Inspiration
I'm Thierry, a mason for 25 years. Health forced me to stop physical work — but not my desire to learn. I named my AI assistant Laetitia after my daughter, stillborn at 6 months; she now lives in every line of code. CodeVaultAI was also born from a simple frustration: snippets scattered across browsers, and not wanting my code in the cloud. I hope my son Kevin, who loves IT, will join the project.
What it does
A secure, 100% offline desktop code vault. It stores, tags and searches snippets, with a Monaco (VS Code) editor, static analysis + Auto-Fix + beautify, full-text search, multi-file import/export and auto-detection of 35+ languages. The local AI assistant, Laetitia, answers questions and reviews code — using the user's own local AI (Ollama/LM Studio) or their own API key. No cloud, no telemetry: your code stays on your machine. A companion web portal offers a public live demo and access requests.
How we built it
Electron 28 desktop app with React 18, Tailwind CSS and Monaco, backed by a local SQLite database (codevault.db). A PHP/Symfony portal serves the public site and access gate, and a browser version (IndexedDB) mirrors the vault. We bundled Tailwind locally, added an AI provider-detection layer, and wired an IPC bridge for tokens and DB operations.
Challenges we ran into
Staying offline-first while still offering AI; detecting/calling local AI servers (Ollama, LM Studio) across CORS; migrating from localStorage to SQLite without data loss; designing a security model where the admin token lives only in the desktop app, never in the browser; and replacing CDN dependencies (Tailwind) with a local build so the UI never breaks offline.
Accomplishments that we're proud of
A fully offline, privacy-first dev tool that actually works; a local AI assistant; a clean security model (desktop-only token, hashed passwords, no telemetry); and an entire ecosystem — vault, web portal, portfolio and dev tools — built solo, assisted by AI.
What we learned
How far AI-assisted development can go for a solo builder; the real trade-offs of offline-first and security; desktop packaging; and that privacy-respecting tools are achievable without sacrificing modern UX.
What's next for CodeVaultAI 2027 — assistant dev IA sécurisé
Populate the app with more developer tools, deepen local-AI integration (models, diff view, smarter auto-fix), add a plugin/skill system, support more languages, refine the public portal — and take it further in upcoming competitions.
Log in or sign up for Devpost to join the conversation.