Ollama with Cloudflare Tunnel

Run Ollama AI models locally with remote access via Cloudflare tunnels.

Prerequisites

• Docker Compose

Services

• ollama: AI model server with GPU support (port 11434)
• ollama-tunnel: Secure Cloudflare tunnel for remote access

Setup

1. Edit cloudflared/config.yml with your tunnel UUID and hostname
2. Run docker compose up -d
3. Access locally at http://localhost:11434 or remotely via Cloudflare
4. Check tunnel status: docker compose logs ollama-tunnel

Installing Models

# Start containers then run:
bash ollama/models.sh

Available Models

• LLMs: llama3.2:1b, gemma3:1b, deepseek-r1:1.5b
• Embeddings: nomic-embed-text, mxbai-embed-large
• Vision: granite3.2-vision:2b

Customization

• Edit ollama/models.sh to add/remove models
• Uncomment models in the script to enable them
• View installed models by uncommenting ollama list in the script

License

This project is licensed under the MIT License.