Example: Placing an outbound call

1
require(Modules.Grok);
2
require(Modules.ApplicationStorage);
3
4
const SYSTEM_PROMPT = `
5
  Your name is Voxi. You are a helpful voice assistant for phone callers representing the company Voximplant (pronounced VOX-im-plant).
6
  Keep responses short and telephony-friendly (usually 1-2 sentences).
7
  If the user asks for a live agent or an operator, call the "forward_to_agent" function.
8
  If the user says goodbye, call the "hangup_call" function.
9
`;
10
11
// -------------------- Grok Voice Agent settings --------------------
12
const SESSION_PARAMETERS = {
13
    session: {
14
        voice: "Ara",
15
        turn_detection: {type: "server_vad"},
16
        instructions: SYSTEM_PROMPT,
17
        tools: [
18
            {type: "web_search"},
19
            {
20
                type: "file_search",
21
                vector_store_ids: ["collection_4c5a63ab-f739-4c13-93d2-05b74095c34a"], // optional RAG
22
                max_num_results: 5,
23
            },
24
            {
25
                type: "x_search",
26
                allowed_x_handles: ["voximplant", "aylarov"],
27
            },
28
            {
29
                type: "function",
30
                name: "forward_to_agent",
31
                description: "Forward the user to a live agent",
32
                parameters: {
33
                    type: "object",
34
                    properties: {},
35
                    required: [],
36
                },
37
            },
38
            {
39
                type: "function",
40
                name: "hangup_call",
41
                description: "Hangup the call",
42
                parameters: {
43
                    type: "object",
44
                    properties: {},
45
                    required: [],
46
                },
47
            },
48
        ],
49
    },
50
};
51
52
// Helper to parse custom data JSON from call setup
53
function parseCustomData() {
54
    const raw = VoxEngine.customData();
55
    if (!raw) return {};
56
    try {
57
        return JSON.parse(raw);
58
    } catch (error) {
59
        Logger.write("===CUSTOM_DATA_PARSE_FAILED===");
60
        Logger.write(error);
61
        return {};
62
    }
63
}
64
65
VoxEngine.addEventListener(AppEvents.Started, async () => {
66
    let voiceAIClient;
67
    let call;
68
    let hangupCall = false;
69
    let forwardToLiveAgent = false;
70
71
    const {destination, callerId} = parseCustomData();
72
73
    if (!destination || !callerId) {
74
        Logger.write(
75
            "Missing destination or callerId. Provide custom data: {\"destination\":\"+16174018889\",\"callerId\":\"+13322776633\"}",
76
        );
77
        VoxEngine.terminate();
78
        return;
79
    }
80
81
    call = VoxEngine.callPSTN(destination, callerId);
82
83
    // Termination functions - add cleanup and logging as needed
84
    call.addEventListener(CallEvents.Disconnected, ()=>VoxEngine.terminate());
85
    call.addEventListener(CallEvents.Failed, ()=>VoxEngine.terminate());
86
87
    try {
88
        voiceAIClient = await Grok.createVoiceAgentAPIClient({
89
            xAIApiKey: (await ApplicationStorage.get("XAI_API_KEY")).value,
90
            onWebSocketClose: (event) => {
91
                Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
92
                VoxEngine.terminate();
93
            },
94
        });
95
96
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.ConversationCreated, (event) => {
97
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
98
            voiceAIClient.sessionUpdate(SESSION_PARAMETERS);
99
        });
100
101
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.SessionUpdated, (event) => {
102
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
103
            voiceAIClient.responseCreate({instructions: "Hello."});
104
        });
105
106
        // Keep it interruption-friendly (barge-in).
107
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.InputAudioBufferSpeechStarted, (event) => {
108
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
109
            voiceAIClient.clearMediaBuffer();
110
        });
111
112
        // Function calling.
113
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.ResponseFunctionCallArgumentsDone, (event) => {
114
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
115
116
            const {name, call_id} = event?.data?.payload || {};
117
            let output;
118
119
            if (name !== "forward_to_agent" && name !== "hangup_call") {
120
                Logger.write(`===Ignoring unhandled function call: ${name}===`);
121
                return;
122
            }
123
124
            if (name === "forward_to_agent") {
125
                forwardToLiveAgent = true;
126
                output = {result: "Forwarding your call to a live agent. Please hold on."};
127
            } else if (name === "hangup_call") {
128
                hangupCall = true;
129
                output = {result: "Have a great day, goodbye!"};
130
            }
131
132
            voiceAIClient.conversationItemCreate({
133
                item: {
134
                    type: "function_call_output",
135
                    call_id,
136
                    output: JSON.stringify(output),
137
                },
138
            });
139
            voiceAIClient.responseCreate({});
140
        });
141
142
        // -------------------- Log Other Events --------------------
143
        [
144
            CallEvents.FirstAudioPacketReceived,
145
            Grok.Events.WebSocketMediaStarted,
146
            Grok.VoiceAgentAPIEvents.InputAudioBufferSpeechStopped,
147
            Grok.VoiceAgentAPIEvents.ConversationItemInputAudioTranscriptionCompleted,
148
            Grok.VoiceAgentAPIEvents.ConversationItemAdded,
149
            Grok.VoiceAgentAPIEvents.ResponseCreated,
150
            Grok.VoiceAgentAPIEvents.ResponseOutputItemAdded,
151
            Grok.VoiceAgentAPIEvents.ResponseDone,
152
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioTranscriptDelta,
153
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioTranscriptDone,
154
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioDelta,
155
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioDone,
156
            Grok.VoiceAgentAPIEvents.ResponseOutputItemDone,
157
            Grok.VoiceAgentAPIEvents.ConnectorInformation,
158
            Grok.VoiceAgentAPIEvents.InputAudioBufferCommitted,
159
            Grok.VoiceAgentAPIEvents.WebSocketError,
160
            Grok.VoiceAgentAPIEvents.Unknown,
161
        ].forEach((evtName) => {
162
            voiceAIClient.addEventListener(evtName, (e) => {
163
                Logger.write(`===${e.name}===>${JSON.stringify(e)}`);
164
            });
165
        });
166
167
        voiceAIClient.addEventListener(Grok.Events.WebSocketMediaStarted, () => {
168
            VoxEngine.sendMediaBetween(call, voiceAIClient);
169
        });
170
171
        voiceAIClient.addEventListener(Grok.Events.WebSocketMediaEnded, (event) => {
172
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
173
            if (hangupCall) callCloseHandler();
174
            else if (forwardToLiveAgent) {
175
                call.say("Here is where I would forward the call via the phone network or SIP.");
176
                call.addEventListener(CallEvents.PlaybackFinished, callCloseHandler);
177
            } else return;
178
        });
179
    } catch (error) {
180
        Logger.write("===SOMETHING_WENT_WRONG===");
181
        Logger.write(error);
182
        VoxEngine.terminate();
183
    }
184
});