Analyse de l'application Offline AI Chat: MemoryAI pour le 6 novembre

MemoryAI — an AI assistant that lives on your phone and works without internet. • Ask questions, write, and summarise your documents — by typing, or by talking hands-free • Your chats stay on your phone: no account, no ads, no subscription • Kid Mode with parental controls: a PIN, a content filter and a locked model Download one AI model once (0.3–4 GB, best over Wi-Fi), then use it anywhere: on a plane, abroad, with no signal. Works on devices with 4 GB+ RAM. 🔒 Complete Privacy Run powerful AI models directly on your phone. Your conversations, documents and memories stay on your device: no AI servers, no login required. The app sends only anonymous diagnostics and usage data, never the content of your chats. 🧠 AI Features • Chat with AI — multi-turn conversations that keep the whole chat in context, up to the model's limit • Agent Memory (optional) — turn it on in Settings and the AI remembers what you tell it, in your own words, stored locally on your device; view, edit or delete memories any time. Requires an embedding model • Chat Search — search and filter your conversations by title or message • Continuous Voice Mode — fully hands-free: speak naturally, and the AI listens, responds, and speaks back without tapping • Voice Input — speak instead of typing with on-device transcription • Text-to-Speech — listen to AI responses read aloud, or enable auto-speak for hands-free use • Document Analysis (RAG) — import PDF, TXT and MD files and ask questions about them • Image Understanding — analyze photos with Vision AI models; select multiple images at once • Chat Templates — 20 pre-built AI personas (code helper, writing coach, tutor, and more) • Benchmarking — measure model speed (tokens/sec) directly on your device • Context Window Display — see current vs. max token usage, so you never exceed the model's limit 🔧 Power User Features • Import custom GGUF models — bring any compatible model from HuggingFace • Dual engine — Gemma (MediaPipe / LiteRT) + llama.cpp; choose the backend that fits your needs • HuggingFace Token Management — save and manage your token for seamless gated model downloads • Model parameters — adjust temperature, top-k, top-p, max tokens per model • Pause & Resume Downloads — control large model downloads to save data or battery • Backup & Restore — save your chats and settings to Google Drive 👨‍👩‍👧 Kid / Safe Mode • PIN-protected parental controls • Content filter blocks inappropriate words before they reach the AI • Model Safety Checks — automatic warnings to prevent unverified imported models from running in Kid Mode • Locks the active model — child cannot switch or change settings • Age-appropriate AI responses, guided by a safety prompt • Parental Unlock — temporarily access settings and documents from the chat screen without disabling Kid Mode 📱 Works Everywhere • Android • 10+ AI models: Gemma, Llama, Phi, Qwen, DeepSeek, and more • Available in 6 languages: English, Italian, German, French, Spanish, Portuguese 💎 Completely Free • Unlimited chat, voice, vision, documents, image OCR • Documents: PDF, TXT, MD and images (JPG, PNG) • No ads, no subscriptions, no in-app purchases Built by Facchinits — privacy-first AI for everyone.