Tawus Local AI

Power & Privacy in Your Pocket: Tawus Local AI

Tawus Local AI is an advanced, privacy-first mobile AI assistant and LLM runner that allows you to execute large language models completely locally on your Android device—no cloud, no subscriptions, and no data tracking. Keep your conversations entirely secure and accessible anywhere, even without an internet connection.

Why Choose Tawus Local AI?

100% Local & Offline AI: Powered by a high-performance llama.cpp engine, the app runs open-source models natively on your hardware. Your data never leaves your device, ensuring absolute data security, ultimate privacy, and zero latency from external servers.

Smart Web Integration (RAG Tools): Break the boundaries of traditional offline AI! With integrated web search and URL parsing capabilities, your local model can fetch live web pages, analyze online content, and provide accurate, up-to-date responses based on real-time data.

Granular Hardware & Performance Tuning: Optimize the app to fit your phone's exact hardware profile. Take full control over CPU threads, batch size, and GPU layers (GPU offloading) to push your device to its peak performance.

Flash Attention & KV Cache Quantization: Run larger models with lower RAM footprints. Enable advanced features like Flash Attention and custom KV cache quantization types to accelerate token generation speeds and maximize memory efficiency.

Custom System Prompts & Chat Templates: Tailor the AI’s behavior to your exact needs. Modify the System Prompt to give the model a specific persona or role, and choose from multiple built-in chat templates (Llama, Qwen, etc.) to match the exact syntax requirements of your imported model.

Key Features:

Multi-Session Organization: Seamlessly create, manage, and switch between separate chat threads with a fully local history log.

Real-Time Performance Statistics: Track your engine's health live with detailed indicators showing native heap memory usage, current context length, KV cache status, and processing state.

Pro-Level Inference Parameters: Fine-tune response output by modifying critical sampling settings including Temperature, Top-P, Min-P, Top-K, and Repeat Penalty.

Modern Material Design UI: Built natively with Jetpack Compose, featuring an immersive, high-performance, and gorgeous dark-themed interface engineered for smooth navigation and eye comfort.

Import your favorite open-source models, dial in your performance configurations, and experience an independent, unconstrained, and highly secure AI studio right in the palm of your hand.

Take back control of your data and intelligence with Tawus Local AI!