Local Assistant is powered by Ollama, which lets you run AI models directly on your own machine, no cloud, no subscriptions, no data leaving your computer. To get started, you need to pull a few models. This guide covers three recommended models that together give you chat, vision, and basic coding capabilities.


Prerequisites

Before pulling models, make sure you have both Ollama and Local Assistant installed and running. If you haven't done that yet, follow the installation guides:



Vision (Image Analysis & Text Extraction)

This model can describe what's in a photo and extract text from images. It's based on Qwen 2.5 VL 7B and is ideal for Local Assistant's vision features.


Model: redule26/huihui_ai_qwen2.5-vl-7b-abliterated:latest

Size: 6.0 GB | Context: 125K tokens | Input: Text & Images


Pull it with this command:

ollama pull redule26/huihui_ai_qwen2.5-vl-7b-abliterated:latest


Read more about this model on its Ollama page

https://ollama.com/redule26/huihui_ai_qwen2.5-vl-7b-abliterated


Chat (Conversations & General Use)

For day-to-day conversations, questions, and general-purpose tasks, LLaVA 13B is an excellent choice. It delivers strong reasoning in a model that runs well on most modern hardware.


ollama pull llava:13b


Code (Basic Coding Assistance)

When you need help writing, reviewing, or explaining code, Llama 3.1 is a solid starting point. It handles most common programming languages and tasks.


ollama pull llama3.1:latest


Once the downloads are complete, simply refresh your Local Assistant interface in the browser. Local Assistant automatically detects all models available in Ollama, no manual configuration needed. You'll see your new models in the model selector dropdown, ready to use.


That's it! Three models, three capabilities. If you run into any issues pulling or running these models, feel free to create a topic and the community will help you out.