cai
2026-08-08 6f96f3de875d3711e6c2d4c8fc3fbdcdb00c9173
src/main.rs
@@ -15,7 +15,9 @@
};
use anyhow::{Context, Result, anyhow};
use asr_realtime::{RealtimeAsrConfig, RealtimeAsrOutcome, RealtimeAsrUpload};
use asr_realtime::{
    AudioIngressMetadata, RealtimeAsrConfig, RealtimeAsrOutcome, RealtimeAsrUpload,
};
use audio::{AudioDiagnostics, load_pre_recorded_frames};
use base64::{Engine as _, engine::general_purpose};
use futures_util::StreamExt;
@@ -28,7 +30,7 @@
    options::TrackPublishOptions,
    prelude::{
        DataPacket, LocalAudioTrack, LocalTrack, ParticipantIdentity, RemoteAudioTrack,
        RemoteTrack, Room, RoomEvent, RoomOptions,
        RemoteParticipant, RemoteTrack, Room, RoomEvent, RoomOptions,
    },
};
use reqwest::Client;
@@ -707,6 +709,7 @@
    let enabled = config.user_audio_observer_enabled;
    let simple_vad_enabled = config.simple_vad_enabled;
    let simple_vad_config = config.simple_vad_config.clone();
    let user_participant_identity = config.user_participant_identity.clone();
    let turn_bridge_config = TurnBridgeConfig::from_config(config);
    tokio::spawn(async move {
@@ -728,6 +731,7 @@
            turn_bridge_config,
            http,
            sink,
            user_participant_identity,
        )
        .await;
    })
@@ -743,6 +747,7 @@
    turn_bridge_config: TurnBridgeConfig,
    http: Client,
    sink: Arc<BotAudioOutputSink>,
    user_participant_identity: Option<String>,
) {
    info!(
        call_id = %call_id,
@@ -766,6 +771,15 @@
                publication: _,
                participant,
            } => {
                if user_participant_identity
                    .as_deref()
                    .is_some_and(|expected| participant.identity().to_string() != expected)
                {
                    warn!(call_id = %call_id, trace_id = %trace_id,
                        metadata_status = "wrong_participant",
                        "runtime helper ignored non-user audio participant");
                    continue;
                }
                let participant_alias = redact(&participant.identity().to_string());
                let track_sid_alias = redact(&track.sid().to_string());
                let track_name = track.name();
@@ -791,6 +805,7 @@
                    turn_bridge_config.clone(),
                    http.clone(),
                    sink.clone(),
                    participant,
                );
            }
            RoomEvent::TrackSubscribed {
@@ -2902,6 +2917,7 @@
    turn_bridge_config: TurnBridgeConfig,
    http: Client,
    sink: Arc<BotAudioOutputSink>,
    participant: RemoteParticipant,
) -> JoinHandle<()> {
    tokio::spawn(async move {
        let mut stream = NativeAudioStream::new(
@@ -3027,6 +3043,17 @@
                    if !was_in_speech && is_in_speech {
                        let turn_id = format!("turn-{:04}", vad.turn_index);
                        let ingress_metadata = match AudioIngressMetadata::from_participant(
                            &participant.attributes(),
                        ) {
                            Ok(value) => value,
                            Err(reason) => {
                                warn!(call_id = %call_id, trace_id = %trace_id,
                                    turn_id = %turn_id, metadata_status = "invalid", reason = reason,
                                    "runtime helper ignored invalid audio ingress metadata");
                                None
                            }
                        };
                        match RealtimeAsrUpload::start(
                            http.clone(),
                            turn_bridge_config.realtime_asr_config(),
@@ -3034,6 +3061,7 @@
                            &trace_id,
                            &turn_id,
                            &vad.speech_samples,
                            ingress_metadata.as_ref(),
                        ) {
                            Ok(upload) => {
                                info!(