Implementation guide · 02
Design an Android capture stack that survives offline use
A native-first architecture for capturing text and media, preserving intent, and delivering to AI workflows after connectivity returns.
- 01
Model the capture intent
Treat selected text, shared media, quick capture, and widget entry as distinct sources with shared metadata.
- 02
Persist before delivery
Write the capture locally before any network operation so interruption does not erase user intent.
- 03
Queue explicit work
Use background work with visible states for prepared, queued, sending, failed, and delivered items.
- 04
Protect configuration
Keep secrets out of source and use encrypted settings or a server-side relay appropriate to the threat model.
- 05
Test the ugly paths
Verify process death, low connectivity, permission denial, large media, and destination changes.
Operating principle