Player not loading? Watch on YouTube
This tutorial configures OpenClaw with Gemma 4 through Ollama, then adds a self-hosted SearXNG web search service in Docker. It assumes an existing OpenClaw installation and demonstrates the local model configuration, model selection and gateway restart.
The speaker describes E2B and E4B as smaller models for phones that accept text, images, video and audio, while 26B and 31B accept text and images. He reports good tool calling results with E4B and says he runs an 18 GB 26B model comfortably on a 24 GB MacBook Pro. The downloaded models shown in Ollama are E4B at 9.6 GB and E2B at 7.2 GB. These examples are his reported experience, rather than general hardware requirements. He also explains an optional Ollama cloud configuration for machines unable to run models locally.
For search, the walkthrough installs Docker, connects SearXNG to OpenClaw and edits settings.yaml to enable JSON alongside HTML. OpenClaw requires the JSON format in this setup. After restarting the container, the first search attempt fails; the speaker retries and reports receiving current news results.
