ollama is a golang LLM server built for ease of use.

To update or switch versions, run webi ollama@stable (or @v0.1.5, etc).

Cheat Sheet

ollama is an LLM serving platform written in golang. It makes LLMs built on Llama standards easy to run with an API.

To get started quickly with the open source LLM Mistral-7b as an example is two commands.

  1. Open TWO Terminals
  2. In the first, start the ollama server
    OLLAMA_ORIGINS='*' OLLAMA_HOST=localhost:11434 ollama serve
  3. In the second, run the ollama CLI (using the Mistral-7b model)
    ollama pull mistral
    ollama run mistral

Table of Contents

  • Files
  • ChatGPT-style Web UI
  • System Notes
  • Models to Try
  • As a Network API


These are the files / directories that are created and/or modified with this install:


How to Use a ChatGPT-style Web Interface

There are many Ollama UIs to choose from, but ollama-webui is easy to start with (and can be built as a static page):

  1. Install node
    webi node@lts
    source ~/.config/envman/PATH.env
  2. Clone and enter the ollama-webui repo
    git clone https://github.com/ollama-webui/ollama-webui.git ./ollama-webui/
    pushd ./ollama-webui/
  3. Install and start the project
    cp -RPp ./example.env ./.env
    npm clean-install
    npm run dev
  4. Enjoy!

Note: Be sure to run ollama with CORS enabled:

OLLAMA_ORIGINS='*' OLLAMA_HOST=localhost:11434 ollama serve

System Notes

You'll need a fairly modern computer. An Apple M1 Air works great.

  • 8GB+ RAM
  • 4GB+ Storage
  • Models range between 3GB and 30GB+
    (they can take a while to download, and several seconds to initialize)

How to Downloads Other Models

See the list at https://ollama.ai/library.

For example, we could try sqlcoder, or orca-mini (because it's small):

ollama pull sqlcoder
ollama run sqlcoder
ollama pull orca-mini
ollama run orca-mini

How to Use as an API on a Network

If you'd like ollama to be accessible beyond localhost (

  • set the host to, which makes it accessible to ALL networks
  • you may wish to limit origins
# fully open to all

# restrict browsers (not APIs) to requests from https://example.com
OLLAMA_ORIGINS='https://example.com' OLLAMA_HOST= ollama serve

See also:


Report an Issue Submit Installer Star on GitHub