01 / Overview
Fashion AI Workbench is a Turborepo that turns a broad AI product idea into an executable engineering foundation. It connects a Next.js interface to a shared type-safe API, relational storage, caching and multiple model providers.
Fashion AI Workbench
A Bun-first workspace for building type-safe AI products with provider switching, streaming responses and production-shaped infrastructure.
- Next.js 16
- Bun
- tRPC
- Drizzle
- Vercel AI SDK
02 / The question
What needed to be understood?
AI prototypes often stop at a chat screen. This project explores how to preserve fast experimentation while keeping data contracts, infrastructure and provider boundaries explicit enough for a real product.
03 / The build
How the project approached it
- Centralized the tRPC router so server and client consumers share one contract.
- Isolated model selection behind a provider layer for Ollama, OpenAI and DeepSeek.
- Kept PostgreSQL, Redis and local Ollama available through one development stack.
- Added workspace-wide lint, typecheck and test tasks through Turborepo and CI.
04 / Current state
What remains
The repository is a reusable, runnable base for AI workflows rather than a disconnected collection of demos. The next product milestone is validating a focused fashion workflow on top of it.
Links