[Discussion] Box — community fork with voice, vision, image gen & more — interest in upstream contributions?

[jegly]

Hi,

I've been maintaining Box, a community fork of AI Edge Gallery, and wanted to
introduce it and gauge interest in contributing some of its features back upstream.

What is Box?
Box layers additional capabilities on top of AI Edge Gallery. It ships as two
builds — stock Android and GrapheneOS/custom ROM support — and has a growing
user base.

---

What Box adds on top of upstream:

Area	What Box adds
Inference engines	llama.cpp (GGUF LLMs), stable-diffusion.cpp (image gen), whisper.cpp (STT) alongside LiteRT
Model import	Import any local GGUF file — not limited to the curated download list
NPU / TPU	All Snapdragon / Tensor / MediaTek variants bundled in one APK
Voice mode	Free talk (continuous hands-free loop) and Vision talk (live camera + voice)
Image generation	On-device Stable Diffusion via GGUF
Speech-to-text	On-device Whisper STT
Document analysis	Attach text files directly in chat
Chat history	Persisted to a SQLCipher-encrypted Room database, resumable across sessions
Security	Biometric app lock, hard offline mode, prompt sanitisation, audit log
Agent skills	20 built-in skills (upstream has 9)
Math rendering	LaTeX expressions rendered as Unicode in chat

---

Question for the team:
Some of these are Box-specific (GGUF, security, GrapheneOS support) but others
feel like natural fits for upstream — particularly voice-to-voice (Whisper STT +
streaming TTS), vision input in AI Chat, and document attachment. Would any of
these be welcome as pull requests? Happy to discuss scope and implementation
before opening anything formally.

Thanks for building AI Edge Gallery — it's been a great foundation.

[dpknag]

Hi @jegly,

Thank you for sharing this! It is great to see how the community is exploring and extending the foundation of the AI Edge Gallery.

Regarding your question about upstream pull requests, please note that the repository is not currently ready for code contributions. We will make a separate announcement when we are ready for OSS users to make contributions to it.

However, we have logged the specific capabilities you mentioned (like Voice-to-Voice and Document Analysis) as feature requests for our team to track.

We really appreciate your detailed feedback and for being such an active part of the community!

[jegly]

Hi @jegly,

Thank you for sharing this! It is great to see how the community is exploring and extending the foundation of the AI Edge Gallery.

Regarding your question about upstream pull requests, please note that the repository is not currently ready for code contributions. We will make a separate announcement when we are ready for OSS users to make contributions to it.

However, we have logged the specific capabilities you mentioned (like Voice-to-Voice and Document Analysis) as feature requests for our team to track.

We really appreciate your detailed feedback and for being such an active part of the community!

Awesome ! you guys and gals are doing amazing work ! Thank you <3

[jegly]

I wanted to flag that Box (issue #779) shipped voice-to-voice + live vision in AI Chat several months ago, and I
raised this in the discussion thread before this PR landed. I'm not claiming this is a copy — the architecture is
different — but it would be great to get a mention in the PR description or README as prior community art. Happy
to contribute the Box approach as a PR too if there's interest