AiMi
Find anime by vibe, not tags. + Viral Receipt Generator.
5 followers
Find anime by vibe, not tags. + Viral Receipt Generator.
5 followers
Traditional search relies on tags; AiMi understands soul. This hardware-aware RAG engine uses Generative AI (HyDE) to find anime by "vibe" (e.g., "Neon-soaked tragedy"). What's New: Dual-Mode: Auto-optimizes for CPU (Lightweight) or GPU (Generative). 108 Years of Data: 8,000+ anime from 1917-2025. Viral Receipts: A standalone tool that turns your watchlist into aesthetic, shareable store receipts.
Hey Hunters! 👋 I'm Divyanshu. I got tired of getting bad anime recommendations, so I spent 2 months building AiMi to fix it. Most RAG apps require expensive GPUs, but I engineered AiMi to run smoothly on a standard CPU using Nomic Embeddings. I'm bootstrapping this project to fund my dream Deep Learning PC ($7k goal), so I'm offering the Full Source Code for anyone who wants to learn how to build production-grade RAG systems. Exclusive for Product Hunt: Get 20% off with code PHLAUNCH20. Let me know what you think of the 'Receipt' aesthetics! 🧾
The Slider View is gorgeous. ✨
It’s crazy that this is built with Streamlit. It usually looks so clunky, but you somehow made it look like a premium streaming app.
Love the 'Hardware-Aware' feature too. It runs buttery smooth on my MacBook Air. Good luck with the workstation goal!
As a dev, I'm honestly more impressed by the backend here. Getting a Hybrid RAG pipeline (Nomic + BM25) to run this smoothly on a CPU is no joke.
Most people just slap a wrapper on OpenAI, but the fact that you engineered this with HyDE for intent understanding is next level.
Just grabbed the Tier 3 package to see how you handled the vector indexing. Solid work.