Module tier3 

Tier 3: deep D-STAR protocol connections and voice capture.

Establishes full D-STAR protocol connections (DPlus, DExtra, DCS) to reflectors via the dstar-gateway crate’s AsyncSession API. Captures voice streams including:

• D-STAR header metadata (callsign, suffix, UR call, RPT fields)
• AMBE voice frames decoded to PCM via mbelib-rs, then encoded to MP3
• Slow data (D-STAR text messages, DPRS position reports)

The capture pipeline for each stream:

AsyncSession<P> -> VoiceStart (header)
                -> VoiceFrame x N (AMBE + slow data)
                -> VoiceEnd
                    |
            mbelib-rs: AMBE -> PCM
                    |
            mp3lame-encoder: PCM -> MP3
                    |
            Postgres: streams table (metadata + audio_mp3)

Connections are managed by a priority queue: user-pinned reflectors take precedence, followed by auto-promoted reflectors ranked by activity score. Idle sessions (no voice activity beyond the configured timeout) are disconnected to free slots for more active reflectors.

Module layout

• decoder — AMBE-to-PCM-to-MP3 audio pipeline using mbelib-rs and mp3lame-encoder.
• capture — Per-stream state tracker (StreamCapture) and cross-stream dispatcher (CaptureManager).

Current status

The audio decode pipeline and capture-state machinery are complete and unit-tested. The remaining orchestration work — maintaining a pool of AsyncSession tasks keyed off the Tier 1/Tier 2 reflector scores — is left to a follow-up. The public run entry point currently logs its configuration and blocks forever; process_completed_stream is the production-grade finalizer that the future orchestrator will wire in.

Modules

capture 🔒

Voice-stream capture state manager.

decoder 🔒

AMBE-to-PCM-to-MP3 audio decode pipeline.

Functions

process_completed_stream 🔒

Finalizes one captured voice stream: MP3-encodes the audio and persists the row to PostgreSQL.

run 🔒

Runs the Tier 3 voice capture orchestrator.