ollama-openwebui

Ollama and Open WebUI

Ollama is a lightweight API server designed for hosting and interacting with machine learning models. Open WebUI is a user-friendly web interface for managing and utilizing these models. Together, they provide a seamless way to deploy, interact with, and manage ML models locally.

I could have used the Open WebUI image that is bundled ollama support, ghcr.io/open-webui/open-webui:ollama, but having a separate container for both allows for more flexibility.

Launching

Make sure NVIDIA Container Toolkit is installed and configured.

To start the Ollama and Open WebUI containers, run

podman compose up --detach

Accessing

Both Ollama and Open WebUI can be accessed independently. Ollama is accessed using its API at localhost:11434, while Open WebUI is accessed via a web interface at localhost:10000.

Check if Ollama is up and running

curl --request GET --location http://localhost:11434

Pull the model deepseek-r1

curl --request POST --location http://localhost:11434/api/pull -d '{
  "model": "deepseek-r1"
}'

Sending a prompt to the model

curl --silent --request POST --location http://localhost:11434/api/generate --data '{
  "model": "deepseek-r1",
  "stream": false,
  "prompt": "Why is the sky blue?"
}' | jq --compact-output

Teardown

To stop and remove the containers, run

podman compose down --volumes