Unified Conversational-AI Web Client for an Enterprise Chatbot Platform

Overview

The consolidated, customer-facing web client for an enterprise conversational-AI / RAG platform. It replaced earlier per-client chatbot forks with a single, themeable Next.js application that serves branded chat experiences to large enterprises behind their own identity providers.

The Challenge

The platform had grown a separate frontend codebase per enterprise customer, each diverging in auth, theming, and streaming behaviour. That made every new client a fresh fork to maintain. A unified, configurable client was needed to deliver consistent chat UX, observability, and SSO across accounts while keeping a single source of truth.

What We Built

A Next.js 14 App Router application (src/app, src/components, src/lib) built and bundled with Bun. It delivers token-by-token assistant responses over SSE (sse.js) with a Socket.IO real-time channel, rich Markdown rendering (react-markdown, remark-gfm, image/HTML sanitization via dompurify), and a component system built on Radix primitives and shadcn-style UI with Tailwind. Authentication runs through Firebase with next-firebase-auth-edge for edge-side session verification. State is handled with Jotai and TanStack Query/Table, forms with React Hook Form + Zod. The app ships with deep production observability wired in (Sentry, LogRocket, FullStory) and deploys to Kubernetes via Docker with separate prod/dev manifests.

Technologies & Approach

Next.js + React + Tailwind + Radix for a fast, accessible, themeable shell; Bun for install/build speed; SSE and Socket.IO for low-latency streaming; Firebase edge auth to slot into enterprise sign-in. The heavy observability stack reflects an enterprise-grade reliability bar rather than a build.

Outcome / Impact

Delivered a single, maintainable frontend that standardized streaming chat, auth, and monitoring across enterprise tenants, collapsing multiple client-specific forks into one product surface.

Capabilities Demonstrated

• Production streaming chat interfaces (SSE + WebSocket)
• Enterprise SSO and edge-side session handling
• Multi-tenant, themeable conversational-AI frontends
• RAG answer rendering with safe Markdown/HTML
• End-to-end product observability (Sentry, LogRocket, FullStory)