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
    // Alternative outbound paths (uncomment to use):
83
    // call = VoxEngine.callUser({username: destination, callerid: callerId});
84
    // call = VoxEngine.callSIP(`sip:${destination}@your-sip-domain`, callerId);
85
    // call = VoxEngine.callWhatsappUser({number: destination, callerid: callerId});
86
87
    // Termination functions - add cleanup and logging as needed
88
    call.addEventListener(CallEvents.Disconnected, ()=>VoxEngine.terminate());
89
    call.addEventListener(CallEvents.Failed, ()=>VoxEngine.terminate());
90
91
    try {
92
        voiceAIClient = await Grok.createVoiceAgentAPIClient({
93
            xAIApiKey: (await ApplicationStorage.get("XAI_API_KEY")).value,
94
            onWebSocketClose: (event) => {
95
                Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
96
                VoxEngine.terminate();
97
            },
98
        });
99
100
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.ConversationCreated, (event) => {
101
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
102
            voiceAIClient.sessionUpdate(SESSION_PARAMETERS);
103
        });
104
105
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.SessionUpdated, (event) => {
106
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
107
            voiceAIClient.responseCreate({instructions: "Hello."});
108
        });
109
110
        // Keep it interruption-friendly (barge-in).
111
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.InputAudioBufferSpeechStarted, (event) => {
112
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
113
            voiceAIClient.clearMediaBuffer();
114
        });
115
116
        // Function calling.
117
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.ResponseFunctionCallArgumentsDone, (event) => {
118
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
119
120
            const {name, call_id} = event?.data?.payload || {};
121
            let output;
122
123
            if (name !== "forward_to_agent" && name !== "hangup_call") {
124
                Logger.write(`===Ignoring unhandled function call: ${name}===`);
125
                return;
126
            }
127
128
            if (name === "forward_to_agent") {
129
                forwardToLiveAgent = true;
130
                output = {result: "Forwarding your call to a live agent. Please hold on."};
131
            } else if (name === "hangup_call") {
132
                hangupCall = true;
133
                output = {result: "Have a great day, goodbye!"};
134
            }
135
136
            voiceAIClient.conversationItemCreate({
137
                item: {
138
                    type: "function_call_output",
139
                    call_id,
140
                    output: JSON.stringify(output),
141
                },
142
            });
143
            voiceAIClient.responseCreate({});
144
        });
145
146
        // -------------------- Log Other Events --------------------
147
        [
148
            CallEvents.FirstAudioPacketReceived,
149
            Grok.Events.WebSocketMediaStarted,
150
            Grok.VoiceAgentAPIEvents.InputAudioBufferSpeechStopped,
151
            Grok.VoiceAgentAPIEvents.ConversationItemInputAudioTranscriptionCompleted,
152
            Grok.VoiceAgentAPIEvents.ConversationItemAdded,
153
            Grok.VoiceAgentAPIEvents.ResponseCreated,
154
            Grok.VoiceAgentAPIEvents.ResponseOutputItemAdded,
155
            Grok.VoiceAgentAPIEvents.ResponseDone,
156
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioTranscriptDelta,
157
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioTranscriptDone,
158
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioDelta,
159
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioDone,
160
            Grok.VoiceAgentAPIEvents.ResponseOutputItemDone,
161
            Grok.VoiceAgentAPIEvents.ConnectorInformation,
162
            Grok.VoiceAgentAPIEvents.InputAudioBufferCommitted,
163
            Grok.VoiceAgentAPIEvents.WebSocketError,
164
            Grok.VoiceAgentAPIEvents.Unknown,
165
        ].forEach((evtName) => {
166
            voiceAIClient.addEventListener(evtName, (e) => {
167
                Logger.write(`===${e.name}===>${JSON.stringify(e)}`);
168
            });
169
        });
170
171
        voiceAIClient.addEventListener(Grok.Events.WebSocketMediaStarted, () => {
172
            VoxEngine.sendMediaBetween(call, voiceAIClient);
173
        });
174
175
        voiceAIClient.addEventListener(Grok.Events.WebSocketMediaEnded, (event) => {
176
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
177
            if (hangupCall) callCloseHandler();
178
            else if (forwardToLiveAgent) {
179
                call.say("Here is where I would forward the call via the phone network or SIP.");
180
                call.addEventListener(CallEvents.PlaybackFinished, callCloseHandler);
181
            } else return;
182
        });
183
    } catch (error) {
184
        Logger.write("===SOMETHING_WENT_WRONG===");
185
        Logger.write(error);
186
        VoxEngine.terminate();
187
    }
188
});