SaaS · 2025
AI Chat Dashboard
A real-time AI-powered chat dashboard featuring multi-model conversation management, usage analytics, team workspaces, and an extensible plugin architecture — built for a SaaS product targeting developer teams.
- Role
- Full stack developer
- Year
- 2025
- Category
- SaaS
- Stack
- Next.jsTypeScriptTailwind CSSOpenAI APIWebSocketPostgreSQL
Key engineering metrics
Streaming
< 200ms
Token-by-token streaming via WebSocket with sub-200ms first-token latency
Models
Multi-provider
OpenAI, Anthropic, and Google models through a unified API layer
Analytics
Real-time
Token usage, cost tracking, and team activity dashboards
The problem
Developer teams using AI assistants had conversations scattered across multiple tools — ChatGPT, Claude, Gemini — with no centralised history, no usage tracking, and no way to share useful prompts or conversations across the team.
The solution
I built a unified dashboard that connects to multiple AI providers through a single interface. Conversations stream in real-time via WebSockets, the sidebar tracks history with full-text search, and an analytics panel shows token usage, cost breakdowns, and team activity. The plugin architecture allows custom model integrations without touching the core.
The outcome
A production SaaS platform handling concurrent AI conversations with sub-200ms streaming latency. Team workspaces reduced duplicate prompt engineering across the organisation, and the analytics layer gave engineering leads visibility into AI spend for the first time.
Architecture
- Frontend
- Next.js
- TypeScript
- Tailwind CSS
- Data
- PostgreSQL
- Tooling
- OpenAI API
- WebSocket
Next project
Portfolio Website
Portfolio · 2026
