NOTE: This Github Action is deprecated. It was born out of a lack of alternatives, but since pydantic now maintains a Ollama Github Action, this one is not needed anymore. Check out The Pydantic alternative instead.

Ollama GitHub Action

A GitHub Action to easily install and run Ollama models in your workflow. Supports Linux, macOS, and Windows runners.

Features

• 🚀 Cross-platform support (Linux, macOS, Windows)
• 🔄 Automatic installation and setup
• 🎯 Run specific models
• ⚡ Fast model pulling and execution

Compatibility Matrix

Runner OS	Architecture	Status
Ubuntu 20.04+	x86_64	✅ Fully Supported
macOS 11+	x86_64, ARM64	✅ Fully Supported
Windows Server 2019+	x86_64	✅ Fully Supported

Usage

Example

- name: Serve Ollama Model
  uses: phil65/ollama-github-action@v1
  with:
    model: "smollm2:135m"

Inputs

Input	Description	Required	Default
model	Ollama model to use (e.g., llama2, codellama, mistral)	Yes	smollm2:135m

Outputs

Output	Description
server-url	URL of the running Ollama server (http://localhost:11434)
status	Status of the Ollama server (running/failed)

Platform-Specific Notes

Linux

• Runs natively using the official installer

macOS

• Installs via Homebrew
• Supports both Intel and Apple Silicon

Windows

• Uses the latest release from GitHub
• Custom installation path at C:\ollama

Example for a workflow

jobs:
  serve-model:
    runs-on: ubuntu-latest
    steps:
      - name: Start Ollama Server
        id: ollama  # Required to reference outputs
        uses: phil65/ollama-github-action@v1
        with:
          model: "smollm2:135m"

      # Example: Use the Ollama server in subsequent steps
      - name: Use Ollama
        run: |
          echo "Server URL: ${{ steps.ollama.outputs.server-url }}"
          echo "Server Status: ${{ steps.ollama.outputs.status }}"

          # Example API call
          curl "${{ steps.ollama.outputs.server-url }}/api/generate" \
            -d '{
              "model": "smollm2:135m",
              "prompt": "What is GitHub Actions?"
            }'

Server Lifecycle

The Ollama server will:

1. Start automatically when the action runs
2. Remain running for subsequent workflow steps
3. Be automatically cleaned up when the job completes

Note: If you need to stop the server manually in your workflow, you can use:

- name: Stop Ollama Server
  if: always()  # Ensures cleanup even if previous steps fail
  run: |
    pkill ollama || true

Environment Variables

The following environment variables are available during the workflow:

• OLLAMA_HOST: localhost
• OLLAMA_PORT: 11434

API Examples

Generate Text

- name: Generate Text
  run: |
    curl "${{ steps.ollama.outputs.server-url }}/api/generate" \
      -d '{
        "model": "smollm2:135m",
        "prompt": "Write a hello world program",
        "stream": false
      }'

List Models

- name: List Models
  run: |
    curl "${{ steps.ollama.outputs.server-url }}/api/tags"

Troubleshooting

Common Issues

Memory Issues

• Use a runner with more RAM
• Try a smaller model
• Close unnecessary processes

Debug Logs

Enable debug logging by setting:

env:
  ACTIONS_STEP_DEBUG: true

Security Considerations

• The server is accessible only on localhost
• Model files are stored in the runner's temporary space
• Cleanup is automatic after workflow completion
• No sensitive data is persisted between runs

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

• 🐛 Issue Tracker