readThe deterministic crisis floor: when you can't trust the model with the highest-stakes callMost of building an AI product is making things better — warmer, smarter, more personal. This post is about the one place where I deliberately made my product dumber, on purpose, and why it's the deci00
VMVirginia Mwegainvirginiamwegahashnodedev.hashnode.dev·6d ago · 7 min readThree-tier memory for a parenting agent — without leaking one family's night into another'sFor the coach to feel like it knows a parent, it has to remember things. That last week was brutal. That they work nights, so "go to the gym" is a joke and a three-minute reset is real. That last time00
VMVirginia Mwegainvirginiamwegahashnodedev.hashnode.dev·Jun 22 · 8 min readYour AI agent needs a second agent whose only job is to say noMost agent tutorials show you how to make a model do something. Call a tool, plan a task, chain a few steps together. That part is mostly solved now, and the content reflects it — there are a thousand00