Player not loading? Watch on YouTube
The original video is in Portuguese. This tutorial connects LM Studio to Visual Studio Code through Roo Code to build a todo web app with a local LLM. The presenter installs LM Studio on Windows, loads a Gemma 4 model, starts the local server, and selects LM Studio as Roo Code's provider. He sets the context length to at least 32K for this workflow and checks the selected model in the extension.
The first attempt uses a smaller model identified in the transcript as Gemma 4B. It generates separate HTML, CSS and JavaScript files, and the presenter tests adding, completing and deleting tasks. A follow-up request for theme switching and animations encounters an error, which he suspects comes from the extension.
He then downloads Gemma 4 26B, described as nearly 18 GB, and starts again with a more detailed prompt. His PC has an RTX 4060 Ti with 8 GB of video memory and 32 GB of system RAM. He reports high memory use with the larger model. After a failed request, he restarts LM Studio and VS Code, then notices that Roo Code still needs its model selection updated.
The final demo includes light and dark themes, task filters and clearing completed tasks. The coding assistant asks for approval before saving changes; the presenter later enables automatic approvals. The walkthrough shows setup and basic browser checks, rather than a controlled performance comparison or comprehensive application testing.
