I built ChatVault because I was frustrated with WhatsApp's native search. It relies on exact keyword matching, which makes finding old messages ("where was that sushi place?") impossible if you don't remember the exact phrasing.

I wanted semantic search, but I wasn't willing to upload my chat history to a cloud vector database or an LLM API.

So I built a strictly local-first search engine using Rust compiled to WebAssembly.

The Architecture: 1. Inference: It runs a quantized BERT model (all-MiniLM-L6-v2) directly in the browser using Candle (Hugging Face's Rust ML framework). No ONNX runtime, just pure Rust Wasm. 2. Vector Store: I implemented a custom in-memory vector store in Rust. It performs normalization on insertion, allowing for fast dot-product similarity searches. 3. Performance: To prevent blocking the main thread during indexing (computing embeddings for thousands of messages), the Wasm module runs inside a Web Worker. 4. The UI (Next.js 16) stays responsive at 60fps. Search: I implemented a hybrid scoring algorithm: (Vector_Score * 0.5) + (Keyword_Score * 0.5) to balance semantic meaning with exact matches.

The app works 100% offline once the model weights (~90MB) are cached.

Live Demo: https://chat-vault-mh.vercel.app/

I'm a student looking to get deeper into systems programming/Rust. I'd love feedback on the memory management between JS and Wasm, as passing strings back and forth was the biggest bottleneck I encountered.

Comments URL: https://news.ycombinator.com/item?id=46685602

Points: 1

# Comments: 1

I started building this startup 2 years ago. My first idea was targeting schools, since kids are the ones who get the biggest benefit out of brain training. However, I struggled to see get any traction.

1 year later, I decided to recreate the product and target it towards Brain Training enthusiasts and people on Self Improvement.

After 1 year and a lot of headaches and fun, the product is finally in a state where anyone can use it to train their brain.

Comments URL: https://news.ycombinator.com/item?id=43076272

Points: 8

# Comments: 4