Hugging Face's logo

Spaces:
mishrabp
/
ollama
Sleeping

App Files Community
ollama / README.md
mishrabp's picture
mishrabp
Upload folder using huggingface_hub
1c1c554 verified
preview code
|
raw
history blame contribute delete
3.18 kB
metadata 
title: Ollama
emoji: ⚡
colorFrom: pink
colorTo: blue
sdk: docker
pinned: false
license: mit
short_description: Ollama (llama3.2:3b) on Hugging Face Spaces (Docker)

🦙 Ollama (llama3.2:3b) on Hugging Face Spaces (Docker)

This Space runs Ollama with the llama3.2:3b model using a Docker-based Hugging Face Space.

The container starts Ollama, pulls the model at runtime, and exposes the Ollama HTTP API so you can interact with the model programmatically.

🚀 Hugging Face Space Configuration

When creating the Space:

  • SDK: Docker
  • Template: Blank
  • Hardware: Free CPU (sufficient for 3B models)
  • Visibility: Public or Private (your choice)

No additional configuration files are required beyond this repository.

🧱 How It Works

  • Ollama runs inside a Docker container
  • The model llama3.2:3b is pulled on startup
  • Ollama listens on port 11434
  • Hugging Face automatically maps the port and exposes the Space URL

📡 API Usage

Once the Space is running, you can interact with Ollama via HTTP.

Check version 

curl https://<your-space>.hf.space/api/version

Generate text 

curl https://<your-space>.hf.space/api/generate \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama3.2:3b",
    "prompt": "Explain Kubernetes like I am five"
  }'

🔄 Automatic Deployment (GitHub → Hugging Face)

You can deploy automatically using GitHub Actions.

Required Secrets (GitHub Repository)

Go to Settings → Secrets and variables → Actions, then add:

Secret Name Description
HF_USERNAME Your Hugging Face username
HF_TOKEN Hugging Face Write access token
SPACE_NAME Name of the Hugging Face Space

The workflow logs in to:

registry.hf.space

and pushes the Docker image, triggering a redeploy of the Space.

🧪 Local Development (Optional)

You can run the same container locally for testing.

Build the image 

docker build -t ollama-local:latest .

If you hit DNS/network issues:

docker build --network=host -t ollama-local:latest .

Run locally 

docker run -p 11434:11434 ollama-local:latest

Test:

curl http://localhost:11434/api/version

☸️ Local Kubernetes Deployment (Optional)

This project can also be deployed to a local Kubernetes cluster (Docker Desktop, Minikube, MicroK8s).

Build the image 

docker build -t ollama-local:latest .

Deploy 

kubectl apply -f local.yml

Access the service

  • Docker Desktop:
    http://localhost:30786
  • Minikube:
    echo "http://$(minikube ip):30786"

Verify:

curl http://localhost:30786/api/version

⚠️ Notes & Limitations

  • Free CPU Spaces are slow for inference (expected for LLMs)
  • Model downloads happen at container startup
  • Hugging Face Spaces may restart containers periodically

📚 References