IntelliDoc Studio

• 

  A screenshot of it parforming its magic

I built AutoDocAI because I kept losing time manually writing README sections whenever a new feature landed; I wanted a single interface that could analyze either uploaded project files or a research topic and surface polished documentation automatically. The idea grew out of exploring Gemini’s reasoning power, pairing it with React + Vite to keep the UI snappy while still letting me tweak prompts and output modes in real time.

The app is made with Vite for lightning-fast builds, React 19 for expressive state-driven UI, and a small Gemini service layer (gemini.ts) that bundles uploads, search prompts, and custom instructions into a request. I learned how to style a cohesive, accessible dashboard with Tailwind-like utility classes, keep the generation button guarded against invalid inputs, and surface errors or source citations via the ResultViewer. Managing file uploads, optional research text, and multiple documentation modes taught me how to keep state predictable while still offering flexibility.

Where it got tricky was error handling: I needed to reset the UI every time generation started so stale results couldn’t mislead users, and I also wanted graceful feedback when Gemini returned nothing useful. Keeping the file list manageable (with remove/clear controls) and letting users swap between upload and research workflows without losing context were design challenges that shaped the final layout.

Built With

• react
• tailwind
• typescript
• vite