AI-Powered Personal Website

This project was created by Marcos Silva to learn about AI, Ollama, and web development.

An AI-powered personal website that answers questions about your professional experience and background. Built with Python, Flask, and Ollama AI models.

---

Features

• Interactive AI assistant powered by Ollama (Gemma model)
• Custom dropdown with predefined questions
• Real-time streaming responses (SSE)
• Responsive, modern UI
• Dockerized for easy development and production deployment
• Automated Hetzner VPS provisioning with secure firewall rules and cloud-init

Customization

To personalize this website for yourself:

1. Personal Information: Edit data/personal_info.txt with your professional background and experience.

2. Environment Variables: Set the following in your deployment:

   • PERSON_NAME: Your full name
   • PERSON_ROLE: Your professional role (e.g., "DevOps Engineer")
   • PERSON_LOCATION: Your location (e.g., "Argentina")

3. Website Content: Update templates/index.html to change:

   • Page title and meta tags
   • Sample questions in the dropdown
   • Social media links and resume
   • Open Graph image (replace static/og-image.png)

4. AI Model: Customize the Modelfile and retrain the model with your data if needed.

---

Quick Start

See CONTRIBUTING.md for development setup and local run instructions.

---

Setting up Hetzner VPS

The infrastructure is managed with Terraform (via Tofu) and includes:

• Automatic SSH key provisioning
• Secure default firewall (deny_all) attached on first boot
• Additional firewall rules (allow_rules) for HTTP, HTTPS, SSH, and ICMP
• Cloud-init configuration for user setup, security hardening, and Docker installation

export TF_VAR_hcloud_token=<YOUR_HETZNER_API_TOKEN>
export TF_VAR_ssh_publickey=<YOUR_SSH_PUBLIC_KEY>
cd infra
tofu init
tofu plan
tofu apply

Firewall Setup

• On first boot, the server is attached to a restrictive deny_all firewall.
• After provisioning, the allow_rules firewall is also attached, enabling HTTP, HTTPS, SSH (22 & 2222), and ICMP (ping).
• See infra/main.tf for details.

Cloud-init

• The server is configured via cloud-init.yaml.tftpl to:
  • Create a secure user
  • Harden SSH
  • Install Docker, fail2ban, and UFW
  • Enable UFW and fail2ban
  • Change SSH port to 2222

---

Deploying to Hetzner VPS with Kamal

Production deployments are managed using Kamal. You need to have Kamal installed and configured for your Hetzner VPS.

Ensure you have exported your KAMAL_REGISTRY_PASSWORD:

export KAMAL_REGISTRY_PASSWORD=<YOUR_REGISTRY_PASSWORD>

NOTE: Ensure you have configured your SSH key and Kamal.

Deploy Ollama service:

kamal deploy -c config/deploy.ollama.yml

Deploy Website service:

kamal deploy -c config/deploy.website.yml

See the config/ directory for deployment configuration files.

---

Configuration Files

• Dockerfile: Single file supports both dev and prod, switches mode based on FLASK_ENV
• compose.dev.yaml: Development setup with hot reload and file watching
• pyproject.toml: Python dependencies and project metadata
• config/deploy.ollama.yml: Kamal deployment config for Ollama service
• config/deploy.website.yml: Kamal deployment config for website service
• infra/main.tf: Terraform configuration for Hetzner VPS, firewall, and cloud-init
• infra/cloud-init.yaml.tftpl: Cloud-init template for secure server setup

---

Technologies Used

• Flask - Web framework
• Ollama - AI model server (Gemma 1B)
• Docker & Docker Compose
• Gunicorn - Production WSGI server
• uv - Fast Python package manager
• Kamal - Deployment tool for any app
• Hetzner Cloud - VPS hosting
• Terraform (via OpenTofu) - Infrastructure as Code

---

Contribution & Conduct

Please see the following documents for details:

• Code of Conduct
• Contributing Guidelines
• License (MIT)

---

This project was built to learn about Ollama, AI models, and Flask. Most of the HTML and JavaScript was written by AI.

To contribute, fork the repo and submit a pull request!

---

License

This project is released under the MIT License.

---

Links

• Star or Fork the Repo
• Ollama
• Ollama AI Models
• Flask Documentation

---

SSH Access to Hetzner VPS

After provisioning, you can connect to your server using the user created by cloud-init. Add the following to your ~/.ssh/config file:

Host YourVPS
  HostName your-domain.com
  User your-user
  Port 2222
  IdentityFile ~/.ssh/your-key

• Replace your-domain.com with your server's public IP or domain if different.
• Ensure your SSH public key matches the one provided to Terraform (TF_VAR_ssh_publickey).
• The SSH port is set to 2222 for security (see cloud-init config).
• The user has passwordless sudo and is the default login user.

Connect with:

ssh HetznerVPS