Getting started

Install on Windows

Download the latest Windows x64 Beta online installer from this site. It supports Windows 10 version 1809 or later and Windows 11 on Intel or AMD x64 computers. Windows ARM64 and 32-bit Windows are not supported by this release.

Choose a writable installation folder in the setup wizard, such as a short folder on D: or E:. The small installer downloads the desktop client; first launch downloads and verifies the backend runtime. Keep an internet connection during setup. No separate Node.js or Python installation is required.

The application, backend runtime, chat history, files and application caches stay in the selected installation folder. Uninstalling retains its data folder; reinstall into the same folder to restore your records. Data from older AppData installations is not moved automatically.

Windows still manages shortcuts, registry entries and a small amount of installer temporary data on the system drive. This Beta installer is not code-signed; Windows may show an unknown publisher.

Use the current macOS build

Download the current macOS build for Apple Silicon (ARM64) from this site. It uses a local test signature and has not been notarized by Apple.

After downloading, drag ValiAlpha from its disk image into Applications and open it. First launch prepares the local data directory and service. You still need to configure your own model connection.

Connect your model

Open Settings → Model connections. Select the provider and model, enter the connection details and supply credentials where required. Test the connection, then check a real chat response.

Choose an authorization mode for your work. Review requested actions when prompted; the mode does not expand an application’s selected data or tools.

Start with a small task

Use Files and folders to attach a file or start a chat for a project folder. Ask a focused question and check the answer against the source. Spaces can organize saved pages, attachments and knowledge material as the work grows.

Create and review an application

Open AI applications → Create application. Chat-based creation currently starts with data queries and analysis. Check the selected sources, method and tools, then read the test evidence. Sample tests use synthetic data; review results from the actual task separately.

Publish after your review. Use Start conversation on the published application to work with the same method again.