Installation

Prerequisites

1. Node.js 18+ — Download
2. Ollama running locally — Download
3. At least one model pulled:

ollama pull llama3.2

Recommended first model

llama3.2 is a great starting point — fast, capable, and small enough to run on most hardware. For reasoning features, also try deepseek-r1.

Install and Run

git clone https://github.com/LogneBudo/llmxray.git
cd llmxray
npm install
npm run dev

Open http://localhost:5173 in your browser. That's it.

How the Proxy Works

LLMxRay's dev server automatically proxies API calls to Ollama:

URL prefix	Proxied to
/api/*	http://localhost:11434/api/*
/v1/*	http://localhost:11434/v1/*

If Ollama is running on a different port or machine, change the URL in Settings > Connection.

Build for Production

npm run build    # Type-check + production build → dist/
npm run preview  # Preview the build locally

The dist/ folder contains a static site you can serve from any web server.

Verify Connection

After starting the app, look at the top-right corner of the header:

• Green indicator + "Connected" — Ollama is reachable
• Red indicator + "Disconnected" — Check that ollama serve is running

You can also go to Settings and click the connection test button.

Pulling More Models

# Chat models
ollama pull llama3.2
ollama pull mistral
ollama pull deepseek-r1       # Reasoning model with <think> blocks

# Embedding models (for Knowledge Base and Embeddings Lab)
ollama pull nomic-embed-text
ollama pull all-minilm

# Vision models (for image attachments in chat)
ollama pull llava

LLMxRay automatically detects model capabilities (thinking, vision, embedding, tool-use) and adapts the UI accordingly.