DDEV Ollama

Overview

This add-on integrates Ollama into your DDEV project.

Ollama allows developers to run LLMs locally.

Installation

ddev add-on get tyler36/ddev-ollama
ddev restart

After installation, make sure to commit the .ddev directory to version control.

Usage

ddev ollama is a helper command allowing developers to interact with the Ollama service. The table below describes some commonly-used commands. Type ddev ollama --help to see all available commands.

Command	Description
ddev ollama run <model>	Run a model. Models will automatically download if not available locally
ddev ollama stop	Stop a running model
ddev ollama list	List models
ddev ollama ps	List running models
ddev ollama --help	Show available commands

Other Commands

Command	Description
ddev describe	View service status and used ports for Ollama
ddev logs -s ollama	Check Ollama logs

Advanced Customization

To change the Docker image:

ddev dotenv set .ddev/.env.ollama --ollama-docker-image="ollama/ollama:latest"
ddev add-on get tyler36/ddev-ollama
ddev restart

Make sure to commit the .ddev/.env.ollama file to version control.

All customization options (use with caution):

Variable	Flag	Default
OLLAMA_DOCKER_IMAGE	--ollama-docker-image	ollama/ollama:latest

Tip

If you do not require GPU support, it is recommended to use alpine/ollama It is significantly smaller, however, it be a week or 2 out of sync with the latest ollama/ollama image.

API Endpoint

When running Ollama locally, it is typically accessed at http://localhost:11434. However when using DDEV and Docker, access the Ollama server via the container name (ollama:11434). For example:

$ ddev exec curl ollama:11434
Ollama is running

Many Ollama-compatible packages recommend setting this directly in an .env:

OLLAMA_BASE_URL=ollama:11434

NVIDIA GPU Support

Note

NVIDIA GPU support is only available for Linux and WSL2.

To allow the Ollama container to use your GPU,

1. Install NVIDIA Container Toolkit.

2. Create .ddev/docker-compose.ollama-gpu.yaml and add the following:

   # .ddev/docker-compose.ollama-gpu.yaml
   services:
     ollama:
       deploy:
         resources:
           reservations:
             devices:
               - driver: nvidia
                 count: all
                 capabilities: [gpu]

3. Restart DDEV to apply changes

   ddev restart

Credits

Contributed and maintained by @tyler36