Hey HN! I built MyDevice AI because I was frustrated with the privacy trade-offs of existing AI chatbots. I wanted the convenience and capability of modern AI without my conversations being sent to remote servers or used for training. What it is: A completely local AI assistant that runs entirely on your iPhone. Think Perplexity, but your data never leaves your device. Key features:

100% local processing - All AI inference happens on-device using optimized models Web search integration - Built-in SearXNG search for up-to-date information while maintaining privacy. Tiny models on phones can hallucinate quite a bit, but web search acts as an anchor point that makes them dramatically smarter and more trustworthy - they stay focused and factual Lightning fast - Async model loading means no waiting around for initialization Thinking Mode - Powered by Qwen 3 for complex problem-solving 30+ day chat history - Your conversations are preserved locally Completely free and open source - No subscriptions, no data collection

Recent updates:

Redesigned UI that's much cleaner Zero-config SearXNG integration (no more setup headaches) Image previews in search results Tons of bug fixes based on user feedback

I've been dogfooding this for months and it's become my primary AI assistant. Response times are impressive on modern iPhones, and knowing my conversations stay private makes all the difference. The code is available at github.com/navedmerchant/MyDeviceAI if you want to contribute or just see how it works under the hood. Would love to hear your thoughts!