feat(mixer): enhance mixer functionality with interaction and tool legs

- Updated mixer to handle participant and isolated leg roles, allowing for IVR and consent interactions.
- Introduced commands for starting and canceling interactions, managing tool legs for recording and transcription.
- Implemented per-source audio handling for tool legs, enabling separate audio processing.
- Enhanced DTMF handling to forward events between participant legs only.
- Added support for PCM recording directly from tool legs, with WAV file generation.
- Updated TypeScript definitions and functions to support new interaction and tool leg features.
This commit is contained in:
2026-04-10 14:54:21 +00:00
parent 6a130db7c7
commit 7d59361352
13 changed files with 1448 additions and 94 deletions

View File

@@ -96,6 +96,16 @@ interface IDeviceStatus {
isBrowser: boolean;
}
interface IActiveLeg {
id: string;
type: 'sip-device' | 'sip-provider' | 'webrtc' | 'tool';
state: string;
codec: string | null;
rtpPort: number | null;
remoteMedia: string | null;
metadata: Record<string, unknown>;
}
interface IActiveCall {
id: string;
direction: string;
@@ -104,6 +114,13 @@ interface IActiveCall {
providerUsed: string | null;
state: string;
startedAt: number;
legs: Map<string, IActiveLeg>;
}
interface IHistoryLeg {
id: string;
type: string;
metadata: Record<string, unknown>;
}
interface ICallHistoryEntry {
@@ -113,6 +130,7 @@ interface ICallHistoryEntry {
calleeNumber: string | null;
startedAt: number;
duration: number;
legs: IHistoryLeg[];
}
const providerStatuses = new Map<string, IProviderStatus>();
@@ -187,7 +205,18 @@ function getStatus() {
calls: [...activeCalls.values()].map((c) => ({
...c,
duration: Math.floor((Date.now() - c.startedAt) / 1000),
legs: [],
legs: [...c.legs.values()].map((l) => ({
id: l.id,
type: l.type,
state: l.state,
codec: l.codec,
rtpPort: l.rtpPort,
remoteMedia: l.remoteMedia,
metadata: l.metadata || {},
pktSent: 0,
pktReceived: 0,
transcoding: false,
})),
})),
callHistory,
contacts: appConfig.contacts || [],
@@ -242,6 +271,7 @@ async function startProxyEngine(): Promise<void> {
providerUsed: data.provider_id,
state: 'ringing',
startedAt: Date.now(),
legs: new Map(),
});
// Notify browsers of incoming call.
@@ -266,6 +296,7 @@ async function startProxyEngine(): Promise<void> {
providerUsed: null,
state: 'setting-up',
startedAt: Date.now(),
legs: new Map(),
});
});
@@ -279,6 +310,7 @@ async function startProxyEngine(): Promise<void> {
providerUsed: data.provider_id,
state: 'setting-up',
startedAt: Date.now(),
legs: new Map(),
});
// Notify all browser devices — they can connect via WebRTC to listen/talk.
@@ -303,6 +335,20 @@ async function startProxyEngine(): Promise<void> {
if (call) {
call.state = 'connected';
log(`[call] ${data.call_id} connected`);
// Enrich provider leg with media info from the answered event.
if (data.provider_media_addr && data.provider_media_port) {
for (const leg of call.legs.values()) {
if (leg.type === 'sip-provider') {
leg.remoteMedia = `${data.provider_media_addr}:${data.provider_media_port}`;
if (data.sip_pt !== undefined) {
const codecNames: Record<number, string> = { 0: 'PCMU', 8: 'PCMA', 9: 'G.722', 111: 'Opus' };
leg.codec = codecNames[data.sip_pt] || `PT${data.sip_pt}`;
}
break;
}
}
}
}
// Try to link WebRTC session to this call for audio bridging.
@@ -331,6 +377,15 @@ async function startProxyEngine(): Promise<void> {
const call = activeCalls.get(data.call_id);
if (call) {
log(`[call] ${data.call_id} ended: ${data.reason} (${data.duration}s)`);
// Snapshot legs with metadata for history.
const historyLegs: IHistoryLeg[] = [];
for (const [, leg] of call.legs) {
historyLegs.push({
id: leg.id,
type: leg.type,
metadata: leg.metadata || {},
});
}
// Move to history.
callHistory.unshift({
id: call.id,
@@ -339,6 +394,7 @@ async function startProxyEngine(): Promise<void> {
calleeNumber: call.calleeNumber,
startedAt: call.startedAt,
duration: data.duration,
legs: historyLegs,
});
if (callHistory.length > MAX_HISTORY) callHistory.pop();
activeCalls.delete(data.call_id);
@@ -361,17 +417,50 @@ async function startProxyEngine(): Promise<void> {
log(`[sip] unhandled ${data.method_or_status} Call-ID=${data.call_id?.slice(0, 20)} from=${data.from_addr}:${data.from_port}`);
});
// Leg events (multiparty).
// Leg events (multiparty) — update shadow state so the dashboard shows legs.
onProxyEvent('leg_added', (data: any) => {
log(`[leg] added: call=${data.call_id} leg=${data.leg_id} kind=${data.kind} state=${data.state}`);
const call = activeCalls.get(data.call_id);
if (call) {
call.legs.set(data.leg_id, {
id: data.leg_id,
type: data.kind,
state: data.state,
codec: null,
rtpPort: null,
remoteMedia: null,
metadata: data.metadata || {},
});
}
});
onProxyEvent('leg_removed', (data: any) => {
log(`[leg] removed: call=${data.call_id} leg=${data.leg_id}`);
activeCalls.get(data.call_id)?.legs.delete(data.leg_id);
});
onProxyEvent('leg_state_changed', (data: any) => {
log(`[leg] state: call=${data.call_id} leg=${data.leg_id}${data.state}`);
const call = activeCalls.get(data.call_id);
if (!call) return;
const leg = call.legs.get(data.leg_id);
if (leg) {
leg.state = data.state;
if (data.metadata) leg.metadata = data.metadata;
} else {
// Initial legs (provider/device) don't emit leg_added — create on first state change.
const legId: string = data.leg_id;
const type = legId.includes('-prov') ? 'sip-provider' : legId.includes('-dev') ? 'sip-device' : 'webrtc';
call.legs.set(data.leg_id, {
id: data.leg_id,
type,
state: data.state,
codec: null,
rtpPort: null,
remoteMedia: null,
metadata: data.metadata || {},
});
}
});
// WebRTC events from Rust — forward ICE candidates to browser via WebSocket.
@@ -484,6 +573,7 @@ initWebUi(
providerUsed: providerId || null,
state: 'setting-up',
startedAt: Date.now(),
legs: new Map(),
});
} else {
log(`[dashboard] call failed for ${number}`);