what is liara-toolkit fr?

Liara Toolkit describes itself as a privacy-first personal AI assistant that runs fully on your own device, either on Android or on desktop, instead of sending your data to a cloud server. According to the project description, it uses a local Qwen3 language model, can call tools to take actions on your behalf, and keeps its memory encrypted so your conversations stay private even if someone gets access to your files. The actual documentation in this repository is very sparse right now. The README file that ships with the project is just the default starter text you get from creating a new Tauri, React, and TypeScript app, the kind of boilerplate a developer sees the moment they scaffold a new project. It points to VS Code extensions for Tauri and Rust but does not explain how to install Liara, how to set up the local model, or what the interface looks like once it is running. Based on the listed topics, the project is built with Rust for the core application logic, paired with Tauri to package it as a lightweight desktop and mobile app, and it references llama.cpp style tooling for running the AI model on-device without internet access. This suggests the team is aiming for something that works offline, which fits the privacy-first claim in the description. The license is listed as PolyForm Noncommercial, which means the source code is visible to read and modify, but you are not allowed to use it for commercial purposes without a separate agreement. Because the project is so new and the documentation has not caught up with the description yet, anyone interested should treat this as an early-stage, work-in-progress toolkit rather than a finished, ready-to-use assistant.