DeployFest 2026

AI Agent built with Vertex AI (Gemini 1.5 Pro) + Cloud Run for the DeployFest hackathon.

Stack

Layer	Tech
Frontend	Next.js 14, TypeScript, Tailwind CSS
Backend	Express, TypeScript,@google-cloud/vertexai
AI	Vertex AI — Gemini 1.5 Pro with function calling
Deploy	Google Cloud Run

Local Development

Prerequisites: Node 22+, a GCP project with Vertex AI API enabled.

# Backend
cd backend
cp .env.example .env        # fill in GCP_PROJECT_ID
npm install
npm run dev                 # http://localhost:3001

# Frontend (separate terminal)
cd frontend
cp .env.example .env.local  # BACKEND_URL=http://localhost:3001
npm install
npm run dev                 # http://localhost:3000

Deploy to Cloud Run

# Authenticate
gcloud auth login
gcloud auth configure-docker

# One-shot deploy
chmod +x infra/deploy.sh
./infra/deploy.sh <YOUR_GCP_PROJECT_ID>

Adapting to the Problem Statement

On hackathon day (9:45 AM reveal):

1. backend/src/agent/prompts.ts — Update SYSTEM_PROMPT to reflect the problem domain.
2. backend/src/agent/tools.ts — Replace stub implementations in toolHandlers with real logic.
3. frontend/app/page.tsx — Update the title and description.
4. Set NEXT_PUBLIC_APP_TITLE in frontend/.env.local.

Project Structure

├── backend/
│   └── src/
│       ├── agent/
│       │   ├── agent.ts      # Agentic loop (Vertex AI function calling)
│       │   ├── tools.ts      # Tool definitions + handlers (swap on day)
│       │   └── prompts.ts    # System prompt (swap on day)
│       └── routes/chat.ts    # POST /api/chat
├── frontend/
│   ├── app/
│   │   ├── page.tsx          # Main page
│   │   └── api/chat/route.ts # Next.js proxy to backend
│   └── components/
│       ├── ChatInterface.tsx  # Chat UI with loading state
│       └── MessageBubble.tsx  # Message + tool call visualizer
└── infra/
    ├── cloudbuild.yaml        # Cloud Build CI/CD
    └── deploy.sh              # Manual one-shot deploy