Project 04 / AI SaaS

ConvoBase

Problem: Businesses struggle to provide 24/7 intelligent customer support. Traditional chatbots are rigid and frustrating for users.

Solution: Built an AI-powered customer support platform with embeddable widgets, real-time messaging, and intelligent conversation routing using GPT-4.

Features

Tech Stack

Next.js 15

Express.js

Preact (Widget)

Turborepo

WebSockets

Architecture

Monorepo

Real-time

Role

Full-stack dev

(0→1)

Frontend

Next.js 15

Server-side rendering with App Router for optimal performance.

Backend

Express.js

Robust API layer with middleware architecture.

Widget

Preact

Lightweight embeddable widget (15KB gzipped).

Architecture

Turborepo

Monorepo with shared packages and remote caching.

Features / 01

AI Chat Widget

Embeddable chat widget powered by GPT-4 for intelligent customer support.

Monorepo Architecture

Turborepo-based architecture with shared packages for widget, admin, and API.

Real-time Messaging

WebSocket-based real-time messaging between customers and support agents.

Knowledge Base

AI-powered knowledge base that learns from past conversations.

Team Collaboration

Multi-agent support with conversation handoff and internal notes.

Analytics Dashboard

Comprehensive analytics on response times, resolution rates, and customer satisfaction.

Monorepo Architecture / 02

apps/web

Next.js Admin Dashboard

apps/widget

Preact Embed Widget

apps/api

Express.js API Server

packages/ui

Shared UI Components

ConvoBase uses a monorepo architecture powered by Turborepo. This allows code sharing between the admin dashboard, embeddable widget, and API while maintaining fast build times with remote caching.

Technology Stack / 03

Frontend

Next.js 15
React
TypeScript
Tailwind CSS
shadcn/ui

Backend

Express.js
Node.js
WebSockets
Redis
PostgreSQL

AI/ML

OpenAI GPT-4
Embeddings
Vector Search
RAG Pipeline

DevOps

Turborepo
Docker
GitHub Actions
Vercel
Railway