Setup

Prerequisites

• .NET 10 SDK. All projects target net10.0. There is no global.json, so any 10.x SDK works.
• A desktop OS supported by Avalonia: Windows, Linux, or macOS.

No other tooling is required for a normal build. The AI features need no setup at build time; models are downloaded at runtime by the app itself.

Build and run

From the repository root:

dotnet build MnemoApp.sln
cd Mnemo.UI
dotnet run

The app starts with a fresh local database in your user data folder on first run (Windows: %LocalAppData%\Mnemo\). Deleting that folder resets the app to a first-run state.

Tests

dotnet test MnemoApp.sln

Tests live in Mnemo.Infrastructure.Tests (xUnit). There is no CI on pull requests; run the tests locally before pushing. See Testing.

IDE notes

Visual Studio, Rider, and VS Code with the C# Dev Kit all work. Views are plain AXAML files; the Avalonia previewer is optional.

Useful scripts

Script	Purpose
scripts/pack-local.ps1	Builds a local Velopack installer for testing the packaged app, without uploading anything. Parameters: -Rid (default win-x64), -Version.
scripts/GeneratePerfNotes/	Console tool that generates .mnemo packages full of large notes (25 to 1500 blocks) for editor performance testing.

Developer mode

The app has a hidden developer settings section (tap the Settings page title seven times). It exposes performance diagnostics, chat dataset logging, and the experimental cloud teacher-model toggles used for AI training work. None of this is needed for ordinary development.