smallm vs OpenHarbor

"I want to use AI on my phone. But I don't want my conversations sent to some server." smallm makes both possible. **Your smartphone becomes AI.** smallm is a **local LLM app** that runs AI models directly on your smartphone. Download and run the latest open-source AI models like Qwen, Gemma, Llama, Phi, Mistral, and DeepSeek right on your device. All processing happens on your device. No cloud API required. ■ Real AI, Running Locally smallm runs genuine large language models (LLMs). Choose from models ranging from 360M to 14B parameters, optimized for your phone's specs. The app automatically detects your device's RAM, storage, and GPU performance, recommending the best model for you. No complicated setup needed. ■ Your Conversations Never Leave Your Device Running locally means **no internet connection required**. Your conversations never leave your smartphone. ・For brainstorming work ideas — Confidential content stays secure ・For editing journals and writing — Private text remains private ・For helping kids with homework — A safe environment for families ・On airplanes or subways with no signal — Offline? No problem ■ AI That Understands Photos With vision-capable models, just attach a photo and the AI reads its contents. Document transcription, scene description, questions about images — your camera connects with AI. ■ No Monthly Fees. One-Time Purchase. No subscription. One purchase unlocks all features. No monthly bills to worry about. ■ For Developers: Local API Server Start the built-in API server to call AI models from other apps or scripts. Perfect for local development and experimentation. 【Free Trial Available】 ・Chat with lightweight models (Qwen3.5 0.8B / SmolLM2 360M) ・Chat history (recent 20 conversations) Experience the power of local AI for free. 【Full Version (One-Time Purchase)】 ・All models unlocked (up to 14B parameters) ・Unlimited chat history ・Local API server ・Data export smallm is developed by a Japanese developer. No user data collection or transmission features included. Real AI on your smartphone. That's smallm. ## What's New - Vision model support: Attach images and ask AI questions - New models added: Phi-4 Mini, Qwen 2.5 14B - Performance improvements - Bug fixes

**OpenHarbor Manager Assistant** is a lightweight management client built for the Lobster Gateway, allowing you to connect to and manage your local or LAN-based AI Agent Gateway anytime, anywhere from your phone. ## Main Features ### WebSocket Gateway Connection Supports both `ws://` and `wss://` protocols. Optional Gateway Token support, compatible with Gateway device authentication (Ed25519) and automatic pairing flow. ### Multiple Lobster Configurations Save and manage multiple Gateway profiles (Lobster 1, Lobster 2, etc.). Quickly switch between them from the top bar, with independent connection status and chat history for each profile. ### Gateway Sessions Connect to the Gateway in operator mode and chat with your AI Agent in real time. Conversation history is shared with the browser Control UI session (`agent:main:main`). ### Status Monitoring Monitor Gateway health, running status, and online devices/channels in real time. ### Model Configuration View and switch AI models (supports popular providers like DeepSeek). Update API Keys with one tap and apply changes instantly without restarting the Gateway. ### Skill Search Search or browse trending Skills through the ClawHub API. Copy installation commands with one tap and run them on the computer hosting the Gateway to install Skills instantly. ## Use Cases * Remotely monitor your local or LAN-based AI Agent from your phone * Switch AI models or update API Keys without opening your computer * Quickly search and install Skills from ClawHub ## Notes This app is a management tool for the Lobster Gateway and does not provide standalone generative AI services. Before using the app, you need to deploy and run your own Lobster Gateway (recommended version: v2026.2.26 or later).