Hixbi — Multi-Model AI Chat Platform
Overview
Hixbi is a unified chat platform that brings together multiple AI models—Gemini, ChatGPT, Claude, and DeepSeek—into a single, seamless interface. Users can switch between models, manage conversations, and leverage edge computing for lightning-fast responses.
Problem & Users
Developers and AI enthusiasts often need to compare responses from different LLMs or use specific models for different tasks. However, managing multiple subscriptions, juggling different interfaces, and dealing with varying APIs creates friction.
Key challenges:
- •Switching between multiple AI platforms is time-consuming
- •Each platform has different authentication and billing systems
- •Response times vary significantly based on geography
- •No unified conversation history across models
Solution Architecture
Tech Stack
Frontend
Next.js 14, TypeScript, Zustand
Edge Layer
Cloudflare Workers
Authentication
NextAuth.js with RBAC
Payments
Stripe subscriptions
System Design
User Request → Next.js App → Cloudflare Worker (Edge) → AI Provider
↓
Zustand State
↓
PostgreSQLKey Technical Decisions
1. Edge Compute for Low Latency
Instead of proxying all AI requests through a single backend server, I deployed separate Cloudflare Workers for each AI provider. This reduces latency by routing requests through Cloudflare global network.
2. Unified State Management
Managing conversation state, model switching, and streaming responses required a robust state solution. Zustand provided the perfect balance of simplicity and power.
3. Stripe Integration
Users can choose between different subscription tiers with Stripe webhooks handling lifecycle events and NextAuth RBAC enforcing feature access.
Impact & Metrics
30%
Faster TTFB
99.9%
Uptime
<100ms
To first token
45
Msgs/day avg
85%
Week 1 retention
4.8/5
User rating
What is Next
- →Organization workspaces: Team collaboration with shared conversations
- →Usage dashboards: Detailed analytics on model performance and costs
- →RAG integration: Upload documents for context-aware responses
- →Mobile apps: Native iOS and Android applications