br-AI-n: Offline Private AI

Bandiago

About this app

br-AI-n is a secure, private, and offline-first AI assistant for your Android device. It allows you to download and run state-of-the-art open-source Large Language Models (LLMs) directly on your phone's hardware with zero cloud dependencies and complete data privacy.

Because br-AI-n executes inference locally, your chats, voice recordings, and image attachments never leave your phone. And whenever you need up-to-the-minute real-world information, optional live Web Search grounds your answers with real-time sources and cited footnotes.

🌟 Key Features
🔒 Offline-First & 100% Private by Default
No cloud APIs, no remote AI servers, and zero data tracking. Your model weights, chat sessions, and prompts remain strictly on your device.

🌐 Optional Live Web Search & Citations
Break free from knowledge cutoff limits! Enable live web search to ground answers with real-time news and verified web sources. Explore cited claims with interactive, superscript footnotes ([1]) and direct source links.

🎭 Customizable & Creative Personas
Tailor your AI’s personality! Choose from built-in personas (Code Expert, Creative Writer, Researcher, Joker, and more) or create your own custom personas. Customize custom prompts, names, and emoji icons with real-time system prompt updates.

🧠 Deep Reasoning & Scratchpad
Watch the model think in real time with transparent reasoning traces and multi-hop search synthesis inside an expandable thinking drawer.

🎙️ Voice Transcription & Real-Time Speech
Speak naturally with local voice transcription, and listen to low-latency streaming text-to-speech (TTS) that reads responses aloud sentence-by-sentence as they generate.

👁️ Vision & Document OCR
Attach images, receipts, diagrams, or documents to transcribe, inspect, and discuss with multimodal visual models.

⚡ Real-Time Hardware & Inference Stats
Monitor real-time token generation speeds (tokens/sec), RAM utilization, context length limits, and storage headroom.

📂 Saved Local Chat History
All conversations are organized locally in a persistent SQLite database. Resume past discussions, search topics, or clear context with ease.

🤖 Broad Open-Source Model Support
Download optimized GGUF quantizations of leading open-weight architectures, including Llama 3.2, Qwen 2.5, Gemma 2, and Phi series directly from Hugging Face.

🛡️ No Accounts, No Subscriptions
No logins, no tracking cookies, and no recurring API fees. Full on-device AI power right in your pocket.

⚙️ Technology
Powered by a high-performance, mobile-optimized C++ inference engine (llama.cpp), br-AI-n leverages your phone's CPU and GPU architecture to deliver state-of-the-art quantized local intelligence.

Note: Generation speed, loading times, and memory requirements depend on your device's processor and total RAM. Check the built-in Hardware Profile in the Model Manager to pick the optimal model size for your device.

What’s new

• Optional Live Web Search: Ground responses with real-time web sources and breaking news.
• Interactive Citations: Tap in-text superscript footnote badges [1] or source cards to open external references.
• Customizable Personas: Edit built-in personas or build custom AI companions with tailored prompts, names, and emoji icons.
• Thinking Drawer: View real-time reasoning steps and search synthesis.
• Stability improvements & bug fixes.