Launch a CPU LLM Server on AWS with Ollama + Open WebUI

Last updated on 
Launch a CPU LLM Server on AWS with Ollama + Open WebUI

Running a private LLM stack on AWS usually means installing model runtime, setting up a UI, configuring services, opening ports, and fixing startup issues before you can even test a prompt. This AMI is designed to remove that setup burden. Prezelfy CPU LLM Server AMI provides a deployment-ready stack with Ollama and Open WebUI, tuned for CPU-based workloads and packaged for fast launch in AWS.

What Do We Offer with Our CPU LLM Server AMI?

This AMI includes more than a standard install. It is preconfigured for faster onboarding, operational consistency, and secure-by-default behavior.

Included in this AMI:

  • Ollama installed and configured as a system service
  • Open WebUI installed and configured as a system service
  • CPU-friendly model preload workflow
  • Amazon Linux 2023 based hardened baseline
  • Firewall configuration for required ports
  • Open WebUI access on port 8080
  • Ollama API access on port 11434
  • Service startup integration and runtime validation

Architectural Design

The architecture is intentionally simple and practical for teams that want a private LLM endpoint without GPU dependency.

  • ollama.service runs local model serving on 127.0.0.1:11434
  • open-webui.service provides browser UI on 0.0.0.0:8080
  • Open WebUI communicates with Ollama over local network
  • Model storage is set to a fixed path for consistency
  • First-boot fallback can preload missing models if needed

This design targets reliability and predictable behavior across instance restarts and new launches.

How to Use the CPU LLM Server

  1. Launch an EC2 instance from the AMI in a subnet with outbound internet access.
  2. Configure security group inbound rules:
    • TCP 22 for SSH (restricted to your IP)
    • TCP 8080 for Open WebUI
    • TCP 11434 for API access (optional, restrict as needed)
  3. Open Open WebUI in browser:
    • http://<public-ip>:8080
  4. Verify services on host:
    • sudo systemctl status ollama
    • sudo systemctl status open-webui
  5. Verify models:
    • ollama list

Example API call:

Access API via HTTP
curl http://127.0.0.1:11434/api/generate -d {"model":"phi3","prompt":"Hello"}'

Troubleshooting

Cannot access Open WebUI from public IP:

  • Confirm open-webui is running
  • Confirm host firewall allows 8080/tcp
  • Confirm security group allows 8080/tcp
  • Confirm subnet route/NACL allows traffic

No models visible in UI:

  • Check ollama list
  • Pull model manually if needed: ollama pull phi3
  • Refresh Open WebUI model view after model pull

Ollama service fails to start:

  • Validate binary path: command -v ollama
  • Check logs: sudo journalctl -u ollama -n 200 --no-pager

Open WebUI local curl fails:

  • Check listener: sudo ss -ltnp | grep 8080
  • Check logs: sudo journalctl -u open-webui -n 200 --no-pager

Ready to Launch?

If you need a practical, private, CPU-based LLM stack on AWS, this AMI gives you a faster path from instance launch to first prompt. It is built to reduce setup friction while keeping deployment flexible for your own network and access policies.