VideoMind
AI-First Video Intelligence
VideoMind turns any video into a fully searchable AI knowledge base. Every video is stored on Shelby Protocol — the decentralised content layer that makes the whole platform possible.
Wallet-signed uploads
Every video is uploaded to Shelby Testnet directly from the user's connected Aptos wallet (Petra). The wallet signs the blob registration transaction — no backend private key involved in video storage.
Decentralised blob storage
Raw video files are stored as Shelby blobs under the path videomind/videos/{id}/raw.{ext}. Each blob is namespaced under the uploader's wallet address and retrievable via the Shelby HTTP gateway.
48-hour expiry + auto-renewal
Shelby Testnet enforces a 48-hour max blob expiration. VideoMind runs a cron daemon every 6 hours that checks for blobs expiring within 12 hours and automatically re-uploads them — keeping your library alive.
ShelbyClientProvider + useUploadBlobs
The frontend wraps the app in ShelbyClientProvider from @shelby-protocol/react. The useUploadBlobs hook handles chunked multipart upload and transaction submission through the wallet adapter.
Shelby Explorer integration
Every uploaded video links directly to the Shelby Explorer for the uploader's account address, giving full on-chain transparency of stored blobs.
Architecture: Shelby as the content layer
VideoMind is designed so Shelby Protocol is the source of truth for all video content. The frontend wallet uploads blobs directly — the backend never proxies video bytes. AI-generated metadata (transcripts, chapters, summaries) lives in Turso and is cross-referenced with Shelby blob addresses, so the full intelligence layer is anchored to on-chain storage.
Upload
Wallet signs blob → Shelby Testnet
Transcribe
OpenAI Whisper — full timestamped transcript
Analyze
Claude AI — chapters, highlights, summary
Store
All outputs persisted in Turso (libSQL)
Chat
Claude answers questions about the video
Search
Semantic search across all videos
VideoMind is built on Shelby Protocol as an advanced integration contribution for the Shelby ecosystem. The full source code is available on GitHub.