Problem
Cloud photo storage is convenient, but expensive and restrictive.
iCloud feels alive because it is searchable, always available, and integrated into the phone. External SSDs and HDDs are cheap and have huge capacity, but they feel dead. They store the original files, but the user cannot easily browse, search, or access them from an iPhone unless the drive is physically connected.
Cloud storage is alive but expensive. Local drives are cheap but dead. Silo tries to combine the best of both.
Solution
Silo gives local storage an intelligent iPhone interface.
It keeps originals on local disks while building an intelligent layer of previews, metadata, and embeddings. The goal is to make external storage feel as searchable and accessible as iCloud, without turning personal memories into a monthly subscription.
How it works:
- Desktop app indexes the original photo and video library on local disks
- iPhone stores lightweight previews, metadata, and embeddings
- AI search works across the entire library
- Originals remain local — the iPhone becomes the intelligent interface to the local archive
Technical Foundation
Silo grew out of work on a high-performance local photo organizer that explored EXIF extraction, media inventory generation, trip clustering, duplicate detection, partial hashing, and safe dry-run sorting. This gives Silo a real technical foundation — it is based on practical experiments with large media libraries and external drives.
Key capabilities:
- EXIF inventory extraction using exiftool
- Partial hashing for fast duplicate detection (sample file chunks instead of reading every byte)
- Trip/session detection using date, time, and GPS metadata
- Proxy compression experiments
- Semantic search model exploration (SigLIP, CoreML, MLX)
- Local-first sync between desktop and iPhone over local network
Architecture
Desktop: Rust / Tauri — indexing, scanning, metadata extraction, local database, media processing
iOS: Swift — browsing, searching, preview cache, mobile-first access
AI / ML: Image and video embeddings, semantic search, possible vector database direction (LanceDB)
Current Status
Silo is currently in active prototype and technical exploration. It is a real product direction with performance tests, architecture decisions, and a clear user problem — not starting from a blank page.