Wektor is a desktop app that answers questions about your CSV or database in plain English, without uploading your actual data anywhere. Only column names and aggregate stats reach the AI model you connect (your own key). The model writes analysis code from that summary, and it runs locally in a sandboxed process against your real data. Charts, trend detection, PDF reports, and a presentation mode included. Free tier works fully offline with sample data, no signup, no API key.
Hey everyone, maker here.
I built Wektor because every AI data tool I tried had the same catch: to get insights, you had to upload your spreadsheet to someone's server first. That felt wrong for anything with real business data in it, so I built this to work the opposite way: the AI never sees your rows, only a schema summary, and all the actual computation happens locally on your machine.
I'm a solo student building this, so please be honest with feedback, that's genuinely more useful to me than kind words right now. Two things upfront: builds are currently unsigned (your OS will show a warning on first launch, source is public if you want to verify before running), and free-tier AI models can be slower than a paid key since they're shared capacity.
Happy to answer anything about the architecture, the privacy approach, or what's coming next.