Open-WebUI to use ollama

Make sure that you have ollama and docker installed.

This will start up a transient container than goes away once you stop it. Useful for testing and not leaving a whole bunch of junk laying around your computer.

1
2
3

docker run -it --rm -p 3000:8080 \
         -v open-webui:/app/backend/data \
         ghcr.io/open-webui/open-webui:main

I like to put it into a script, so you can fire it up on the command line. When you kill that process, the container will remove itself.

If you want to have the latest version

1

docker pull ghcr.io/open-webui/open-webui:main

And restart.

First thing you need to do is to create an admin account. This is for local stuff.

Once this is done, you can see an interface similar to the gpt interfaces we all know and love.

One fun trick while evaulating models and prompts is to run completion over multiple models. Create a new chat, then select a model at the top. Use the plus sign to add additional models.

The models will be run in turn and you'll be able to compare all the results next to each other.

It's worth creating a OpenWebUI account to be able to build off of other's work.

Lets go through the process of adding a tool. Go to keyless weather, and press Get. For the url, put in http://localhost:3000 and import it.

Then go to workspace -> models -> create. I'm calling it weather, based on llama3.2:latest, and then select enable the tool you jsut selected. Save it.

Head over to a new chat, select the model, and then ask it what the weather is in a specific location. For example

whats the weater in nyc next wek

And you should see a response, with the tool called out at the bottom.

… To summarize, the week ahead in NYC will see mostly clear skies with some cloudiness, ranging from temperatures of 21.7 °F to 31.5 °F.

TOOL:weather/get_future_weather_week

There's a lot more to explore, but this is a nice little start.