Dandiset Metadata Assistant

A web application for viewing and editing DANDI Archive dandiset metadata with AI assistance.

https://magland.github.io/dandiset-metadata-assistant/

Features

• Load dandiset metadata from the DANDI Archive API by ID and version
• View metadata in a structured, expandable format
• AI chat interface (placeholder) for AI-assisted metadata editing
• Track pending changes with inline color-coded diffs
• API key management stored in browser localStorage
• Resizable split-panel layout for chat and metadata views

Getting Started

npm install
npm run dev

Usage

1. Enter a Dandiset ID (e.g., 000003) and select a version
2. Click "Load" to fetch the metadata
3. (Optional) Configure your DANDI API key to access embargoed dandisets and enable committing changes
4. Use the AI assistant to propose metadata changes
5. Review pending changes (highlighted inline) and commit when ready

Proxy Server

This project includes a Cloudflare Worker proxy server that enables metadata commits by handling CORS and performing server-side validation. See the proxy/README.md for detailed setup and deployment instructions.

Quick Start (Proxy)

1. Install dependencies:

   cd proxy
   npm install

2. Run locally:

   npm run dev

3. Configure the client to use the proxy:

   # In the root directory
   cp .env.example .env
   # Edit .env and set VITE_PROXY_URL if needed

4. Deploy to Cloudflare:

   cd proxy
   npx wrangler login
   npm run deploy

Tech Stack

• React + TypeScript
• Vite
• Material UI
• Cloudflare Workers (proxy server)

License

Apache 2.0

Authors

• Jeremy Magland, Center for Computational Mathematics, Flatiron Institute
• Ben Dichter, CatalystNeuro