Switch

llmnextjstypescriptsupabasepgvectorstreaming

A full-stack multi-model AI platform serving 1,000+ users and 13K+ daily database requests. Switch between models mid-conversation without losing context.

What it is

A chat platform that treats OpenAI and Anthropic as interchangeable behind a single normalized message schema, so a thread can hop between models with full continuity. A custom prompt-routing classifier splits traffic 70/30 between economy and premium models, and Cortex — a persistent memory system on Supabase Postgres with pgvector embeddings — gives every conversation low-latency semantic recall across 13K+ daily database requests.

Why it matters

Every provider has a different API shape, token budget, and streaming protocol, so people end up locked into one model and one context window. Switch removes that fragmentation: pick the best model for each message and keep the conversation intact.

Inspiration

Constantly copy-pasting the same conversation between ChatGPT and Claude to compare answers, and losing the thread every time.

Stack

Next.js 14TypeScriptSupabasepgvectorOpenAI SDKAnthropic SDKTailwind CSS

View on GitHub →Live site →

← All projects