Installation

Prerequisites

1. Ollama running locally — Download
2. At least one model pulled:

ollama pull llama3.2

Recommended first model

llama3.2 is a good starting point: fast, capable, small enough for most hardware. For reasoning features, also try deepseek-r1. For the Tool calls page, a model that supports tools, such as qwen2.5 or gemma4.

One command

With Node.js 22 or newer:

npx llmxray

Open http://localhost:5174. The app, and the observe port on 11435 for the Live page, start together.

Docker

docker run -p 5174:5174 -p 11435:11435 djovaneli/llmxray

On Linux the container cannot reach Ollama through host.docker.internal; pass the daemon's address:

docker run -p 5174:5174 -p 11435:11435 -e OLLAMA_URL=http://172.17.0.1:11434 djovaneli/llmxray

Leave the 11435 mapping out if you do not need the Live page.

Options

Flag	Default	What it does
--port <n>	5174	The app's port (env PORT)
--ollama-url <url>	http://localhost:11434	Where Ollama is (env OLLAMA_URL)
--observe-port <n>	11435	The Live page's observe port (env OBSERVE_PORT)
--no-observe		Do not open the observe port (env OBSERVE=0)
--host <address>	127.0.0.1	Interface to listen on (env HOST); 0.0.0.0 opens both ports to other machines on your network

Both ports answer only this machine by default, because the Live log holds whatever your tools sent.

Verify the connection

The top-right corner of the header shows a green dot and "Ollama connected" when the daemon is reachable, red when it is not. If it is red, check that ollama serve is running and that --ollama-url points at it. The URL can also be changed under Settings > General.

From source

For development, or to change the code:

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

Open http://localhost:5173. The dev server proxies /api/* and /v1/* to Ollama and opens the same observe port.

npm run build    # type-check and production build into dist/
npm start        # the shipped server, as npx llmxray runs it

Pulling more models

ollama pull llama3.2
ollama pull mistral
ollama pull deepseek-r1       # reasoning model with <think> blocks
ollama pull qwen2.5           # tool calling
ollama pull nomic-embed-text  # embeddings, for Knowledge Base and Embeddings
ollama pull llava             # vision, for image attachments in chat

LLMxRay reads each model's capabilities (thinking, vision, embedding, tools) from Ollama and adapts the pages accordingly.