timmyy123/LLM-Hub

What it solves

LLM Hub is a privacy-focused mobile application that allows users to run large language models (LLMs) and other generative AI tools entirely on-device. It eliminates the need for cloud-based AI services, ensuring that conversations and data never leave the device, while providing a comprehensive suite of AI capabilities across Android and iOS.

How it works

The app leverages mobile-optimized runtimes and SDKs to accelerate inference using the device's CPU, GPU, and NPU. It supports multiple model formats (including .gguf, .litertlm, and .task) and allows users to import custom models or download them directly from HuggingFace.

On Android, it integrates with Termux to enable an autonomous AI agent capable of executing shell commands and managing files. On iOS, it utilizes the Run Anywhere SDK and MediaGenerationKit for local LLM and media generation.

Who it’s for

• Privacy-conscious users who want local AI without data collection.
• Developers and power users who want to experiment with local LLMs, image/video generation, and AI agents on mobile devices.
• Mobile AI enthusiasts looking for a comprehensive toolset for on-device transcription, translation, and media search.

Highlights

• Multimodal AI Suite: Local chat, image generation (Stable Diffusion), video generation (Stable Video Diffusion), and music generation.
• Autonomous AI Agent: Supports function calling, MCP tools, and Termux terminal command generation with an auto-correction loop on Android.
• Local Media Intelligence: On-device indexing of photos and videos for instant search and "moment finding" using EmbeddingGemma 2.
• Advanced Audio Tools: On-device transcription (Whisper), multilingual TTS, and local voice cloning (Pocket TTS).
• Vibe Coder: A local coding assistant with live HTML/JS previews and Codex integration via Termux on Android.
• 100% Private: No internet required for inference, no accounts, and no tracking.