Added connection support for both omi + omi glasses

[MithilSaiReddy]

@aaravgarg @beastoin check this

Multi-device support (Omi + OmiGlass)

description
Enable simultaneous connection of Omi (audio) and OmiGlass (camera) so both streams feed a single multimodal conversation.

What this delivers

• Connect audio + camera devices at the same time (no more silent disconnects).
• Audio transcripts and image frames stream concurrently to the same conversation.
• Pause/resume affects audio only; camera capture continues.
• Backend now receives multimodal source when both devices are active.
• Backwards compatible when only one device is connected.

Key implementation

• Replaced single device connection with typed device map in devices.dart.
• Split recording state into _audioDevice and _glassDevice in capture_provider.dart.
• Separated audio and photo streaming pipelines.
• Fixed bugs from prior partial migration (null access, incorrect disconnect calls).

Impact
Unlocks real multimodal capture using existing WebSocket + conversation structure with minimal architectural change.

Files

• lib/services/devices/devices.dart
• lib/providers/capture_provider.dart

[gemini-code-assist]

Code Review

This pull request successfully refactors the application to support simultaneous connections for an audio device (Omi) and a camera device (OmiGlass), which is a great step towards multimodal conversations. The changes in CaptureProvider correctly separate the logic for audio and photo streaming. However, the core DeviceService has critical issues in the implementation of multi-device support. The _connectToDevice and ensureConnection methods contain incorrect logic, typos, and duplicated code that will cause compilation and runtime errors. I have provided detailed comments and suggestions to address these critical problems.

[MithilSaiReddy]

@aaravgarg @mdmohsin7 can u guys check this pr ?

[beastoin]

Hey 👋 — thanks for putting this together! Before we can review, could you share a quick live demo (screenshot, screen recording, or terminal output) showing this working on your local or dev environment?

In the AI era, writing code is the easy part — what really makes a PR stand out is proof that it works end-to-end. A short video or even a screenshot goes a long way in helping reviewers feel confident about merging.

Feel free to update this PR whenever you have something to show. Thanks! 🙏