Gemma 4 and Ollama: VS Code setup with Continue

This tutorial connects Gemma 4 to Visual Studio Code through Ollama and the Continue extension. The speaker sets up a local LLM coding assistant and tests it against tasks in a prepared Markdown file for a simple web to-do app.

The demonstration uses the 8 billion parameter model on a MacBook with an M4 Pro chip and 24 GB of RAM. The speaker considers the 26 billion and 31 billion parameter variants too large for this machine. He downloads the chosen model with Ollama pull and checks that it responds in terminal chat before configuring the editor.

Continue detects the installed Ollama models in this setup. The tutorial also shows where to add a provider manually if automatic detection fails. The speaker describes Continue as open source and uses its agent mode to create and edit an HTML file.

Permissions require some troubleshooting. He enables automatic file operations while keeping confirmation for terminal commands, but the model still pauses until he approves its plan. Further approvals appear during the next task. The resulting page displays a task list, then gains an input form that the speaker tests in the browser.

Only the first two tasks are demonstrated. The speaker judges Gemma 4 useful for small, well-defined jobs and describes local models as a complement to paid models rather than a full replacement.