canary-digital

Really cool project! Tauri is fantastic for keeping the memory footprint small compared to Electron. You mentioned 'Vibecoding' and building this to get used to AI coding tools. I'm curious: which parts of the Tauri/Rust boundary were the hardest for the AI to get right? In my experience, LLMs are great at the frontend JS/React, but sometimes hallucinate bad IPC (Inter-Process Communication) calls when talking to the Rust backend. Did you run into that issue?

Unlike Germany (or most European countries), the US does not have a centralized, state-run database or federal ministry for evaluating foreign degrees. Education is highly decentralized here.

Instead, the US relies entirely on private credential evaluation agencies. The most universally recognized ones are members of NACES (National Association of Credential Evaluation Services).

If you need an evaluation for immigration (USCIS), university admissions, or a job, you usually have to pay a service like WES (World Education Services) or ECE (Educational Credential Evaluators). They maintain their own massive, proprietary databases of global institutions which act as the de-facto 'Anabin' for the US, but unfortunately, those databases aren't public to browse.

This is a wonderfully elegant use of Service Workers. Intercepting the request at the browser level to do instant local routing instead of waiting for a round-trip to the search provider is one of those 'why didn't I think of that' hacks.

Quick question on the implementation: How are you managing the storage and lookup for the 14,000+ bangs? Are they just bundled as a static JSON in the SW payload, and does that have any noticeable impact on the browser's memory footprint when the Service Worker activates?