ChatGPT's interface, Ollama's backend, your hardware

What it does

This Next.js application wraps Ollama’s local API in a chat interface that looks and behaves like ChatGPT. Conversations live in browser localStorage, so there is no database to manage, and you can pull, delete, and switch between models directly from the sidebar. It is designed to run fully offline once Ollama is available on your network.

The interesting bit

It is largely a polished frontend layer—Next.js, Tailwind, and shadcn-ui wired to Ollama’s endpoints—but that simplicity is the point. The author also maintains a separate shadcn-chat component library, which suggests this project doubles as a proof-of-concept for reusable chat primitives. The README’s honest admission that this is a hobby project, with a gentle nudge toward Open WebUI for power users, is a rarity in the LLM tooling gold rush.

Key highlights

• Runs fully offline with local Ollama instances; no cloud dependencies
• Chat history persists in localStorage without database setup
• Built-in model management: download, delete, and switch models from the UI
• Supports vision models via image prompts and includes voice input
• Responsive layout with light/dark mode and code syntax highlighting

Caveats

• Chat import and export is still on the upcoming-features list
• The author explicitly recommends Open WebUI for anyone needing a more complete experience
• Non-localhost deployments require manual CORS configuration on the Ollama server side

Verdict

Ideal for developers who want a clean, single-user chat frontend over Ollama without infrastructure overhead. Look elsewhere if you need multi-user support, server-side persistence, or enterprise-grade features.