Add interoperable LXMF voice messages

[torlando-tech]

Summary

• add interoperable asynchronous LXMF voice messages using the standardized FIELD_AUDIO representation with Ogg/Opus payloads
• record finalized audio through LXST v0.0.8, with permission handling, preview, cancellation, failure recovery, persistence, and retry
• derive duration from validated Ogg/Opus metadata and waveform levels from decoded PCM
• add accessible compact playback bubbles with play/pause, waveform progress, duration, timestamp, and delivery state
• enforce bounded payload/cache behavior and tokenized mutual exclusion between calls and voice recording
• preserve voice and generic attachments through Binder transport and failed-message retry

Fixes #463

Verification completed

• LXST unit tests and physical recorder instrumentation passed on a Samsung SM-G998U1
• recorded Ogg/Opus independently decoded successfully
• official Python LXMF to Columba and Columba to official Python voice-message delivery passed
• focused Ktlint, Detekt, recorder, player, persistence, retry, call/recording concurrency, and ViewModel suites passed
• Kotlin and Python backend variants compile with LOCAL_LXST_KT unset against published LXST v0.0.8
• physical recording, preview, cancellation, failed-send retention, playback, decoded-PCM waveform, and compact-bubble checks passed on the development candidate

Verification still running

• final clean full matrix against exact head 10fc50a9a483594821044b23f645499c048ba9d7
• independent exact-head review
• exact-head APK installation and final physical idle/playing visual confirmation

Risk and rollback

The highest-risk areas are Android microphone lifecycle ownership, binary attachment persistence, and playback metadata concurrency. The implementation uses tokenized microphone leases, bounded loading, atomic metadata admission, stale-cache cleanup, and regression coverage for duplicate starts, stale releases, retry reconstruction, and failed recorder finalization. Rollback is the removal of this feature stack and the LXST dependency pin update.

[greptile-apps]

Greptile Summary

The PR adds interoperable Ogg/Opus LXMF voice messages across recording, playback, persistence, retry, and Binder transport.

• Adds bounded audio parsing, loading, metadata extraction, waveform decoding, and playback UI.
• Preserves standardized audio fields through attachment extraction and failed-message retry.
• Coordinates microphone ownership between voice recording and calls, including synchronous teardown cleanup.
• Extends IPC transport and interoperability tests for binary audio attachments.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
app/src/main/java/network/columba/app/ui/model/MessageMapper.kt	Maps standardized audio fields into bounded UI payload references; the previously reported unbounded nested traversal is fixed.
app/src/main/java/network/columba/app/ui/model/AudioAttachmentLoader.kt	Resolves bounded inline and repository-owned file payloads with canonical-path, size, cycle, and nesting checks.
data/src/main/java/network/columba/app/data/repository/ConversationRepository.kt	Extracts large audio payloads while preserving the enclosing mode-and-payload array required by playback and retry.
app/src/main/java/network/columba/app/viewmodel/MessagingViewModel.kt	Integrates recording, sending, retry, and microphone admission; teardown now releases microphone ownership in a synchronous finally block.
rns-ipc/src/main/kotlin/network/columba/app/rns/ipc/AttachmentBlob.kt	Extends out-of-band Binder transport to preserve binary audio and generic attachments without oversized inline transactions.
app/src/main/java/network/columba/app/audio/VoiceMessagePlayer.kt	Adds lifecycle-managed Ogg/Opus loading, metadata caching, waveform analysis, temporary-file playback, and cleanup.

Sequence Diagram

sequenceDiagram
    participant UI as Messaging UI
    participant VM as MessagingViewModel
    participant Audio as Recorder / Player
    participant Repo as ConversationRepository
    participant IPC as RNS IPC
    participant Peer as LXMF Peer
    UI->>VM: Record and send voice message
    VM->>Audio: Record finalized Ogg/Opus
    Audio-->>VM: Audio bytes
    VM->>Repo: Persist FIELD_AUDIO
    VM->>IPC: Send [AM_OPUS_OGG, bytes]
    IPC->>Peer: Deliver LXMF message
    Peer-->>IPC: Receive FIELD_AUDIO
    IPC->>Repo: Persist inline or extracted payload
    Repo-->>VM: Load message
    VM->>Audio: Resolve, validate, and play payload
    Audio-->>UI: Metadata, waveform, and playback state

Loading

_{Reviews (4): Last reviewed commit: "fix: always release voice recording leas..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!