Skip to content
twinny

Quick start

Prerequisites

To use twinny, you need access to an inference provider. An inference provider is a local or cloud-hosted server that runs the AI models.

The recommended way to set this up is by using Ollama. Ollama simplifies running AI models locally and exposes them through an OpenAI-compatible API. Performance will vary based on your hardware and the chosen model; please refer to Ollama’s documentation for more details.

Installing the extension

  1. Install the Visual Studio Code extension here or for VSCodium here.

Installing Ollama as an inference provider

  1. Visit the Ollama installation page and follow the instructions to install Ollama on your machine.
  2. Download a model to use with twinny. A good general-purpose model to start with is codellama:7b-instruct. You can run the following command in your terminal:
Terminal window
ollama pull codellama:7b-instruct

To run this model, or another of your choice (see the Supported Models page for more options), use:

Terminal window
ollama run your-chosen-model-name

For example, to run codellama:7b-instruct:

Terminal window
ollama run codellama:7b-instruct

Make sure the model is running in Ollama before proceeding.

Using twinny

Once both the extension and an Ollama model are running, you can start using twinny:

  1. Open VS Code. If it was already open during the extension installation, a restart might be necessary.
  2. Look for the twinny icon in the side panel.

You should also see the 🤖 icon in the status bar, indicating that twinny is ready. This icon will change to a spinner when twinny is communicating with the inference provider.