Local-first AI · Vite · OpenAI-compatible

AI Resume Creator

AI Resume Creator is a local-first resume tool built by Crovyx. The app runs in the browser, talks to an OpenAI-compatible model through a Vite proxy, and keeps the user’s resume in local save/restore — with in-app diagnostics that show exactly what is happening, so AI writing assistance never quietly ships personal data to a SaaS backend the user can’t see.

AI Resume Creator — local-first resume tool with OpenAI-compatible model proxy and diagnostics
Local editor → Vite OpenAI-compatible proxy → model call → diagnostics panel → save / restore. Nothing leaves the user’s control invisibly.
Problem

“AI resume tools” are mostly data harvesting in disguise.

Most online resume generators want every job, salary and identifier the user has, store it on someone else’s servers, and quietly route everything through whichever model the vendor picked this quarter — with no way for the user to see what is going on.

  • Personal employment history is uploaded into opaque vendor backends.
  • Model calls are invisible — no way to see the request, response or latency.
  • “Save” means “create an account on our platform.”
  • Switching models or providers is impossible without rewriting the app.
What it does

AI assistance with the data still on the user’s side.

  • Local-first editor — the user’s resume lives in the browser, not in a vendor account.
  • OpenAI-compatible Vite proxy so any compatible model endpoint can be plugged in without rewriting the app.
  • In-app diagnostics panel showing model requests, responses, latency and error states.
  • Explicit save / restore — the user keeps their own file, not a row in someone else’s database.
Capabilities demonstrated

What this project proves.

Local-first AI UX

An AI-assisted tool where the user’s data stays under their control and the model is a service the app uses, not a vendor it surrenders to.

Pluggable model integration

An OpenAI-compatible proxy pattern that makes the underlying model a configuration choice, not a hard-coded dependency.

Observable by design

Diagnostics are part of the product, not a hidden devtools tab — so the user can actually see what the AI layer is doing on their behalf.

Build something similar Want AI in your product without surrendering user data?

Whenever the requirement is “AI assistance, but our users’ data stays ours” — the architecture pattern is the same: local-first, pluggable model layer, visible diagnostics. Crovyx ships these systems end-to-end.