Bloomind

Bloomind - On-device AI that structures your thinking. Fully offline.

by
Bloomind turns voice or text into structured "Brain Flowers" — entirely on your iPhone, with zero network access. Powered by MiniCPM-V (1.3B) via llama.cpp + Metal. Your thoughts never leave the device. • 100% offline — works in airplane mode • Metal GPU inference • GBNF grammar for consistent structured output • On-device RAG for contextual chat One-time purchase: $3.99 (launch month $0.99). No subscription.

Add a comment

Replies

Best
Bloomind
Maker
📌

Hey Product Hunt!
👋I'm an indie iOS dev, and Bloomind is a project born from a personal frustration:

I wanted to use AI to organize my scattered thoughts, but I wasn't comfortable sending my inner reflections to any cloud.So I spent months figuring out how to run a real LLM entirely on-device.

Here's what's under the hood:

MiniCPM-V 1.3B parameters, Q4_K_M quantized
llama.cpp with Metal GPU acceleration
GBNF grammar constraints for structured output
On-device RAG for contextual memory

The result:
you speak or type raw thoughts, and the on-device AI transforms them into structured "Brain Flowers" — each with a core insight and reflections. You can then chat deeper with any thought.The app has literally zero network entitlements — it's not just "we promise not to send data," it physically cannot transmit anything. You can verify this in airplane mode.

A few things I'd love feedback on:
How does the processing speed feel on your device?
Are the structured outputs useful, or do you prefer a different format?
Any thoughts on the Brain Flower metaphor?

Launch month price is $0.99 (normally $3.99).

One-time purchase, no subscription ever.Happy to answer any questions about on-device inference, the architecture, or anything else!

Apple Store: https://apps.apple.com/us/app/bloomind-brain-flower/id6772862305


Built on MiniCPM-V — huge thanks to the OpenBMB team for making on-device multimodal LLMs possible.