Skip to main content
Back to Projects

Project case study

Mamba Chat — Streaming Chat UI

AI/ML

A React/TypeScript streaming chat SPA with rich content blocks and offline-first IndexedDB persistence; pairs with Mamba Server

Live DemoView Source
TypeScriptReactViteVercel AI SDKDexieRadix UIshadcn/uiTailwind CSSRechartsShiki

A modern, feature-rich AI chatbot SPA built with React and TypeScript. Mamba Chat delivers a polished conversational experience with real-time streaming, rich interactive content, and offline-first persistence, and connects to a Mamba Server backend.

The architecture emphasizes clean separation of concerns: a service-layer pattern abstracts the AI backend (with mock and real transports), a repository pattern via custom hooks handles data access, and context-based state management ties it together.

Key capabilities:

  • Character-by-character streaming of rich content blocks, including interactive forms, charts, code snippets, and cards rendered from AI tool calls
  • Offline-first storage of conversations and messages in IndexedDB via Dexie, with automatic conversation-title generation
  • Multi-model architecture, dark/light theming with flash prevention, and a mobile-first responsive layout