what is atlas fr?

Atlas is a personal assistant that runs mainly on your own computer, using a folder of Markdown notes, in the format used by the note-taking app Obsidian, as its memory. It can be talked to by voice or by typing, from a desktop chat window or through WhatsApp, and it remembers past conversations rather than starting fresh each time. If the AI model it is using errors out mid conversation, Atlas automatically switches to a backup model and tells you it did so. The assistant can read images, PDFs, and voice notes, converting each into something the model can use: describing a picture, pulling text out of a PDF, or transcribing spoken audio. Every conversation is saved as a journal entry, and Atlas writes its own permanent notes too, marking them as facts, reusable skills, or lessons from mistakes, then linking each new note to related ones. When it lacks a capability it needs, Atlas can write a new tool for itself following a standard called Model Context Protocol, and the app checks that the tool works before turning it on. Voice input uses a wake word, hey atlas by default, and speech is transcribed locally on your machine, so only the resulting text goes to whichever model you have configured. You can point Atlas at a model running locally through Ollama or LM Studio, so nothing leaves your computer, or at any OpenAI-compatible service. WhatsApp support is set up with a one-time QR code scan and then starts automatically every time Atlas runs, and there is an optional connector for searching files in Google Drive. Getting started means cloning the repository and running a setup script that creates a Python environment, builds the memory vault folder, detects or configures a model backend, and optionally walks through WhatsApp and Google Drive setup. It needs Python 3.11 or newer, and Node.js if WhatsApp is wanted, with every setup step skippable and rerunnable later. Every action Atlas takes is logged to a file for anyone who wants a full audit trail.