FileNest

Inspiration

Folders keep growing, but the useful context inside them disappears. I wanted a practical desktop companion that could help me find, understand, and organize the files I already have—without making a cloud upload the default.

What it does

FileNest is a local-first desktop app for macOS and Windows. It watches chosen folders, waits until new files are stable, and indexes them in the background. It extracts document text, runs OCR when needed, and keeps searchable metadata, notes, paths, and content on the device.

Search works when I remember a filename and when I only remember the meaning. FileNest combines filenames, paths, extracted content, notes, and semantic retrieval to surface relevant files. Find with Chat starts with matching files and continues as a conversation; Chat with File keeps answers grounded in one selected document.

How we built it

The macOS app uses Swift and SwiftUI; the Windows app uses Electron, React, and TypeScript. The product combines local indexing, SQLite-backed storage, OCR, retrieval-augmented generation, vector search, and configurable local models through Ollama. Cloud providers are optional and only used when the user chooses them.

For OpenAI Build Week, I worked with Codex powered by GPT-5.6 to extend FileNest's indexing, retrieval, chat, model-management, and cross-platform workflows. Codex accelerated implementation, debugging, and testing, while I made the product, privacy, and architecture decisions.

Challenges we ran into

The hard part was keeping the experience responsive while background indexing, OCR, and retrieval operate on real folders. I also had to make search results explainable, preserve a local-first default, and keep the product coherent across macOS and Windows.

Accomplishments that we're proud of

FileNest turns a passive file collection into something searchable and conversational while keeping the original source files close at hand. The app makes it possible to retrieve a document by what it contains, not just what it was named.

What we learned

Reliable AI file workflows need careful boundaries: stable-file detection, clear evidence for chat answers, and privacy controls are as important as the model itself.

What's next for FileNest

Next I plan to expand integrations, refine organization suggestions, improve model management, and continue making on-device knowledge retrieval faster and easier to trust.

Built With

• openai
• vector