Ollama Chat Assistant

Next.js chat application using assistant-ui with local LLM integration via Ollama.

Prerequisites

• Ollama must be installed and running locally
• Install from: https://ollama.ai
• Default endpoint: http://localhost:11434

Getting Started

1. Start Ollama and pull at least one model:

ollama pull gemma3:latest
# or any other model

2. Install dependencies:

npm install

3. Run development server:

npm run dev

4. Open http://localhost:3000

Features

• 🔄 Dynamic model selection from available Ollama models
• 💬 Multi-thread conversation management
• 🎨 Markdown rendering with syntax highlighting
• 💾 Thread persistence with localStorage
• ⚡ Real-time model switching

Tech Stack

• Next.js 15.5 (Static Export / CSR/SPA)
• assistant-ui (React chat components)
• ollama/browser (Direct browser-to-Ollama API)
• Zustand (State management)
• Radix UI + Tailwind CSS

Build and Deploy

# Build for local testing (localhost:11434)
npm run build:local

# Build for production (hospital IP)
npm run build:prod

# Serve static build
npm run serve

GitHub Actions: Automated workflows build both local and production variants on push/release.

Deployment

The app exports to static HTML/CSS/JS in the /out directory. Deploy to any HTTP server (nginx, Apache, etc.) with access to an Ollama server. No Node.js runtime required.

See CLAUDE.md for detailed documentation.