Hackathon II β€” PanaversitySpec-Driven Development

Evolution of Todo

$
5
Phases
1000
Total Pts
5
MCP Tools
3
AI Layers
// HACKATHON PHASES

Project Progression

250 / 1000 pts earned
PHASE 1
Console App
In-memory CLI todo app
Python+100pts
PHASE 2
Full-Stack Web
REST API + Neon DB + Auth
Next.js + FastAPI+150pts
PHASE 3
AI Chatbot
Natural language todo mgmt
Agents SDK + MCP+200pts
// MCP SERVER

AI Tool Layer

5 tools exposed to OpenAI Agents SDK for natural language task management

add_task()

Create new todos via natural language

list_tasks()

Fetch & filter task lists

complete_task()

Mark tasks as done

update_task()

Modify existing tasks

delete_task()

Remove tasks permanently

// SYSTEM DESIGN

Architecture

πŸ’¬
Chat UI
Next.js
β†’
⚑
FastAPI
Backend
β†’
🧠
AI Agent
OpenAI SDK
β†’
πŸ”Œ
MCP Server
5 Tools
β†’
πŸ—„οΈ
Neon DB
PostgreSQL
// BUILT WITH

Tech Stack

⚑Next.js
πŸš€FastAPI
🧠OpenAI Agents SDK
πŸ”ŒOfficial MCP SDK
πŸ—„οΈNeon PostgreSQL
πŸ—οΈSQLModel
πŸ”Better Auth
🐳Docker
☸️Kubernetes
πŸ€–Claude Code
πŸ“‹Spec-Kit Plus
πŸ“¨Kafka
// WORKFLOW EXPLAINED

How It Works?

Step 01

Natural Language Input

User types a command like 'Add a task to buy groceries tomorrow' using everyday language.

Step 02

Agentic Reasoning

The AI agent parses the text, determining the required action and identifying key parameters via OpenAI models.

Step 03

Secure Context

The backend FastMCP server receives the action context, automatically tagging it with your unique JWT User ID.

Step 04

Database Execution

Neon PostgreSQL runs the specific CRUD command strictly under your isolated schema. Everything stays private.