Example: Using Grok features

1
require(Modules.Grok);
2
const SYSTEM_PROMPT = `
3
  Your name is Voxi. You are a helpful voice assistant for phone callers representing the company Voximplant (pronounced VOX-im-plant).
4
  You can answer questions about the company, its voice AI integrations, and X/Twitter posts from the "voximplant" and "aylarov" handles.
5
  "aylarov" is the X handle for Alexey Aylarov, Voximplant's CEO.
6
7
  Keep responses short and telephony-friendly (usually 1-2 sentences).
8
  If the user asks for a live agent or an operator, call the "forward_to_agent" function.
9
  If the user says goodbye, call the "hangup_call" function.
10
11
  When answering a company/product question, prefer searching the knowledge base first.
12
  `;
13
const COLLECTION_ID = "collection_4c5a63ab-f739-4c13-93d2-05b74095c34a"; // uploaded documents to show RAG
14
15
// -------------------- Grok Voice Agent settings --------------------
16
const SESSION_PARAMETERS = {
17
    session: {
18
        voice: "Ara",
19
        turn_detection: {type: "server_vad"},
20
        instructions: SYSTEM_PROMPT,
21
        tools: [
22
            {type: "web_search"},
23
            {
24
                type: "file_search",
25
                vector_store_ids: [COLLECTION_ID],
26
                max_num_results: 5,
27
            },
28
            {
29
                type: "x_search",
30
                allowed_x_handles: ["voximplant", "aylarov"],
31
            },
32
            {
33
                type: "function",
34
                name: "forward_to_agent",
35
                description: "Forward the user to a live agent",
36
                parameters: {
37
                    type: "object",
38
                    properties: {},
39
                    required: [],
40
                },
41
            },
42
            {
43
                type: "function",
44
                name: "hangup_call",
45
                description: "Hangup the call",
46
                parameters: {
47
                    type: "object",
48
                    properties: {},
49
                    required: [],
50
                },
51
            },
52
        ],
53
    },
54
};
55
56
VoxEngine.addEventListener(AppEvents.CallAlerting, async ({call}) => {
57
    let voiceAIClient = undefined;
58
    let hangupCall = false,
59
        forwardToLiveAgent = false;
60
61
    call.answer();
62
    call.record({hd_audio: true, stereo: true}); // optional: call recording
63
64
    const callCloseHandler = () => {
65
        voiceAIClient?.close();
66
        VoxEngine.terminate();
67
    };
68
    call.addEventListener(CallEvents.Disconnected, callCloseHandler);
69
    call.addEventListener(CallEvents.Failed, callCloseHandler);
70
71
    try {
72
        voiceAIClient = await Grok.createVoiceAgentAPIClient({
73
            xAIApiKey: VoxEngine.getSecretValue('GROK_API_KEY'),
74
            onWebSocketClose: (event) => {
75
                Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
76
                VoxEngine.terminate();
77
            },
78
        });
79
80
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.ConversationCreated, (event) => {
81
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
82
83
84
            voiceAIClient.sessionUpdate(SESSION_PARAMETERS);
85
        });
86
87
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.SessionUpdated, (event) => {
88
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
89
            VoxEngine.sendMediaBetween(call, voiceAIClient);
90
            voiceAIClient.responseCreate({instructions: "Hello."});
91
        });
92
93
        // -------------------- Barge-in (keep it interruption-friendly) --------------------
94
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.InputAudioBufferSpeechStarted, (event) => {
95
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
96
            voiceAIClient.clearMediaBuffer();
97
        });
98
99
        // -------------------- Function calling --------------------
100
        voiceAIClient.addEventListener(Grok.VoiceAgentAPIEvents.ResponseFunctionCallArgumentsDone, (event) => {
101
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
102
103
            const {name, call_id} = event?.data?.payload || {};
104
            let output;
105
106
            // Ignore server-side tools like collections_search / web_search / x_search
107
            if (name !== "forward_to_agent" && name !== "hangup_call") {
108
                Logger.write(`===Ignoring unhandled function call: ${name}===`);
109
                return;
110
            }
111
112
            if (name === "forward_to_agent") {
113
                forwardToLiveAgent = true;
114
                output = {result: "Forwarding your call to a live agent. Please hold on."};
115
            } else if (name === "hangup_call") {
116
                hangupCall = true;
117
                output = {result: "Have a great day, goodbye!"};
118
            }
119
120
            // Create a conversationItem and send it
121
            voiceAIClient.conversationItemCreate({
122
                item: {
123
                    type: "function_call_output",
124
                    call_id,
125
                    output: JSON.stringify(output),
126
                },
127
            });
128
            voiceAIClient.responseCreate({});
129
        });
130
131
        // -------------------- Log Other Events --------------------
132
        [
133
            CallEvents.FirstAudioPacketReceived,
134
            Grok.Events.WebSocketMediaStarted,
135
            Grok.VoiceAgentAPIEvents.InputAudioBufferSpeechStopped,
136
            Grok.VoiceAgentAPIEvents.ConversationItemInputAudioTranscriptionCompleted,
137
            Grok.VoiceAgentAPIEvents.ConversationItemAdded,
138
            Grok.VoiceAgentAPIEvents.ResponseCreated,
139
            Grok.VoiceAgentAPIEvents.ResponseOutputItemAdded,
140
            Grok.VoiceAgentAPIEvents.ResponseDone,
141
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioTranscriptDelta,
142
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioTranscriptDone,
143
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioDelta, // Not in enum
144
            Grok.VoiceAgentAPIEvents.ResponseOutputAudioDone,
145
            Grok.VoiceAgentAPIEvents.ResponseOutputItemDone,
146
            Grok.VoiceAgentAPIEvents.ConnectorInformation,
147
            Grok.VoiceAgentAPIEvents.InputAudioBufferCommitted,
148
            Grok.VoiceAgentAPIEvents.WebSocketError,
149
            Grok.VoiceAgentAPIEvents.Unknown,
150
        ].forEach((evtName) => {
151
            voiceAIClient.addEventListener(evtName, (e) => {
152
                Logger.write(`===${e.name}===>${JSON.stringify(e)}`);
153
            });
154
        });
155
156
        voiceAIClient.addEventListener(Grok.Events.WebSocketMediaEnded, (event) => {
157
            Logger.write(`===${event.name}===>${JSON.stringify(event.data)}`);
158
            if (hangupCall) callCloseHandler();
159
            else if (forwardToLiveAgent) {
160
                call.say("Here is where I would forward the call via the phone network, SIP, or WhatsApp.");
161
                // See the forwardCallToPSTN, forwardCallToSIP, forwardCallToUser, and  handleBlindTransfer call methods
162
                // For this simple demo, we will just close and hang-up
163
                call.addEventListener(CallEvents.PlaybackFinished, callCloseHandler);
164
            }
165
        });
166
    } catch (error) {
167
        Logger.write("===SOMETHING_WENT_WRONG===");
168
        Logger.write(error);
169
        VoxEngine.terminate();
170
    }
171
});