Skip to content
MH
All projects
Full StackGenerative AI

ChatCraftAI

Full Stack / Real-Time / AI Product

A real-time messaging platform with AI woven into the conversation rather than bolted onto the side — built as a full-stack TypeScript application covering the realtime transport, the data layer and the interface.

Year
2025
Role
Full-stack developer
Source
Public repository
ChatCraftAI — concept diagram

Technologies

  • TypeScript
  • React
  • Node.js
  • WebSockets
  • LLMs
  • Real-Time

The problem

Most AI chat interfaces are single-player: one person talking to a model. Real conversations have several people in them, and adding AI to a multi-party realtime channel raises questions a single-user chat never has to answer.

The solution

A messaging platform built realtime-first, with AI as a participant in the message stream. State is shared and consistent across connected clients, and AI assistance appears inside the conversation rather than in a separate panel.

Architecture

How it is put together.

01

Realtime transport

A persistent connection layer keeps every connected client in sync as messages arrive.

02

Message pipeline

Messages are persisted and broadcast, with ordering preserved across clients.

03

AI augmentation

AI responses are produced against conversation context and streamed into the same channel.

04

Interface

A React front end renders live conversation state with streaming updates.

Key features

What it actually does.

  • Real-time multi-client messaging
  • AI participation inside the conversation
  • Streaming responses rather than blocking waits
  • End-to-end TypeScript across client and server

Engineering challenges

The parts that were hard.

Consistency under concurrency

Multiple clients and an async AI producer all writing to one stream makes ordering and de-duplication a real problem rather than a detail.

Latency is the product

In chat, perceived speed is the experience. Streaming was a design requirement, not an optimisation.

Outcomes

What came out of it.

Qualitative outcomes only — no invented benchmarks or metrics.

01

A working realtime AI messaging application

02

Streaming AI responses in a multi-party channel

03

Full-stack ownership from transport to interface

TravelMate — Generative AI / RAG / Final Year Project

Next project

TravelMate

An AI travel assistant built as a final year project.

Let’s build

Want something like this for your business?

Tell me the problem and I will come back with an architecture, a scope and a timeline.