Example: Answering an incoming call

1
voiceAIClient.sessionUpdate({
2
  session: {
3
    type: "realtime",
4
    instructions: SYSTEM_PROMPT,
5
    voice: "alloy",
6
    output_modalities: ["audio"],
7
    turn_detection: { type: "server_vad", interrupt_response: true },
8
  },
9
});

1
VoxEngine.sendMediaBetween(call, voiceAIClient);

1
voiceAIClient.addEventListener(OpenAI.RealtimeAPIEvents.InputAudioBufferSpeechStarted, () => {
2
  voiceAIClient.clearMediaBuffer();
3
});

1
/**
2
 * Voximplant + OpenAI Realtime API connector demo
3
 * Scenario: answer an incoming call and bridge it to OpenAI Realtime.
4
 */
5
6
require(Modules.OpenAI);
7
require(Modules.ApplicationStorage);
8
9
const SYSTEM_PROMPT = `
10
You are Voxi, a helpful voice assistant for phone callers.
11
Keep responses short and telephony-friendly (usually 1-2 sentences).
12
`;
13
14
const SESSION_CONFIG = {
15
    session: {
16
        type: "realtime",
17
        instructions: SYSTEM_PROMPT,
18
        voice: "alloy",
19
        output_modalities: ["audio"],
20
        turn_detection: {type: "server_vad", interrupt_response: true},
21
    },
22
};
23
24
VoxEngine.addEventListener(AppEvents.CallAlerting, async ({call}) => {
25
    let voiceAIClient;
26
27
    call.addEventListener(CallEvents.Disconnected, () => VoxEngine.terminate());
28
    call.addEventListener(CallEvents.Failed, () => VoxEngine.terminate());
29
30
    try {
31
        call.answer();
32
        // call.record({hd_audio: true, stereo: true}); // Optional: record the call
33
34
        voiceAIClient = await OpenAI.createRealtimeAPIClient({
35
            apiKey: (await ApplicationStorage.get("OPENAI_API_KEY")).value,
36
            model: "gpt-realtime-1.5",
37
            onWebSocketClose: (event) => {
38
                Logger.write("===OpenAI.WebSocket.Close===");
39
                if (event) Logger.write(JSON.stringify(event));
40
                VoxEngine.terminate();
41
            },
42
        });
43
44
        //---------------------- Event handlers -----------------------
45
        voiceAIClient.addEventListener(OpenAI.RealtimeAPIEvents.SessionCreated, () => {
46
            voiceAIClient.sessionUpdate(SESSION_CONFIG);
47
        });
48
49
        voiceAIClient.addEventListener(OpenAI.RealtimeAPIEvents.SessionUpdated, () => {
50
            VoxEngine.sendMediaBetween(call, voiceAIClient);
51
            voiceAIClient.responseCreate({instructions: "Hello! How can I help today?"});
52
        });
53
54
        // Barge-in: clear buffered audio when the caller starts speaking
55
        voiceAIClient.addEventListener(OpenAI.RealtimeAPIEvents.InputAudioBufferSpeechStarted, () => {
56
            Logger.write("===BARGE-IN: OpenAI.InputAudioBufferSpeechStarted===");
57
            voiceAIClient.clearMediaBuffer();
58
        });
59
60
        voiceAIClient.addEventListener(
61
            OpenAI.RealtimeAPIEvents.ConversationItemInputAudioTranscriptionCompleted,
62
            (event) => {
63
                const payload = event?.data?.payload || event?.data || {};
64
                const transcript = payload.transcript || payload.text || payload.delta;
65
                if (transcript) Logger.write(`===USER=== ${transcript}`);
66
            }
67
        );
68
69
        voiceAIClient.addEventListener(
70
            OpenAI.RealtimeAPIEvents.ResponseOutputAudioTranscriptDelta,
71
            (event) => {
72
                const payload = event?.data?.payload || event?.data || {};
73
                if (payload.delta) Logger.write(`===AGENT_DELTA=== ${payload.delta}`);
74
            }
75
        );
76
77
        voiceAIClient.addEventListener(
78
            OpenAI.RealtimeAPIEvents.ResponseOutputAudioTranscriptDone,
79
            (event) => {
80
                const payload = event?.data?.payload || event?.data || {};
81
                const transcript = payload.transcript || payload.text;
82
                if (transcript) Logger.write(`===AGENT=== ${transcript}`);
83
            }
84
        );
85
86
        // Consolidated "log-only" handlers - key OpenAI/VoxEngine debugging events
87
        [
88
            OpenAI.RealtimeAPIEvents.ResponseCreated,
89
            OpenAI.RealtimeAPIEvents.ResponseDone,
90
            OpenAI.RealtimeAPIEvents.ResponseOutputAudioDone,
91
            OpenAI.RealtimeAPIEvents.ConnectorInformation,
92
            OpenAI.RealtimeAPIEvents.HTTPResponse,
93
            OpenAI.RealtimeAPIEvents.WebSocketError,
94
            OpenAI.RealtimeAPIEvents.Unknown,
95
            OpenAI.Events.WebSocketMediaStarted,
96
            OpenAI.Events.WebSocketMediaEnded,
97
        ].forEach((eventName) => {
98
            voiceAIClient.addEventListener(eventName, (event) => {
99
                Logger.write(`===${event.name}===`);
100
                if (event?.data) Logger.write(JSON.stringify(event.data));
101
            });
102
        });
103
    } catch (error) {
104
        Logger.write("===UNHANDLED_ERROR===");
105
        Logger.write(error);
106
        voiceAIClient?.close();
107
        VoxEngine.terminate();
108
    }
109
});