Continue and twinny: Ollama setup in VS Code

This Spanish-language tutorial sets up a local AI coding assistant in Visual Studio Code using Continue and twinny separately. The presenter starts with VS Code and Ollama already installed on Windows, with models available locally.

For Continue, the walkthrough selects Ollama as the provider, enables model autodetection and saves the generated configuration. The presenter then uses the same landing-page prompt with 4-billion- and 12-billion-parameter models. On the reported RTX 3070 with 8 GB of video memory, generation takes about 1 minute 14 seconds and 4 minutes 36 seconds respectively. These timings describe the demonstrated runs, rather than general performance expectations.

The browser checks expose faults in both outputs. The smaller model's page has a nonworking services link and theme toggle. The larger model's theme switch works, but some navigation and consultation controls do nothing. The presenter calls the result a tie.

In the twinny demonstration, the extension detects Ollama and its installed models without additional configuration. The presenter selects the 12B model and requests a single HTML presentation explaining Docker, with native JavaScript and no external dependencies. The walkthrough saves the generated file and tests its slides and clickable components. It shows how to run models locally for code generation, while the failed controls illustrate why generated pages still need functional checks.