Welcome to my homelab on the web. This is where I tinker with free or low-cost tech that is fun to learn at home, but powerful enough to handle real-world challenges.
Local AI Research Gateway: Ubuntu, Docker, & SearXNG
What is a Local AI Research Gateway? Building a local research stack gives you total privacy by keeping your data and LLM execution entirely on your hardware. By combining Ollama with Open WebUI and SearXNG, you can create a localized version of modern chat interfaces capable of real-time web search. Key Components: Ollama: The backend engine handling model execution locally. Open WebUI: The interface for chat and agent orchestration. SearXNG: A self-hosted metasearch engine that feeds real-time web data directly to the LLM. How to get started: Host Ollama: Run Ollama natively on your Ubuntu host to maximize GPU access for your local models (like gemma2:2b). Deploy via Docker: Use a docker-compose.yml to spin up Open WebUI and SearXNG containers. Bridge the Connection: In Open WebUI, navigate to Admin Panel > Settings > Web Search and set the Query URL to http://host.docker.internal:8080/search?q=<query>. Configure Search: Set the “Search Result Count” to 3 to ensure memory usage stays within the limits of constrained hardware. ...
Running OpenClaw 2026 on a 4GB Laptop GPU
The Challenge: The 4GB VRAM Wall Running modern LLMs like Phi-4 or Qwen3 on local hardware is becoming a necessity as cloud AI costs continue to add up for power users. However, OpenClaw 2026 now requires a large context window (minimum 12k-16k) to handle its agentic workflows and tool-calling capabilities. On an entry-level NVIDIA T500 (4GB), these large windows usually force the system to spill over into System RAM (CPU mode). When this happens, generation speed drops from sluggish to 5+ minutes per response, making the bot unresponsive. ...
Privacy-First AI: Running Ollama
Why Run AI Locally? Ollama allows you to run Large Language Models (LLMs) directly on your desktop. While cloud solutions like ChatGPT or Gemini offer massive horsepower, they require you to send your data to external servers. Running a local LLM provides: Data Sovereignty: Your prompts and data never leave your machine. Zero Cost: No monthly subscriptions or API usage fees. Offline Access: Work without an internet connection. Security: Ideal for analyzing sensitive documents or private codebases. ...
VPN for Your Network with Cloudflare
Overview [Cloudflare] (https://www.cloudflare.com/zero-trust/products/access/) offers a powerful solution for connecting your devices across different networks. This free service allows you to securely link your home office setup, remote cloud images, and laptop through Cloudflare’s global network without exposing your infrastructure to the public internet. What is Cloudflare Tunnel? Cloudflare Tunnel is a service that creates secure, outbound-only connections from your devices to Cloudflare’s edge network. Unlike traditional VPNs, you don’t need to open ports on your firewall or manage complex routing—the tunnel handles all the connectivity for you. ...
Strava API Dashboard
This page pulls data directly from the Strava API. Every few hours, a GitHub Action wakes up, grabs any updates, and saves them to a data file. Hugo then does any converting and using Leaflet.js plots my last starting point on a map. ...