From 949be8e5f1a75286b8b517dacda76ed27646e77b Mon Sep 17 00:00:00 2001
From: cai <cai@nbcai.cc>
Date: Tue, 11 Aug 2026 21:23:54 +0800
Subject: [PATCH] fix(helper): gate audio observer on fixture ack
---
src/main.rs | 1030 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--
1 files changed, 986 insertions(+), 44 deletions(-)
diff --git a/src/main.rs b/src/main.rs
index d357e4a..94b6589 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -59,6 +59,123 @@
const STREAM_TIMING_FIRST_CHUNK: u64 = 1;
const STREAM_TIMING_MAX_ELAPSED_MS: u64 = 5_000;
const STREAM_TIMING_SOURCE: &str = "stream_anchor_monotonic";
+const CONTROLLED_FIXTURE_PROBE_TOPIC: &str = "controlled_fixture_attribute_probe";
+const CONTROLLED_FIXTURE_ACK_TOPIC: &str = "controlled_fixture_attribute_ack";
+const CONTROLLED_FIXTURE_PROTOCOL_VERSION: u64 = 1;
+const CONTROLLED_FIXTURE_GENERATION: u64 = 1;
+const CONTROLLED_FIXTURE_PROBE_RECHECKS: usize = 3;
+const CONTROLLED_FIXTURE_PROBE_RECHECK_DELAY: Duration = Duration::from_millis(25);
+const CONTROLLED_FIXTURE_PROBE_TTL: Duration = Duration::from_millis(250);
+
+#[derive(Debug, Deserialize)]
+#[serde(rename_all = "camelCase")]
+struct ControlledFixtureAttributeProbe {
+ #[serde(rename = "type")]
+ message_type: String,
+ protocol_version: u64,
+ call_id_hash: String,
+ call_trace_id_hash: String,
+ generation: u64,
+ client_fixture_sequence: String,
+}
+
+#[derive(Debug, Serialize)]
+#[serde(rename_all = "camelCase")]
+struct ControlledFixtureAttributeAck {
+ #[serde(rename = "type")]
+ message_type: &'static str,
+ protocol_version: u64,
+ call_id_hash: String,
+ call_trace_id_hash: String,
+ generation: u64,
+ client_fixture_sequence: String,
+ result: &'static str,
+ #[serde(skip_serializing_if = "Option::is_none")]
+ input_source_category: Option<&'static str>,
+ #[serde(skip_serializing_if = "Option::is_none")]
+ reject_reason: Option<&'static str>,
+}
+
+#[derive(Debug)]
+struct PendingControlledFixtureProbe {
+ sender: ParticipantIdentity,
+ sequence: String,
+ expires_at: Instant,
+}
+
+fn sha256_hex(value: &str) -> String {
+ let mut digest = Sha256::new();
+ digest.update(value.as_bytes());
+ digest
+ .finalize()
+ .iter()
+ .map(|byte| format!("{byte:02x}"))
+ .collect()
+}
+
+fn controlled_fixture_probe(
+ payload: &[u8],
+ call_id: &str,
+ trace_id: &str,
+ sender: &ParticipantIdentity,
+ expected_participant: Option<&str>,
+) -> Option<PendingControlledFixtureProbe> {
+ if !is_bound_user_participant(sender.as_str(), expected_participant) {
+ return None;
+ }
+ let probe: ControlledFixtureAttributeProbe = serde_json::from_slice(payload).ok()?;
+ if probe.message_type != CONTROLLED_FIXTURE_PROBE_TOPIC
+ || probe.protocol_version != CONTROLLED_FIXTURE_PROTOCOL_VERSION
+ || probe.generation != CONTROLLED_FIXTURE_GENERATION
+ || probe.call_id_hash != sha256_hex(call_id)
+ || probe.call_trace_id_hash != sha256_hex(trace_id)
+ || probe.client_fixture_sequence.trim().is_empty()
+ {
+ return None;
+ }
+ Some(PendingControlledFixtureProbe {
+ sender: sender.clone(),
+ sequence: probe.client_fixture_sequence,
+ expires_at: Instant::now() + CONTROLLED_FIXTURE_PROBE_TTL,
+ })
+}
+
+fn classify_controlled_fixture_attributes(
+ actual_participant: &str,
+ expected_participant: Option<&str>,
+ attributes: &std::collections::HashMap<String, String>,
+ requested_sequence: &str,
+) -> Result<(), &'static str> {
+ if !is_bound_user_participant(actual_participant, expected_participant) {
+ return Err("wrong_participant");
+ }
+ let source = attributes
+ .get("inputSourceCategory")
+ .map(String::as_str)
+ .unwrap_or("");
+ if source != "controlled_fixture" {
+ return Err(if source.is_empty() {
+ "missing_attributes"
+ } else {
+ "wrong_source"
+ });
+ }
+ let sequence = attributes
+ .get("clientFixtureSequence")
+ .map(String::as_str)
+ .unwrap_or("");
+ if sequence.trim().is_empty() {
+ return Err("missing_sequence");
+ }
+ if sequence != requested_sequence {
+ return Err("wrong_sequence");
+ }
+ Ok(())
+}
+
+fn controlled_fixture_observer_gate(pending_probe: bool, probe_result: Option<bool>) -> bool {
+ !pending_probe || probe_result == Some(true)
+}
#[tokio::main(flavor = "multi_thread")]
async fn main() -> Result<()> {
@@ -737,6 +854,10 @@
})
}
+fn is_bound_user_participant(identity: &str, expected: Option<&str>) -> bool {
+ expected.is_none_or(|value| identity == value)
+}
+
async fn observe_user_audio_events(
mut events: UnboundedReceiver<RoomEvent>,
call_id: String,
@@ -764,6 +885,10 @@
"runtime helper user_track_subscribe_requested"
);
+ let mut current_user_participant: Option<RemoteParticipant> = None;
+ let mut pending_probe: Option<PendingControlledFixtureProbe> = None;
+ let mut acknowledged_probe_sequences = HashSet::new();
+
while let Some(event) = events.recv().await {
match event {
RoomEvent::TrackSubscribed {
@@ -771,10 +896,10 @@
publication: _,
participant,
} => {
- if user_participant_identity
- .as_deref()
- .is_some_and(|expected| participant.identity().to_string() != expected)
- {
+ if !is_bound_user_participant(
+ &participant.identity().to_string(),
+ user_participant_identity.as_deref(),
+ ) {
warn!(call_id = %call_id, trace_id = %trace_id,
metadata_status = "wrong_participant",
"runtime helper ignored non-user audio participant");
@@ -793,6 +918,26 @@
track_source = %track_source,
"runtime helper user_track_subscribed"
);
+ let pending_sequence = pending_probe.as_ref().map(|probe| probe.sequence.clone());
+ let participant_for_probe = participant.clone();
+ let probe_result = process_controlled_fixture_probe(
+ &mut pending_probe,
+ &mut acknowledged_probe_sequences,
+ Some(&participant_for_probe),
+ &sink,
+ &call_id,
+ &trace_id,
+ user_participant_identity.as_deref(),
+ )
+ .await;
+ if !controlled_fixture_observer_gate(pending_sequence.is_some(), probe_result) {
+ warn!(
+ call_id = %call_id,
+ trace_id = %trace_id,
+ "runtime helper withheld audio observer until controlled fixture ACK"
+ );
+ continue;
+ }
spawn_user_audio_frame_observer(
track,
call_id.clone(),
@@ -805,8 +950,41 @@
turn_bridge_config.clone(),
http.clone(),
sink.clone(),
+ user_participant_identity.clone(),
participant,
);
+ current_user_participant = Some(participant_for_probe);
+ }
+ RoomEvent::DataReceived {
+ payload,
+ topic: Some(topic),
+ participant: Some(sender),
+ ..
+ } if topic == CONTROLLED_FIXTURE_PROBE_TOPIC => {
+ if let Ok(probe) =
+ serde_json::from_slice::<ControlledFixtureAttributeProbe>(&payload)
+ {
+ if acknowledged_probe_sequences.contains(&probe.client_fixture_sequence) {
+ continue;
+ }
+ }
+ pending_probe = controlled_fixture_probe(
+ &payload,
+ &call_id,
+ &trace_id,
+ &sender.identity(),
+ user_participant_identity.as_deref(),
+ );
+ process_controlled_fixture_probe(
+ &mut pending_probe,
+ &mut acknowledged_probe_sequences,
+ current_user_participant.as_ref(),
+ &sink,
+ &call_id,
+ &trace_id,
+ user_participant_identity.as_deref(),
+ )
+ .await;
}
RoomEvent::TrackSubscribed {
track: RemoteTrack::Video(track),
@@ -847,6 +1025,79 @@
_ => {}
}
}
+}
+
+async fn process_controlled_fixture_probe(
+ pending_probe: &mut Option<PendingControlledFixtureProbe>,
+ acknowledged_probe_sequences: &mut HashSet<String>,
+ participant: Option<&RemoteParticipant>,
+ sink: &BotAudioOutputSink,
+ call_id: &str,
+ trace_id: &str,
+ expected_participant: Option<&str>,
+) -> Option<bool> {
+ let Some(probe) = pending_probe.take() else {
+ return None;
+ };
+ if acknowledged_probe_sequences.contains(&probe.sequence) || Instant::now() > probe.expires_at {
+ return Some(false);
+ }
+ let Some(participant) = participant else {
+ *pending_probe = Some(probe);
+ return None;
+ };
+ if participant.identity() != probe.sender {
+ return Some(false);
+ }
+ let mut decision = Err("timeout");
+ for attempt in 0..CONTROLLED_FIXTURE_PROBE_RECHECKS {
+ if Instant::now() > probe.expires_at {
+ break;
+ }
+ let attributes = participant.attributes();
+ decision = classify_controlled_fixture_attributes(
+ &participant.identity().to_string(),
+ expected_participant,
+ &attributes,
+ &probe.sequence,
+ );
+ if decision.is_ok() || !matches!(decision, Err("missing_attributes")) {
+ break;
+ }
+ if attempt + 1 < CONTROLLED_FIXTURE_PROBE_RECHECKS {
+ sleep(CONTROLLED_FIXTURE_PROBE_RECHECK_DELAY).await;
+ }
+ }
+ let (result, input_source_category, reject_reason) = match decision {
+ Ok(()) => ("observed", Some("controlled_fixture"), None),
+ Err(reason) => ("rejected", None, Some(reason)),
+ };
+ let ack = ControlledFixtureAttributeAck {
+ message_type: CONTROLLED_FIXTURE_ACK_TOPIC,
+ protocol_version: CONTROLLED_FIXTURE_PROTOCOL_VERSION,
+ call_id_hash: sha256_hex(call_id),
+ call_trace_id_hash: sha256_hex(trace_id),
+ generation: CONTROLLED_FIXTURE_GENERATION,
+ client_fixture_sequence: probe.sequence.clone(),
+ result,
+ input_source_category,
+ reject_reason,
+ };
+ let payload = match serde_json::to_vec(&ack) {
+ Ok(payload) => payload,
+ Err(_) => return Some(false),
+ };
+ let local_participant = sink.room.local_participant();
+ let publish = local_participant.publish_data(DataPacket {
+ payload,
+ topic: Some(CONTROLLED_FIXTURE_ACK_TOPIC.to_string()),
+ reliable: true,
+ destination_identities: vec![probe.sender],
+ });
+ if publish.await.is_ok() {
+ acknowledged_probe_sequences.insert(probe.sequence);
+ }
+ Some(result == "observed")
}
async fn handle_finished_turn(
@@ -2917,6 +3168,7 @@
turn_bridge_config: TurnBridgeConfig,
http: Client,
sink: Arc<BotAudioOutputSink>,
+ expected_participant_identity: Option<String>,
participant: RemoteParticipant,
) -> JoinHandle<()> {
tokio::spawn(async move {
@@ -2996,6 +3248,7 @@
None
};
let mut realtime_asr_upload: Option<RealtimeAsrUpload> = None;
+ let mut last_fixture_sequence: Option<String> = None;
while let Some(drained) = frame_rx.recv().await {
let frame = drained.frame;
@@ -3029,8 +3282,11 @@
if let Some(vad) = simple_vad.as_mut() {
if vad_enabled_gate.load(Ordering::Acquire) {
- let was_in_speech = vad.in_speech;
- let turn = vad.observe_frame(
+ let participant_identity = participant.identity().to_string();
+ let (was_in_speech, is_in_speech, turn) = observe_bound_participant_frame(
+ &participant_identity,
+ expected_participant_identity.as_deref(),
+ vad,
&call_id,
&trace_id,
&participant_alias,
@@ -3038,41 +3294,15 @@
frame_count,
elapsed_ms,
&frame,
+ http.clone(),
+ turn_bridge_config.realtime_asr_config(),
+ || participant.attributes(),
+ &mut realtime_asr_upload,
+ &mut last_fixture_sequence,
+ turn_bridge_config.asr_realtime_enabled,
);
- let is_in_speech = vad.in_speech;
- if !was_in_speech && is_in_speech {
- let turn_id = format!("turn-{:04}", vad.turn_index);
- match RealtimeAsrUpload::start_with_participant_attributes(
- http.clone(),
- turn_bridge_config.realtime_asr_config(),
- &call_id,
- &trace_id,
- &turn_id,
- &vad.speech_samples,
- || participant.attributes(),
- ) {
- Ok(upload) => {
- info!(
- call_id = %call_id,
- trace_id = %trace_id,
- turn_id = %turn_id,
- "runtime helper asr_realtime_session_started"
- );
- realtime_asr_upload = Some(upload);
- }
- Err(error) if turn_bridge_config.asr_realtime_enabled => {
- warn!(
- call_id = %call_id,
- trace_id = %trace_id,
- turn_id = %turn_id,
- error = %safe_error(&error.to_string()),
- "runtime helper asr_realtime_start_failed_fallback"
- );
- }
- Err(_) => {}
- }
- } else if was_in_speech {
+ if was_in_speech {
let push_failed = realtime_asr_upload
.as_mut()
.and_then(|upload| upload.push_48k_samples(frame.data.as_ref()).err());
@@ -3189,6 +3419,176 @@
"runtime helper user_audio_stream_ended"
);
})
+}
+
+fn observe_frame_and_start_session<F>(
+ vad: &mut SimpleVad,
+ call_id: &str,
+ trace_id: &str,
+ participant_alias: &str,
+ track_sid_alias: &str,
+ frame_count: u64,
+ elapsed_ms: u64,
+ frame: &AudioFrame<'_>,
+ http: Client,
+ config: RealtimeAsrConfig,
+ read_attributes: F,
+ upload_slot: &mut Option<RealtimeAsrUpload>,
+ last_fixture_sequence: &mut Option<String>,
+ realtime_enabled: bool,
+) -> (bool, bool, Option<FinishedSpeechTurn>)
+where
+ F: FnOnce() -> std::collections::HashMap<String, String>,
+{
+ let was_in_speech = vad.in_speech;
+ let turn = vad.observe_frame(
+ call_id,
+ trace_id,
+ participant_alias,
+ track_sid_alias,
+ frame_count,
+ elapsed_ms,
+ frame,
+ );
+ let is_in_speech = vad.in_speech;
+ if !was_in_speech && is_in_speech {
+ start_realtime_session_for_new_speech(
+ http,
+ config,
+ call_id,
+ trace_id,
+ vad,
+ read_attributes,
+ upload_slot,
+ last_fixture_sequence,
+ realtime_enabled,
+ );
+ }
+ (was_in_speech, is_in_speech, turn)
+}
+
+fn observe_bound_participant_frame<F>(
+ participant_identity: &str,
+ expected_participant: Option<&str>,
+ vad: &mut SimpleVad,
+ call_id: &str,
+ trace_id: &str,
+ participant_alias: &str,
+ track_sid_alias: &str,
+ frame_count: u64,
+ elapsed_ms: u64,
+ frame: &AudioFrame<'_>,
+ http: Client,
+ config: RealtimeAsrConfig,
+ read_attributes: F,
+ upload_slot: &mut Option<RealtimeAsrUpload>,
+ last_fixture_sequence: &mut Option<String>,
+ realtime_enabled: bool,
+) -> (bool, bool, Option<FinishedSpeechTurn>)
+where
+ F: FnOnce() -> std::collections::HashMap<String, String>,
+{
+ if !is_bound_user_participant(participant_identity, expected_participant) {
+ warn!(
+ "audioIngressOriginStatus" = "wrong_participant_or_track",
+ "runtime helper rejected audio participant before VAD/session"
+ );
+ return (vad.in_speech, vad.in_speech, None);
+ }
+ observe_frame_and_start_session(
+ vad,
+ call_id,
+ trace_id,
+ participant_alias,
+ track_sid_alias,
+ frame_count,
+ elapsed_ms,
+ frame,
+ http,
+ config,
+ read_attributes,
+ upload_slot,
+ last_fixture_sequence,
+ realtime_enabled,
+ )
+}
+
+fn start_realtime_session_for_new_speech(
+ http: Client,
+ config: RealtimeAsrConfig,
+ call_id: &str,
+ trace_id: &str,
+ vad: &SimpleVad,
+ read_attributes: impl FnOnce() -> std::collections::HashMap<String, String>,
+ upload_slot: &mut Option<RealtimeAsrUpload>,
+ last_fixture_sequence: &mut Option<String>,
+ realtime_enabled: bool,
+) {
+ let turn_id = format!("turn-{:04}", vad.turn_index);
+ let attributes = read_attributes();
+ let origin_status = AudioIngressMetadata::origin_status(&attributes);
+ let metadata = match AudioIngressMetadata::from_participant(&attributes) {
+ Ok(metadata) => metadata,
+ Err(reason) => {
+ warn!(call_id = %call_id, trace_id = %trace_id, turn_id = %turn_id,
+ reason, audioIngressOriginStatus = %AudioIngressMetadata::rejected_origin_status(
+ reason, &attributes
+ ),
+ "runtime helper asr_realtime_metadata_rejected");
+ return;
+ }
+ };
+ if let Some(metadata) = metadata.as_ref() {
+ if !fixture_sequence_is_new(
+ last_fixture_sequence.as_deref(),
+ &metadata.client_fixture_sequence,
+ ) {
+ warn!(call_id = %call_id, trace_id = %trace_id, turn_id = %turn_id,
+ audioIngressOriginStatus = "sequence_replayed_or_regressed",
+ "runtime helper asr_realtime_metadata_sequence_rejected");
+ return;
+ }
+ }
+ match RealtimeAsrUpload::start(
+ http,
+ config,
+ call_id,
+ trace_id,
+ &turn_id,
+ &vad.speech_samples,
+ metadata.as_ref(),
+ ) {
+ Ok(upload) => {
+ if let Some(metadata) = metadata {
+ *last_fixture_sequence = Some(metadata.client_fixture_sequence);
+ }
+ info!(call_id = %call_id, trace_id = %trace_id, turn_id = %turn_id,
+ origin_status, "runtime helper asr_realtime_session_started");
+ *upload_slot = Some(upload);
+ }
+ Err(error) if realtime_enabled => {
+ warn!(call_id = %call_id, trace_id = %trace_id, turn_id = %turn_id,
+ error = %safe_error(&error.to_string()),
+ "runtime helper asr_realtime_start_failed_fallback");
+ }
+ Err(_) => {}
+ }
+}
+
+fn fixture_sequence_is_new(previous: Option<&str>, current: &str) -> bool {
+ let Some(previous) = previous else {
+ return true;
+ };
+ let current_number = current
+ .rsplit_once('-')
+ .and_then(|(_, value)| value.parse::<u64>().ok());
+ let previous_number = previous
+ .rsplit_once('-')
+ .and_then(|(_, value)| value.parse::<u64>().ok());
+ match (previous_number, current_number) {
+ (Some(previous), Some(current)) => current > previous,
+ _ => previous != current,
+ }
}
struct DrainedUserAudioFrame {
@@ -3943,12 +4343,408 @@
#[cfg(test)]
mod tests {
- use super::{
- ReplyChunkMarker, ReplyChunkMarkerState, RuntimeTurnDeviceOutput, RuntimeTurnStreamEvent,
- RuntimeTurnStreamState, RuntimeTurnStreamTimingPhase, runtime_session_nonce_hash,
- should_publish_device_output,
+ use super::*;
+ use std::{
+ collections::HashSet,
+ io::{Read, Write},
+ net::TcpListener,
+ sync::{
+ Arc,
+ atomic::{AtomicUsize, Ordering},
+ mpsc,
+ },
+ thread,
+ time::Duration,
};
- use std::collections::HashSet;
+
+ #[test]
+ fn production_vad_session_boundary_reads_updated_attributes() {
+ let config = SimpleVadConfig {
+ rms_threshold: 0.001,
+ peak_threshold: 0.01,
+ start_frames: 2,
+ end_silence_ms: 100,
+ min_speech_ms: 1,
+ max_turn_ms: 1_000,
+ initial_ignore_ms: 0,
+ };
+ let mut vad = SimpleVad::new(config);
+ let samples = vec![1_000i16; 160];
+ let frame = AudioFrame {
+ data: samples.as_slice().into(),
+ sample_rate: 16_000,
+ num_channels: 1,
+ samples_per_channel: 160,
+ };
+ let mut attributes = std::collections::HashMap::from([
+ (
+ "inputSourceCategory".to_string(),
+ "controlled_fixture".to_string(),
+ ),
+ (
+ "clientFixtureSequence".to_string(),
+ "fixture-01".to_string(),
+ ),
+ ]);
+ let mut starts = Vec::new();
+ for (session_index, sequence) in [(1, "fixture-01"), (2, "fixture-02")] {
+ let was_in_speech = vad.in_speech;
+ vad.observe_frame(
+ "call-001",
+ "trace-001",
+ "participant",
+ "track",
+ session_index * 2 - 1,
+ 1_000 * session_index,
+ &frame,
+ );
+ vad.observe_frame(
+ "call-001",
+ "trace-001",
+ "participant",
+ "track",
+ session_index * 2,
+ 1_000 * session_index + 10,
+ &frame,
+ );
+ let is_in_speech = vad.in_speech;
+ assert!(!was_in_speech && is_in_speech);
+ attributes.insert("clientFixtureSequence".to_string(), sequence.to_string());
+ let metadata = AudioIngressMetadata::from_participant(&attributes)
+ .expect("valid participant attributes")
+ .expect("controlled fixture metadata");
+ let session_line = asr_realtime::session_start_line(
+ "call-001",
+ "trace-001",
+ &format!("turn-{session_index:04}"),
+ "nonce-001",
+ Some(&metadata),
+ )
+ .expect("session start line");
+ let session_json: serde_json::Value =
+ serde_json::from_slice(&session_line).expect("session start json");
+ assert_eq!(sequence, session_json["clientFixtureSequence"]);
+ starts.push(metadata.client_fixture_sequence);
+ vad.reset_current_turn();
+ }
+ assert_eq!(vec!["fixture-01", "fixture-02"], starts);
+ attributes.insert("inputSourceCategory".to_string(), "other".to_string());
+ assert!(AudioIngressMetadata::from_participant(&attributes).is_err());
+ assert!(
+ AudioIngressMetadata::from_participant(&std::collections::HashMap::new())
+ .expect("missing attributes is absent")
+ .is_none()
+ );
+ attributes.insert(
+ "clientFixtureSequence".to_string(),
+ "fixture-01".to_string(),
+ );
+ assert!(AudioIngressMetadata::from_participant(&attributes).is_err());
+ }
+
+ #[test]
+ fn production_observer_rejects_wrong_participant_before_vad_session() {
+ assert!(!is_bound_user_participant(
+ "participant-other",
+ Some("participant-user")
+ ));
+ assert!(is_bound_user_participant(
+ "participant-user",
+ Some("participant-user")
+ ));
+ assert!(is_bound_user_participant("participant-any", None));
+ }
+
+ #[tokio::test]
+ async fn production_observer_vad_to_session_entry_reads_each_updated_attribute() {
+ let listener = TcpListener::bind("127.0.0.1:0").expect("bind local ASR fixture");
+ let address = listener.local_addr().expect("fixture address");
+ let (request_tx, request_rx) = mpsc::channel::<String>();
+ let captured_count = Arc::new(AtomicUsize::new(0));
+ let captured_count_for_server = Arc::clone(&captured_count);
+ let server = thread::spawn(move || {
+ for _ in 0..2 {
+ let (mut stream, _) = listener.accept().expect("accept ASR session");
+ stream
+ .set_read_timeout(Some(Duration::from_secs(2)))
+ .expect("set fixture timeout");
+ let mut bytes = Vec::new();
+ let mut buffer = [0_u8; 4096];
+ loop {
+ match stream.read(&mut buffer) {
+ Ok(0) => break,
+ Ok(size) => {
+ bytes.extend_from_slice(&buffer[..size]);
+ if bytes.windows(7).any(|window| window == b"0\r\n\r\n") {
+ break;
+ }
+ }
+ Err(_) => break,
+ }
+ }
+ request_tx
+ .send(String::from_utf8_lossy(&bytes).into_owned())
+ .expect("capture ASR request");
+ captured_count_for_server.fetch_add(1, Ordering::SeqCst);
+ stream
+ .write_all(b"HTTP/1.1 200 OK\r\ncontent-type: application/json\r\ncontent-length: 39\r\nconnection: close\r\n\r\n{\"code\":0,\"data\":{\"status\":\"ok\"}}")
+ .expect("write fixture response");
+ }
+ });
+ let mut vad = SimpleVad::new(SimpleVadConfig {
+ rms_threshold: 0.001,
+ peak_threshold: 0.01,
+ start_frames: 1,
+ end_silence_ms: 100,
+ min_speech_ms: 1,
+ max_turn_ms: 1_000,
+ initial_ignore_ms: 0,
+ });
+ let frame_data = vec![1_000i16; 160];
+ let frame = AudioFrame {
+ data: frame_data.as_slice().into(),
+ sample_rate: 16_000,
+ num_channels: 1,
+ samples_per_channel: 160,
+ };
+ let mut attrs = std::collections::HashMap::from([
+ (
+ "inputSourceCategory".to_string(),
+ "controlled_fixture".to_string(),
+ ),
+ (
+ "clientFixtureSequence".to_string(),
+ "fixture-01".to_string(),
+ ),
+ ]);
+ let mut upload = None;
+ let mut last_fixture_sequence = None;
+ let config = RealtimeAsrConfig {
+ enabled: true,
+ url: Some(format!("http://{address}/runtime/asr/realtime")),
+ runtime_token: Some("test".to_string()),
+ runtime_session_nonce: Some("test".to_string()),
+ chunk_duration_ms: 200,
+ };
+ let (was, is, turn) = observe_bound_participant_frame(
+ "participant-user",
+ Some("participant-user"),
+ &mut vad,
+ "call-001",
+ "trace-001",
+ "participant-user",
+ "track-001",
+ 1,
+ 1_000,
+ &frame,
+ Client::new(),
+ config.clone(),
+ || attrs.clone(),
+ &mut upload,
+ &mut last_fixture_sequence,
+ true,
+ );
+ assert!(!was && is && turn.is_none());
+ assert!(upload.is_some());
+ upload.take().unwrap().cancel("test").await;
+ vad.reset_current_turn();
+ attrs.insert(
+ "clientFixtureSequence".to_string(),
+ "fixture-02".to_string(),
+ );
+ let (was, is, turn) = observe_bound_participant_frame(
+ "participant-user",
+ Some("participant-user"),
+ &mut vad,
+ "call-001",
+ "trace-001",
+ "participant-user",
+ "track-001",
+ 2,
+ 2_000,
+ &frame,
+ Client::new(),
+ config,
+ || attrs.clone(),
+ &mut upload,
+ &mut last_fixture_sequence,
+ true,
+ );
+ assert!(!was && is && turn.is_none());
+ assert!(upload.is_some());
+ upload.take().unwrap().cancel("test").await;
+
+ let first_request = request_rx
+ .recv_timeout(Duration::from_secs(2))
+ .expect("first session request");
+ let second_request = request_rx
+ .recv_timeout(Duration::from_secs(2))
+ .expect("second session request");
+ assert!(first_request.contains("\"clientFixtureSequence\":\"fixture-01\""));
+ assert!(second_request.contains("\"clientFixtureSequence\":\"fixture-02\""));
+ assert!(
+ first_request.contains("\"audioIngressOriginStatus\":\"controlled_fixture_bound\"")
+ );
+ assert!(
+ second_request.contains("\"audioIngressOriginStatus\":\"controlled_fixture_bound\"")
+ );
+ // The same production boundary rejects a wrong participant before VAD/session creation.
+ assert!(!is_bound_user_participant(
+ "participant-other",
+ Some("participant-user")
+ ));
+ assert_eq!(0, request_rx.try_iter().count());
+ vad.reset_current_turn();
+ attrs.insert(
+ "clientFixtureSequence".to_string(),
+ "fixture-03".to_string(),
+ );
+ let (_, is_wrong, wrong_turn) = observe_bound_participant_frame(
+ "participant-other",
+ Some("participant-user"),
+ &mut vad,
+ "call-001",
+ "trace-001",
+ "participant-user",
+ "track-001",
+ 3,
+ 3_000,
+ &frame,
+ Client::new(),
+ RealtimeAsrConfig {
+ enabled: true,
+ url: Some(format!("http://{address}/runtime/asr/realtime")),
+ runtime_token: Some("test".to_string()),
+ runtime_session_nonce: Some("test".to_string()),
+ chunk_duration_ms: 200,
+ },
+ || attrs.clone(),
+ &mut upload,
+ &mut last_fixture_sequence,
+ true,
+ );
+ assert!(!is_wrong && wrong_turn.is_none() && upload.is_none());
+ vad.reset_current_turn();
+ attrs.insert(
+ "clientFixtureSequence".to_string(),
+ "fixture-01".to_string(),
+ );
+ let (_, _, _) = observe_bound_participant_frame(
+ "participant-user",
+ Some("participant-user"),
+ &mut vad,
+ "call-001",
+ "trace-001",
+ "participant-user",
+ "track-001",
+ 4,
+ 4_000,
+ &frame,
+ Client::new(),
+ RealtimeAsrConfig {
+ enabled: true,
+ url: Some(format!("http://{address}/runtime/asr/realtime")),
+ runtime_token: Some("test".to_string()),
+ runtime_session_nonce: Some("test".to_string()),
+ chunk_duration_ms: 200,
+ },
+ || attrs.clone(),
+ &mut upload,
+ &mut last_fixture_sequence,
+ true,
+ );
+ assert!(upload.is_none());
+ vad.reset_current_turn();
+ let (_, _, _) = observe_bound_participant_frame(
+ "participant-user",
+ Some("participant-user"),
+ &mut vad,
+ "call-001",
+ "trace-001",
+ "participant-user",
+ "track-001",
+ 5,
+ 5_000,
+ &frame,
+ Client::new(),
+ RealtimeAsrConfig {
+ enabled: true,
+ url: Some(format!("http://{address}/runtime/asr/realtime")),
+ runtime_token: Some("test".to_string()),
+ runtime_session_nonce: Some("test".to_string()),
+ chunk_duration_ms: 200,
+ },
+ || attrs.clone(),
+ &mut upload,
+ &mut last_fixture_sequence,
+ true,
+ );
+ assert!(upload.is_none());
+ vad.reset_current_turn();
+ attrs.remove("clientFixtureSequence");
+ let (_, _, missing_sequence_turn) = observe_bound_participant_frame(
+ "participant-user",
+ Some("participant-user"),
+ &mut vad,
+ "call-001",
+ "trace-001",
+ "participant-user",
+ "track-001",
+ 6,
+ 6_000,
+ &frame,
+ Client::new(),
+ RealtimeAsrConfig {
+ enabled: true,
+ url: Some(format!("http://{address}/runtime/asr/realtime")),
+ runtime_token: Some("test".to_string()),
+ runtime_session_nonce: Some("test".to_string()),
+ chunk_duration_ms: 200,
+ },
+ || attrs.clone(),
+ &mut upload,
+ &mut last_fixture_sequence,
+ true,
+ );
+ assert!(missing_sequence_turn.is_none() && upload.is_none());
+ assert_eq!(
+ "sequence_absent",
+ AudioIngressMetadata::rejected_origin_status("incomplete_metadata", &attrs)
+ );
+ assert_eq!(2, captured_count.load(Ordering::SeqCst));
+ vad.reset_current_turn();
+ attrs.insert("inputSourceCategory".to_string(), "other".to_string());
+ let mut invalid_upload = None;
+ let (_, _, invalid_turn) = observe_bound_participant_frame(
+ "participant-user",
+ Some("participant-user"),
+ &mut vad,
+ "call-001",
+ "trace-001",
+ "participant-other",
+ "track-001",
+ 3,
+ 3_000,
+ &frame,
+ Client::new(),
+ RealtimeAsrConfig {
+ enabled: true,
+ url: Some("http://127.0.0.1:9".to_string()),
+ runtime_token: Some("test".to_string()),
+ runtime_session_nonce: Some("test".to_string()),
+ chunk_duration_ms: 200,
+ },
+ || attrs,
+ &mut invalid_upload,
+ &mut last_fixture_sequence,
+ true,
+ );
+ assert!(invalid_turn.is_none());
+ assert!(invalid_upload.is_none());
+ assert_eq!(2, captured_count.load(Ordering::SeqCst));
+ assert_eq!(0, request_rx.try_iter().count());
+ server.join().expect("fixture server");
+ }
#[test]
fn reply_chunk_marker_state_emits_turn_first_once_and_later_segment_first_once() {
@@ -4078,4 +4874,150 @@
assert!(published.is_empty());
}
}
+
+ #[test]
+ fn controlled_fixture_probe_requires_bound_hashes_and_protocol() {
+ let call_id = "call-ack-1";
+ let trace_id = "trace-ack-1";
+ let payload = serde_json::to_vec(&json!({
+ "type": CONTROLLED_FIXTURE_PROBE_TOPIC,
+ "protocolVersion": CONTROLLED_FIXTURE_PROTOCOL_VERSION,
+ "callIdHash": sha256_hex(call_id),
+ "callTraceIdHash": sha256_hex(trace_id),
+ "generation": CONTROLLED_FIXTURE_GENERATION,
+ "clientFixtureSequence": "fixture-01"
+ }))
+ .unwrap();
+ let sender = ParticipantIdentity("user-1".to_string());
+ let pending =
+ controlled_fixture_probe(&payload, call_id, trace_id, &sender, Some("user-1"))
+ .expect("valid probe");
+ assert_eq!(pending.sequence, "fixture-01");
+ assert!(
+ controlled_fixture_probe(&payload, call_id, "other-trace", &sender, Some("user-1"),)
+ .is_none()
+ );
+ assert!(
+ controlled_fixture_probe(
+ &payload,
+ call_id,
+ trace_id,
+ &ParticipantIdentity("other-user".to_string()),
+ Some("user-1"),
+ )
+ .is_none()
+ );
+ }
+
+ #[test]
+ fn controlled_fixture_attributes_ack_only_on_exact_current_sequence() {
+ let mut attributes = std::collections::HashMap::from([
+ (
+ "inputSourceCategory".to_string(),
+ "controlled_fixture".to_string(),
+ ),
+ (
+ "clientFixtureSequence".to_string(),
+ "fixture-01".to_string(),
+ ),
+ ]);
+ assert!(
+ classify_controlled_fixture_attributes(
+ "user-1",
+ Some("user-1"),
+ &attributes,
+ "fixture-01"
+ )
+ .is_ok()
+ );
+ attributes.insert(
+ "clientFixtureSequence".to_string(),
+ "fixture-02".to_string(),
+ );
+ assert_eq!(
+ classify_controlled_fixture_attributes(
+ "user-1",
+ Some("user-1"),
+ &attributes,
+ "fixture-01"
+ ),
+ Err("wrong_sequence")
+ );
+ attributes.remove("clientFixtureSequence");
+ assert_eq!(
+ classify_controlled_fixture_attributes(
+ "user-1",
+ Some("user-1"),
+ &attributes,
+ "fixture-01"
+ ),
+ Err("missing_sequence")
+ );
+ assert_eq!(
+ classify_controlled_fixture_attributes(
+ "other-user",
+ Some("user-1"),
+ &attributes,
+ "fixture-01"
+ ),
+ Err("wrong_participant")
+ );
+ }
+
+ #[test]
+ fn controlled_fixture_ack_payload_is_reliable_and_redacted() {
+ let ack = ControlledFixtureAttributeAck {
+ message_type: CONTROLLED_FIXTURE_ACK_TOPIC,
+ protocol_version: CONTROLLED_FIXTURE_PROTOCOL_VERSION,
+ call_id_hash: sha256_hex("call-1"),
+ call_trace_id_hash: sha256_hex("trace-1"),
+ generation: CONTROLLED_FIXTURE_GENERATION,
+ client_fixture_sequence: "fixture-01".to_string(),
+ result: "observed",
+ input_source_category: Some("controlled_fixture"),
+ reject_reason: None,
+ };
+ let encoded = serde_json::to_vec(&ack).unwrap();
+ let decoded: serde_json::Value = serde_json::from_slice(&encoded).unwrap();
+ assert_eq!(decoded["type"], CONTROLLED_FIXTURE_ACK_TOPIC);
+ assert_eq!(
+ decoded["protocolVersion"],
+ CONTROLLED_FIXTURE_PROTOCOL_VERSION
+ );
+ assert_eq!(decoded["result"], "observed");
+ assert!(decoded.get("callId").is_none());
+ assert!(decoded.get("traceId").is_none());
+ assert!(decoded.get("participantIdentity").is_none());
+ assert!(decoded.get("audio").is_none());
+ }
+
+ #[test]
+ fn controlled_fixture_probe_does_not_create_session_or_audio_side_effects() {
+ let attributes = std::collections::HashMap::new();
+ assert_eq!(
+ classify_controlled_fixture_attributes(
+ "user-1",
+ Some("user-1"),
+ &attributes,
+ "fixture-01"
+ ),
+ Err("missing_attributes")
+ );
+ assert_eq!(
+ CONTROLLED_FIXTURE_PROBE_TOPIC,
+ "controlled_fixture_attribute_probe"
+ );
+ assert_eq!(
+ CONTROLLED_FIXTURE_ACK_TOPIC,
+ "controlled_fixture_attribute_ack"
+ );
+ }
+
+ #[test]
+ fn controlled_fixture_probe_must_be_observed_before_audio_observer() {
+ assert!(controlled_fixture_observer_gate(false, None));
+ assert!(controlled_fixture_observer_gate(true, Some(true)));
+ assert!(!controlled_fixture_observer_gate(true, Some(false)));
+ assert!(!controlled_fixture_observer_gate(true, None));
+ }
}
--
Gitblit v1.9.3