how this site works

builtbysuko.com is itself a project — a demonstration of the same AI and engineering thinking behind the other four. here's how it's built.

architecture

// request flow

visitor → GoDaddy DNS → Cloud Run (Next.js)

// AI features

chat message → embed (text-embedding-004)

→ KNN search (Firestore)

→ AI model + Google Search grounding

→ streamed response (SSE)

// blog generation (admin-triggered)

topic + details → KB retrieval → voice profile

→ AI model → draft → publish

// CI/CD

git push → Cloud Build → Docker build → Artifact Registry → Cloud Run

the knowledge base

the "ask suko" chat and the AI blog are both powered by a personal knowledge base — not a system prompt, an actual RAG system. it's built from three sources:

100+ question interview

career history, technical opinions, personality, working style — answered in free text, chunked and embedded into Firestore

personal writings

emails, notes, linkedin posts, cover letters, readme files — anything that captures how suko actually writes and thinks

voice profile

an AI-extracted style guide from all uploaded writing samples: sentence rhythm, vocabulary, humor patterns, what suko never sounds like

coding tools

ai-assisted by default — Cursor is the primary coding environment, Claude Code for agentic terminal workflows. Google AI Studio and Replit for quick ideation and prototyping.

Cursorprimary IDE — AI-assisted coding, daily driver

Claude Codeagentic coding via terminal — complex multi-file tasks

Google AI Studioquick ideation, prompt testing, model exploration

Replitrapid prototyping and throwaway experiments

model toolkit

LLM-agnostic — Claude agents are the default for most AI work, Gemini where GCP-native integration makes sense. rows marked site power builtbysuko.com directly.

Anthropic

Claude Sonnet 4.6

primary AI agent work — reasoning, planning, code generation

Anthropic

Claude Opus 4.6

deep research and complex multi-step reasoning tasks

OpenAI

codex-mini-latest

agentic coding via Codex CLI

Google

Gemini 2.5 Flash

ask suko chat — RAG retrieval + Google Search grounding

site

Google

Gemini 2.5 Pro

blog post generation — nuanced style matching

site

Google

Gemini 3.1 Flash-Lite

KB document processing — batch ingestion at low cost

site

Google

text-embedding-004

vector embeddings — 768-dim, native Firestore KNN search

site

GCP services

Cloud Runhosts the Next.js app as a container — scales to zero when idle

Cloud BuildCI/CD — triggers on push to main, builds Docker image, deploys

Artifact Registrystores Docker images between builds

Firestoreknowledge base (vector search), blog posts, contact submissions

Cloud Storagedemo videos, project images, uploaded documents

Secret ManagerAI API keys, Google OAuth credentials

Vertex AItext-embedding-004 for all vector embedding operations

frontend

Next.js 16

App Router, SSR + ISR

React 19

UI layer

Tailwind CSS v4

CSS-first config, no tailwind.config.ts

Framer Motion

entrance & hover animations

IBM Plex Mono

body font

Bebas Neue

display font