Skip to main content

Chat & RAG

Chat & RAG consists of the open source solutions OpenWebUI and Ollama.

OpenWebUI is a user-friendly web interface for interacting with various local large language models, offering a customizable and engaging user experience.
This includes chatting with models, analyzing documents, or generating images.

Ollama is a tool that enables the easy and fast execution of large language models (LLMs) such as Gemma3 on the local meloki appliance.

When you start Chat & RAG for the first time, we will already download some open source models for you so that you can get started right away.

Initial models​

These models are already available in Ollama after the first start and can be used directly in OpenWebUI.

Installed models

gemma3:27-it-qat
This is currently Google's best model.
Gemma 3 offers multimodality, supports image-text inputs and text outputs. It processes context windows of up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilities.

phi4:14b
This is Microsoft's current top model.
Phi-4 is particularly good at mathematical tasks.

llama3.2:1b
This is a lightweight text model that can also run on edge and mobile devices.

Additional models​

If you want to use additional models, you can use OpenWebUI Administration.
There you can download new models for Ollama.

To do this, log in to OpenWebUI as an administrator and open the Admin Panel.
Then navigate to Settings -> Models.

On the right-hand side, you will find a “Download Icon”.

Installed models
Here you can conveniently download additional models.

Further documentation​

Further information on OpenWebUI or Ollama can be found in the official links.

OpenWebUI documentation: https://docs.openwebui.com/
Ollama documentation: https://ollama.com/