Local and custom models

Prepare the endpoint

Start your model server using its own documentation and load a model before connecting Virillio. Record the compatible API base URL, exact model identifier, and any required credential.

A local server must be reachable from the computer running Virillio Code. A remote server must use a trustworthy endpoint and appropriate transport security.

Add a custom provider

1. Open Settings → Providers.
2. Choose the custom-provider option.
3. Enter a recognizable name and the server’s OpenAI-compatible base URL.
4. Add authentication if the server requires it, and configure the model identifier offered by that server.
5. Choose that provider and model in a session, then send a short prompt.

For example, a server running on your own computer might advertise http://localhost:1234/v1. This is an illustrative address, not a server bundled with Virillio. Use the actual URL from your model server.

Check capabilities

Text generation alone does not guarantee compatibility with coding workflows. Tool calling, image inputs, streaming, and context sizes vary between models and servers. Test a small read-only project request before relying on a model for edits.

Troubleshoot the connection

If the server cannot be reached, check that it is running and that the host and port are correct. If the model is missing, compare the identifier with the server’s model list. If tool calls fail, check the model’s documented tool support.

Running a model locally can keep inference on your machine, but connecting other remote plugins or services can still send data elsewhere. See privacy and data handling.