Skip to content

DEVLOG.md

Latest commit

 

History

History
86 lines (56 loc) · 2.46 KB

DEVLOG.md

File metadata and controls

86 lines (56 loc) · 2.46 KB

nvvri devlog

A running log of what I built, what I learned, and what broke along the way. No polish. Just honest notes from the process.

How to read this

Each entry has:

  • What I did -- the actual change
  • Why -- the reason behind the decision
  • What I learned -- the thing I will carry forward
  • What broke / what was hard -- honest account of where I struggled

Entries

[2026-05-19] -- Project reset: documenting everything from here

What I did

Set up this devlog. Going back to basics -- not just shipping features, but understanding every decision I make.

Why

Two reasons. First, I want to actually learn, not just copy and paste until something works. Second, a well-documented project tells a story. Anyone reading the code should be able to understand not just what I built but why I built it that way.

What I learned

Explaining something forces you to understand it. If I cannot write a clear sentence about why I made a decision, I probably do not understand it well enough yet.

Current state of the project

Stack:

  • Next.js 15, App Router, Turbopack
  • React 19
  • TypeScript strict mode, zero any
  • Neon (PostgreSQL) + Prisma v5
  • OpenRouter for AI search
  • Resend for transactional email
  • Leaflet + OpenStreetMap for the map view
  • Playwright for E2E tests, runs in CI
  • Deployed on Vercel at nvvri.co.uk

What is already built:

  • AI natural language search with multi-model fallback chain (Llama 3.3 -> Mistral 7B -> Gemma 3 -> openrouter/auto -> local parser)
  • Local parser fallback -- search works even if every AI model is down
  • PostgreSQL database with Prisma, 6 nurseries seeded with real coordinates
  • Real enquiry backend via Resend, sends two emails (parent confirmation + admin notification)
  • Mobile responsive UI with bottom-sheet modal
  • Map view with Leaflet, custom markers, enquire CTA in popup
  • Playwright E2E tests covering list and map enquiry flows
  • SEO foundation: SSR, sitemap.xml, robots.txt, JSON-LD Preschool schema
  • AI answer card -- zero-click summary at the top of results
  • Per-nursery detail pages with server rendering and static generation

What is next:

  • Parent auth (shortlists and saved nurseries)
  • Compare nurseries side by side
  • Search analytics (log queries, track model success rates)
  • Nursery admin portal