add websocket relayer for openai realtime api

This commit is contained in:
RockYang 2024-10-16 18:16:09 +08:00
parent 155c56f502
commit 2102e1afbb
12 changed files with 1337 additions and 624 deletions

View File

@ -221,6 +221,7 @@ func needLogin(c *gin.Context) bool {
c.Request.URL.Path == "/api/suno/detail" || c.Request.URL.Path == "/api/suno/detail" ||
c.Request.URL.Path == "/api/suno/play" || c.Request.URL.Path == "/api/suno/play" ||
c.Request.URL.Path == "/api/download" || c.Request.URL.Path == "/api/download" ||
c.Request.URL.Path == "/api/realtime" ||
strings.HasPrefix(c.Request.URL.Path, "/api/test") || strings.HasPrefix(c.Request.URL.Path, "/api/test") ||
strings.HasPrefix(c.Request.URL.Path, "/api/payment/notify/") || strings.HasPrefix(c.Request.URL.Path, "/api/payment/notify/") ||
strings.HasPrefix(c.Request.URL.Path, "/api/user/clogin") || strings.HasPrefix(c.Request.URL.Path, "/api/user/clogin") ||

View File

@ -0,0 +1,173 @@
package handler
import (
"github.com/gin-gonic/gin"
"github.com/gorilla/websocket"
"log"
"net/http"
)
// * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
// * Copyright 2023 The Geek-AI Authors. All rights reserved.
// * Use of this source code is governed by a Apache-2.0 license
// * that can be found in the LICENSE file.
// * @Author yangjian102621@163.com
// * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
// 实时 API 中继器
type RealtimeHandler struct {
BaseHandler
}
func NewRealtimeHandler() *RealtimeHandler {
return &RealtimeHandler{}
}
func (h *RealtimeHandler) Connection(c *gin.Context) {
// 获取客户端请求中指定的子协议
clientProtocols := c.GetHeader("Sec-WebSocket-Protocol")
logger.Info(clientProtocols)
// 升级HTTP连接为WebSocket并传入客户端请求的子协议
upgrader := websocket.Upgrader{
CheckOrigin: func(r *http.Request) bool { return true },
Subprotocols: []string{clientProtocols},
}
ws, err := upgrader.Upgrade(c.Writer, c.Request, nil)
if err != nil {
logger.Error(err)
c.Abort()
return
}
defer ws.Close()
// 连接到真实的后端服务器,传入相同的子协议
headers := http.Header{}
if clientProtocols != "" {
headers.Set("Sec-WebSocket-Protocol", clientProtocols)
}
for key, values := range headers {
for _, value := range values {
logger.Infof("%s: %s", key, value)
}
}
backendConn, _, err := websocket.DefaultDialer.Dial("wss://api.geekai.pro/v1/realtime?model=gpt-4o-realtime-preview-2024-10-01", headers)
if err != nil {
log.Printf("Failed to connect to backend: %v", err)
return
}
defer backendConn.Close()
//logger.Info(ws.Subprotocol(), ",", backendConn.Subprotocol())
//// 确保协议一致性,如果失败返回
//if ws.Subprotocol() != backendConn.Subprotocol() {
// log.Println("Subprotocol mismatch")
// return
//}
// 开始双向转发
errorChan := make(chan error, 2)
go relay(ws, backendConn, errorChan)
go relay(backendConn, ws, errorChan)
// 等待其中一个连接关闭
<-errorChan
log.Println("Relay ended")
}
func relay(src, dst *websocket.Conn, errorChan chan error) {
for {
messageType, message, err := src.ReadMessage()
if err != nil {
errorChan <- err
return
}
err = dst.WriteMessage(messageType, message)
if err != nil {
errorChan <- err
return
}
}
}
//func (h *RealtimeHandler) handleMessage(client *RealtimeClient, message []byte) {
// var event Event
// err := json.Unmarshal(message, &event)
// if err != nil {
// logger.Infof("Error parsing event from client: %s", message)
// return
// }
// logger.Infof("Relaying %q to OpenAI", event.Type)
// client.Send(event)
//}
//
//func relay(src, dst *websocket.Conn, errorChan chan error) {
// for {
// messageType, message, err := src.ReadMessage()
// if err != nil {
// errorChan <- err
// return
// }
// err = dst.WriteMessage(messageType, message)
// if err != nil {
// errorChan <- err
// return
// }
// }
//}
//
//func NewRealtimeClient(apiKey string) *RealtimeClient {
// return &RealtimeClient{
// APIKey: apiKey,
// send: make(chan Event, 100),
// }
//}
//
//func (rc *RealtimeClient) Connect() error {
// u := url.URL{Scheme: "wss", Host: "api.geekai.pro", Path: "v1/realtime", RawQuery: "model=gpt-4o-realtime-preview-2024-10-01"}
// c, _, err := websocket.DefaultDialer.Dial(u.String(), nil)
// if err != nil {
// return err
// }
// rc.conn = c
//
// go rc.readPump()
// go rc.writePump()
//
// return nil
//}
//
//func (rc *RealtimeClient) readPump() {
// defer rc.conn.Close()
// for {
// _, message, err := rc.conn.ReadMessage()
// if err != nil {
// log.Println("read error:", err)
// return
// }
// var event Event
// err = json.Unmarshal(message, &event)
// if err != nil {
// log.Println("parse error:", err)
// continue
// }
// rc.send <- event
// }
//}
//
//func (rc *RealtimeClient) writePump() {
// defer rc.conn.Close()
// for event := range rc.send {
// err := rc.conn.WriteJSON(event)
// if err != nil {
// log.Println("write error:", err)
// return
// }
// }
//}
//
//func (rc *RealtimeClient) Send(event Event) {
// rc.send <- event
//}

View File

@ -554,6 +554,10 @@ func main() {
group.POST("/list/luma", h.LumaList) group.POST("/list/luma", h.LumaList)
group.GET("/remove", h.Remove) group.GET("/remove", h.Remove)
}), }),
fx.Provide(handler.NewRealtimeHandler),
fx.Invoke(func(s *core.AppServer, h *handler.RealtimeHandler) {
s.Engine.Any("/api/realtime", h.Connection)
}),
) )
// 启动应用程序 // 启动应用程序
go func() { go func() {

View File

@ -0,0 +1 @@
ALTER TABLE `chatgpt_chat_models` CHANGE `value` `value` VARCHAR(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_0900_ai_ci NOT NULL COMMENT '模型值';

View File

@ -0,0 +1,196 @@
.realtime-conversation {
/********************** connection ****************************/
.connection-container {
background-color: #000;
display: flex;
flex-direction: column;
justify-content: center;
align-items: center;
margin: 0;
overflow: hidden;
font-family: Arial, sans-serif;
width 100vw
.phone-container {
position: relative;
width: 200px;
height: 200px;
}
.phone {
position: absolute;
top: 50%;
left: 50%;
transform: translate(-50%, -50%);
width: 60px;
height: 60px;
background-color: #00ffcc;
mask: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 24 24'%3E%3Cpath d='M20 15.5c-1.25 0-2.45-.2-3.57-.57a1.02 1.02 0 0 0-1.02.24l-2.2 2.2a15.074 15.074 0 0 1-6.59-6.59l2.2-2.2c.27-.27.35-.68.24-1.02a11.36 11.36 0 0 1-.57-3.57c0-.55-.45-1-1-1H4c-.55 0-1 .45-1 1 0 9.39 7.61 17 17 17 .55 0 1-.45 1-1v-3.5c0-.55-.45-1-1-1zM5.03 5h1.5c.07.89.22 1.76.46 2.59l-1.2 1.2c-.41-1.2-.67-2.47-.76-3.79zM19 18.97c-1.32-.09-2.59-.35-3.8-.75l1.2-1.2c.85.24 1.72.39 2.6.45v1.5z'/%3E%3C/svg%3E") no-repeat 50% 50%;
mask-size: cover;
-webkit-mask: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 24 24'%3E%3Cpath d='M20 15.5c-1.25 0-2.45-.2-3.57-.57a1.02 1.02 0 0 0-1.02.24l-2.2 2.2a15.074 15.074 0 0 1-6.59-6.59l2.2-2.2c.27-.27.35-.68.24-1.02a11.36 11.36 0 0 1-.57-3.57c0-.55-.45-1-1-1H4c-.55 0-1 .45-1 1 0 9.39 7.61 17 17 17 .55 0 1-.45 1-1v-3.5c0-.55-.45-1-1-1zM5.03 5h1.5c.07.89.22 1.76.46 2.59l-1.2 1.2c-.41-1.2-.67-2.47-.76-3.79zM19 18.97c-1.32-.09-2.59-.35-3.8-.75l1.2-1.2c.85.24 1.72.39 2.6.45v1.5z'/%3E%3C/svg%3E") no-repeat 50% 50%;
-webkit-mask-size: cover;
animation: shake 0.5s ease-in-out infinite;
}
.signal {
position: absolute;
top: 50%;
left: 50%;
transform: translate(-50%, -50%);
width: 100px;
height: 100px;
border: 2px dashed #00ffcc;
border-radius: 50%;
opacity: 0;
animation: signal 2s linear infinite;
}
.signal:nth-child(2) {
animation-delay: 0.5s;
}
.signal:nth-child(3) {
animation-delay: 1s;
}
.status-text {
color: #00ffcc;
font-size: 18px;
margin-top: 20px;
height: 1.2em;
overflow: hidden;
}
@keyframes shake {
0%, 100% { transform: translate(-50%, -50%) rotate(0deg); }
25% { transform: translate(-52%, -48%) rotate(-5deg); }
75% { transform: translate(-48%, -52%) rotate(5deg); }
}
@keyframes signal {
0% {
width: 60px;
height: 60px;
opacity: 1;
}
100% {
width: 200px;
height: 200px;
opacity: 0;
}
}
}
/*********** end of connection ************/
.conversation-container {
background: linear-gradient(to right, #2c3e50, #4a5568, #6b46c1);
display: flex;
height 100%
flex-direction: column;
justify-content: space-between;
align-items: center;
padding: 0;
width 100vw
.wave-container {
padding 3rem
.wave-animation {
display: flex;
justify-content: center;
align-items: center;
gap: 10px;
.wave-ellipse {
width: 40px;
height: 40px;
background-color: white;
border-radius: 20px;
animation: wave 0.8s infinite ease-in-out;
}
.wave-ellipse:nth-child(odd) {
height: 60px;
}
.wave-ellipse:nth-child(even) {
height: 80px;
}
}
}
@keyframes wave {
0%, 100% {
transform: scaleY(0.8);
}
50% {
transform: scaleY(1.2);
}
}
.wave-ellipse:nth-child(2) {
animation-delay: 0.1s;
}
.wave-ellipse:nth-child(3) {
animation-delay: 0.2s;
}
.wave-ellipse:nth-child(4) {
animation-delay: 0.3s;
}
.wave-ellipse:nth-child(5) {
animation-delay: 0.4s;
}
.voice-indicators {
display flex
flex-flow row
justify-content: space-between;
width 100%
.left {
margin-left 3rem
}
.right {
margin-right 3rem
}
}
.call-controls {
display: flex;
justify-content: center;
gap: 3rem;
padding 3rem
.call-button {
width: 60px;
height: 60px;
border-radius: 50%;
border: none;
display: flex;
justify-content: center;
align-items: center;
font-size: 24px;
color: white;
cursor: pointer;
.iconfont {
font-size 24px
}
}
.hangup {
background-color: #e74c3c;
}
.answer {
background-color: #2ecc71;
}
.icon {
font-size: 28px;
}
}
}
}

View File

@ -1,161 +0,0 @@
<template>
<!--语音通话组件-->
<div class="video-call-container" :style="{height: height}">
<div class="wave-container">
<div class="wave-animation">
<div v-for="i in 5" :key="i" class="wave-ellipse"></div>
</div>
</div>
<!-- 其余部分保持不变 -->
<div class="voice-indicators">
<div class="voice-indicator left">
<canvas ref="canvasClientRef" width="600" height="200"></canvas>
</div>
<div class="voice-indicator right">
<canvas ref="canvasServerRef" width="600" height="200"></canvas>
</div>
</div>
<div class="call-controls">
<button class="call-button hangup" @click="hangUp">
<i class="iconfont icon-hung-up"></i>
</button>
</div>
</div>
</template>
<script setup>
import {onMounted, onUnmounted, ref} from "vue";
const leftVoiceActive = ref(false);
const rightVoiceActive = ref(false);
const props = defineProps({
height: {
type: String,
default: '100vh'
}
})
const emits = defineEmits(['hangUp']);
const animateVoice = () => {
leftVoiceActive.value = Math.random() > 0.5;
rightVoiceActive.value = Math.random() > 0.5;
};
let voiceInterval;
onMounted(() => {
voiceInterval = setInterval(animateVoice, 500);
});
onUnmounted(() => {
clearInterval(voiceInterval);
});
const hangUp = () => {
console.log('Call hung up');
emits('hangUp')
};
</script>
<style scoped lang="stylus">
.video-call-container {
background: linear-gradient(to right, #2c3e50, #4a5568, #6b46c1);
display: flex;
flex-direction: column;
justify-content: space-between;
align-items: center;
padding: 0;
width 100vw
.wave-container {
padding 3rem
.wave-animation {
display: flex;
justify-content: center;
align-items: center;
gap: 10px;
}
}
.wave-ellipse {
width: 40px;
height: 40px;
background-color: white;
border-radius: 20px;
animation: wave 0.8s infinite ease-in-out;
}
.wave-ellipse:nth-child(odd) {
height: 60px;
}
.wave-ellipse:nth-child(even) {
height: 80px;
}
@keyframes wave {
0%, 100% {
transform: scaleY(0.8);
}
50% {
transform: scaleY(1.2);
}
}
.wave-ellipse:nth-child(2) {
animation-delay: 0.1s;
}
.wave-ellipse:nth-child(3) {
animation-delay: 0.2s;
}
.wave-ellipse:nth-child(4) {
animation-delay: 0.3s;
}
.wave-ellipse:nth-child(5) {
animation-delay: 0.4s;
}
.call-controls {
display: flex;
justify-content: center;
gap: 3rem;
padding 3rem
.call-button {
width: 60px;
height: 60px;
border-radius: 50%;
border: none;
display: flex;
justify-content: center;
align-items: center;
font-size: 24px;
color: white;
cursor: pointer;
.iconfont {
font-size 24px
}
}
.hangup {
background-color: #e74c3c;
}
.answer {
background-color: #2ecc71;
}
.icon {
font-size: 28px;
}
}
}
</style>

View File

@ -0,0 +1,357 @@
<template>
<el-container class="realtime-conversation" :style="{height: height}">
<!-- connection animation -->
<el-container class="connection-container" v-if="!isConnected">
<div class="phone-container">
<div class="signal"></div>
<div class="signal"></div>
<div class="signal"></div>
<div class="phone"></div>
</div>
<div class="status-text">{{ connectingText }}</div>
</el-container>
<!-- conversation body -->
<div class="conversation-container" v-else>
<div class="wave-container">
<div class="wave-animation">
<div v-for="i in 5" :key="i" class="wave-ellipse"></div>
</div>
</div>
<!-- 其余部分保持不变 -->
<div class="voice-indicators">
<div class="voice-indicator left">
<canvas ref="clientCanvasRef"></canvas>
</div>
<div class="voice-indicator right">
<canvas ref="serverCanvasRef"></canvas>
</div>
</div>
<div class="call-controls">
<el-tooltip content="长按发送语音" placement="top" effect="light">
<ripple-button>
<button class="call-button answer" @mousedown="startRecording" @mouseup="stopRecording">
<i class="iconfont icon-mic-bold"></i>
</button>
</ripple-button>
</el-tooltip>
<el-tooltip content="结束通话" placement="top" effect="light">
<button class="call-button hangup" @click="hangUp">
<i class="iconfont icon-hung-up"></i>
</button>
</el-tooltip>
</div>
</div>
</el-container>
</template>
<script setup>
import RippleButton from "@/components/ui/RippleButton.vue";
import { ref, onMounted, onUnmounted } from 'vue';
import { RealtimeClient } from '@openai/realtime-api-beta';
import { WavRecorder, WavStreamPlayer } from '@/lib/wavtools/index.js';
import { instructions } from '@/utils/conversation_config.js';
import { WavRenderer } from '@/utils/wav_renderer';
import {showMessageError} from "@/utils/dialog";
// eslint-disable-next-line no-unused-vars,no-undef
const props = defineProps({
height: {
type: String,
default: '100vh'
}
})
// eslint-disable-next-line no-undef
const emits = defineEmits(['close']);
/********************** connection animation code *************************/
const fullText = "正在接通中...";
const connectingText = ref("")
let index = 0;
const typeText = () => {
if (index < fullText.length) {
connectingText.value += fullText[index];
index++;
setTimeout(typeText, 300); // 300
} else {
setTimeout(() => {
connectingText.value = '';
index = 0;
typeText();
}, 1000); // 1
}
}
/*************************** end of code ****************************************/
/********************** conversation process code ***************************/
const leftVoiceActive = ref(false);
const rightVoiceActive = ref(false);
const animateVoice = () => {
leftVoiceActive.value = Math.random() > 0.5;
rightVoiceActive.value = Math.random() > 0.5;
};
const wavRecorder = ref(new WavRecorder({ sampleRate: 24000 }));
const wavStreamPlayer = ref(new WavStreamPlayer({ sampleRate: 24000 }));
const client = ref(
new RealtimeClient({
url: "ws://localhost:5678/api/realtime",
apiKey: "sk-Gc5cEzDzGQLIqxWA9d62089350F3454bB359C4A3Fa21B3E4",
dangerouslyAllowAPIKeyInBrowser: true,
})
);
// // Set up client instructions and transcription
client.value.updateSession({
instructions: instructions,
turn_detection: null,
input_audio_transcription: { model: 'whisper-1' },
voice: 'alloy',
});
// set voice wave canvas
const clientCanvasRef = ref(null);
const serverCanvasRef = ref(null);
// const eventsScrollRef = ref(null);
// const startTime = ref(new Date().toISOString());
// const items = ref([]);
// const realtimeEvents = ref([]);
// const expandedEvents = reactive({});
const isConnected = ref(false);
// const canPushToTalk = ref(true);
const isRecording = ref(false);
// const memoryKv = ref({});
// const coords = ref({ lat: 37.775593, lng: -122.418137 });
// const marker = ref(null);
// Methods
// const formatTime = (timestamp) => {
// const t0 = new Date(startTime.value).valueOf();
// const t1 = new Date(timestamp).valueOf();
// const delta = t1 - t0;
// const hs = Math.floor(delta / 10) % 100;
// const s = Math.floor(delta / 1000) % 60;
// const m = Math.floor(delta / 60_000) % 60;
// const pad = (n) => {
// let s = n + '';
// while (s.length < 2) {
// s = '0' + s;
// }
// return s;
// };
// return `${pad(m)}:${pad(s)}.${pad(hs)}`;
// };
const connect = async () => {
// startTime.value = new Date().toISOString();
// realtimeEvents.value = [];
// items.value = client.value.conversation.getItems();
if (isConnected.value) {
return
}
try {
await client.value.connect();
await wavRecorder.value.begin();
await wavStreamPlayer.value.connect();
isConnected.value = true;
console.log("对话连接成功!")
client.value.sendUserMessageContent([
{
type: 'input_text',
text: '你好,我是老阳!',
},
]);
if (client.value.getTurnDetectionType() === 'server_vad') {
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
}
} catch (e) {
showMessageError(e.message)
}
};
// const disconnectConversation = async () => {
// isConnected.value = false;
// // realtimeEvents.value = [];
// // items.value = [];
// // memoryKv.value = {};
// // coords.value = { lat: 37.775593, lng: -122.418137 };
// // marker.value = null;
//
// client.value.disconnect();
// await wavRecorder.value.end();
// await wavStreamPlayer.value.interrupt();
// };
// const deleteConversationItem = async (id) => {
// client.value.deleteItem(id);
// };
const startRecording = async () => {
isRecording.value = true;
const trackSampleOffset = await wavStreamPlayer.value.interrupt();
if (trackSampleOffset?.trackId) {
const { trackId, offset } = trackSampleOffset;
client.value.cancelResponse(trackId, offset);
}
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
};
const stopRecording = async () => {
isRecording.value = false;
await wavRecorder.value.pause();
client.value.createResponse();
};
// const changeTurnEndType = async (value) => {
// if (value === 'none' && wavRecorder.value.getStatus() === 'recording') {
// await wavRecorder.value.pause();
// }
// client.value.updateSession({
// turn_detection: value === 'none' ? null : { type: 'server_vad' },
// });
// if (value === 'server_vad' && client.value.isConnected()) {
// await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
// }
// canPushToTalk.value = value === 'none';
// };
//
// const toggleEventDetails = (eventId) => {
// if (expandedEvents[eventId]) {
// delete expandedEvents[eventId];
// } else {
// expandedEvents[eventId] = true;
// }
// };
// Lifecycle hooks and watchers
const initialize = async () => {
// Set up render loops for the visualization canvas
let isLoaded = true;
const render = () => {
if (isLoaded) {
if (clientCanvasRef.value) {
const canvas = clientCanvasRef.value;
if (!canvas.width || !canvas.height) {
canvas.width = canvas.offsetWidth;
canvas.height = canvas.offsetHeight;
}
const ctx = canvas.getContext('2d');
if (ctx) {
ctx.clearRect(0, 0, canvas.width, canvas.height);
const result = wavRecorder.value.recording
? wavRecorder.value.getFrequencies('voice')
: { values: new Float32Array([0]) };
WavRenderer.drawBars(canvas, ctx, result.values, '#0099ff', 10, 0, 8);
}
}
if (serverCanvasRef.value) {
const canvas = serverCanvasRef.value;
if (!canvas.width || !canvas.height) {
canvas.width = canvas.offsetWidth;
canvas.height = canvas.offsetHeight;
}
const ctx = canvas.getContext('2d');
if (ctx) {
ctx.clearRect(0, 0, canvas.width, canvas.height);
const result = wavStreamPlayer.value.analyser
? wavStreamPlayer.value.getFrequencies('voice')
: { values: new Float32Array([0]) };
WavRenderer.drawBars(canvas, ctx, result.values, '#009900', 10, 0, 8);
}
}
requestAnimationFrame(render);
}
};
render();
// Set up client event listeners
client.value.on('realtime.event', (realtimeEvent) => {
// realtimeEvents.value = realtimeEvents.value.slice();
// const lastEvent = realtimeEvents.value[realtimeEvents.value.length - 1];
// if (lastEvent?.event.type === realtimeEvent.event.type) {
// lastEvent.count = (lastEvent.count || 0) + 1;
// realtimeEvents.value.splice(-1, 1, lastEvent);
// } else {
// realtimeEvents.value.push(realtimeEvent);
// }
// console.log(realtimeEvent)
});
client.value.on('error', (event) => console.error(event));
client.value.on('conversation.interrupted', async () => {
const trackSampleOffset = await wavStreamPlayer.value.interrupt();
if (trackSampleOffset?.trackId) {
const { trackId, offset } = trackSampleOffset;
client.value.cancelResponse(trackId, offset);
}
});
client.value.on('conversation.updated', async ({ item, delta }) => {
console.log('item updated', item, delta)
if (delta?.audio) {
wavStreamPlayer.value.add16BitPCM(delta.audio, item.id);
}
if (item.status === 'completed' && item.formatted.audio?.length) {
const wavFile = await WavRecorder.decode(
item.formatted.audio,
24000,
24000
);
item.formatted.file = wavFile;
}
});
}
// Watchers
// watch(realtimeEvents, () => {
// if (eventsScrollRef.value) {
// const eventsEl = eventsScrollRef.value;
// eventsEl.scrollTop = eventsEl.scrollHeight;
// }
// });
// watch(items, () => {
// const conversationEls = document.querySelectorAll('[data-conversation-content]');
// conversationEls.forEach((el) => {
// el.scrollTop = el.scrollHeight;
// });
// });
const voiceInterval = ref(null);
onMounted(() => {
initialize()
voiceInterval.value = setInterval(animateVoice, 500);
typeText()
});
onUnmounted(() => {
clearInterval(voiceInterval.value);
client.value.reset();
});
const hangUp = async () => {
emits('close')
isConnected.value = false;
client.value.disconnect();
await wavRecorder.value.end();
await wavStreamPlayer.value.interrupt();
};
// eslint-disable-next-line no-undef
defineExpose({ connect });
</script>
<style scoped lang="stylus">
@import "@/assets/css/realtime.styl"
</style>

View File

@ -0,0 +1,102 @@
<template>
<button
class="ripple-button"
@mousedown="startRipples"
@mouseup="stopRipples"
@mouseleave="stopRipples"
>
<slot></slot>
<span
v-for="ripple in ripples"
:key="ripple.id"
class="ripple"
:style="getRippleStyle(ripple)"
></span>
</button>
</template>
<script setup>
import { ref } from 'vue';
const ripples = ref([]);
let rippleCount = 0;
let animationId;
const startRipples = (event) => {
const button = event.currentTarget;
const rect = button.getBoundingClientRect();
const size = Math.max(rect.width, rect.height);
// const x = event.clientX - rect.left;
// const y = event.clientY - rect.top;
const x = rect.right - rect.left - size/2;
const y = rect.bottom - rect.top - size/2;
const createRipple = () => {
ripples.value.push({
id: rippleCount++,
x,
y,
size: 0,
opacity: 0.5
});
if (ripples.value.length > 3) {
ripples.value.shift();
}
};
const animate = () => {
ripples.value.forEach(ripple => {
ripple.size += 2;
ripple.opacity -= 0.01;
});
ripples.value = ripples.value.filter(ripple => ripple.opacity > 0);
if (ripples.value.length < 3) {
createRipple();
}
animationId = requestAnimationFrame(animate);
};
createRipple();
animate();
};
const stopRipples = () => {
cancelAnimationFrame(animationId);
ripples.value = [];
};
const getRippleStyle = (ripple) => ({
left: `${ripple.x}px`,
top: `${ripple.y}px`,
width: `${ripple.size}px`,
height: `${ripple.size}px`,
opacity: ripple.opacity
});
</script>
<style scoped lang="stylus">
.ripple-button {
position: relative;
overflow: hidden;
border: none;
background none;
color: white;
cursor: pointer;
border-radius: 50%;
outline: none;
margin 0
padding 0
}
.ripple {
position: absolute;
border-radius: 50%;
background-color: rgba(255, 255, 255, 0.7);
transform: translate(-50%, -50%);
pointer-events: none;
}
</style>

View File

@ -328,7 +328,7 @@ const routes = [
name: 'test2', name: 'test2',
path: '/test2', path: '/test2',
meta: {title: '测试页面'}, meta: {title: '测试页面'},
component: () => import('@/views/Test2.vue'), component: () => import('@/views/RealtimeTest.vue'),
}, },
{ {
name: 'NotFound', name: 'NotFound',

View File

@ -0,0 +1,471 @@
<template>
<div data-component="ConsolePage">
<div class="content-top">
<div class="content-title">
<img src="/openai-logomark.svg" alt="OpenAI Logo" />
<span>realtime console</span>
</div>
</div>
<div class="content-main">
<div class="content-logs">
<div class="content-block events">
<div class="visualization">
<div class="visualization-entry client">
<canvas ref="clientCanvasRef" />
</div>
<div class="visualization-entry server">
<canvas ref="serverCanvasRef" />
</div>
</div>
<div class="content-block-title">events</div>
<div class="content-block-body" ref="eventsScrollRef">
<template v-if="!realtimeEvents.length">
awaiting connection...
</template>
<template v-else>
<div v-for="(realtimeEvent, i) in realtimeEvents" :key="realtimeEvent.event.event_id" class="event">
<div class="event-timestamp">
{{ formatTime(realtimeEvent.time) }}
</div>
<div class="event-details">
<div
class="event-summary"
@click="toggleEventDetails(realtimeEvent.event.event_id)"
>
<div
:class="[
'event-source',
realtimeEvent.event.type === 'error'
? 'error'
: realtimeEvent.source,
]"
>
<component :is="realtimeEvent.source === 'client' ? ArrowUp : ArrowDown" />
<span>
{{ realtimeEvent.event.type === 'error'
? 'error!'
: realtimeEvent.source }}
</span>
</div>
<div class="event-type">
{{ realtimeEvent.event.type }}
{{ realtimeEvent.count ? `(${realtimeEvent.count})` : '' }}
</div>
</div>
<div
v-if="expandedEvents[realtimeEvent.event.event_id]"
class="event-payload"
>
{{ JSON.stringify(realtimeEvent.event, null, 2) }}
</div>
</div>
</div>
</template>
</div>
</div>
<div class="content-block conversation">
<div class="content-block-title">conversation</div>
<div class="content-block-body" data-conversation-content>
<template v-if="!items.length">
awaiting connection...
</template>
<template v-else>
<div
v-for="(conversationItem, i) in items"
:key="conversationItem.id"
class="conversation-item"
>
<div :class="['speaker', conversationItem.role || '']">
<div>
{{
(conversationItem.role || conversationItem.type).replaceAll(
'_',
' '
)
}}
</div>
<div class="close" @click="deleteConversationItem(conversationItem.id)">
<X />
</div>
</div>
<div class="speaker-content">
<!-- tool response -->
<div v-if="conversationItem.type === 'function_call_output'">
{{ conversationItem.formatted.output }}
</div>
<!-- tool call -->
<div v-if="conversationItem.formatted.tool">
{{ conversationItem.formatted.tool.name }}(
{{ conversationItem.formatted.tool.arguments }})
</div>
<div
v-if="
!conversationItem.formatted.tool &&
conversationItem.role === 'user'
"
>
{{
conversationItem.formatted.transcript ||
(conversationItem.formatted.audio?.length
? '(awaiting transcript)'
: conversationItem.formatted.text || '(item sent)')
}}
</div>
<div
v-if="
!conversationItem.formatted.tool &&
conversationItem.role === 'assistant'
"
>
{{
conversationItem.formatted.transcript ||
conversationItem.formatted.text ||
'(truncated)'
}}
</div>
<audio
v-if="conversationItem.formatted.file"
:src="conversationItem.formatted.file.url"
controls
/>
</div>
</div>
</template>
</div>
</div>
<div class="content-actions" style="position:absolute; top: 0; left: 0">
<el-button
:type="isConnected ? '' : 'primary'"
@click="connectConversation"
>
{{isConnected ? '断开连接' : '连接对话'}}
</el-button>
<el-button @mousedown="startRecording" @mouseup="stopRecording">开始讲话</el-button>
</div>
</div>
</div>
</div>
</template>
<script setup>
import { ref, reactive, onMounted, onUnmounted, watch } from 'vue';
import { RealtimeClient } from '@openai/realtime-api-beta';
import { WavRecorder, WavStreamPlayer } from '@/lib/wavtools/index.js';
import { instructions } from '@/utils/conversation_config.js';
import { WavRenderer } from '@/utils/wav_renderer';
// Constants
const LOCAL_RELAY_SERVER_URL = process.env.REACT_APP_LOCAL_RELAY_SERVER_URL || '';
// Reactive state
const apiKey = ref(
LOCAL_RELAY_SERVER_URL
? ''
: localStorage.getItem('tmp::voice_api_key') || prompt('OpenAI API Key') || ''
);
const wavRecorder = ref(new WavRecorder({ sampleRate: 24000 }));
const wavStreamPlayer = ref(new WavStreamPlayer({ sampleRate: 24000 }));
const client = ref(
new RealtimeClient({
url: "wss://api.geekai.pro/v1/realtime",
apiKey: "sk-Gc5cEzDzGQLIqxWA9d62089350F3454bB359C4A3Fa21B3E4",
dangerouslyAllowAPIKeyInBrowser: true,
})
);
const clientCanvasRef = ref(null);
const serverCanvasRef = ref(null);
const eventsScrollRef = ref(null);
const startTime = ref(new Date().toISOString());
const items = ref([]);
const realtimeEvents = ref([]);
const expandedEvents = reactive({});
const isConnected = ref(false);
const canPushToTalk = ref(true);
const isRecording = ref(false);
const memoryKv = ref({});
const coords = ref({ lat: 37.775593, lng: -122.418137 });
const marker = ref(null);
// Methods
const formatTime = (timestamp) => {
const t0 = new Date(startTime.value).valueOf();
const t1 = new Date(timestamp).valueOf();
const delta = t1 - t0;
const hs = Math.floor(delta / 10) % 100;
const s = Math.floor(delta / 1000) % 60;
const m = Math.floor(delta / 60_000) % 60;
const pad = (n) => {
let s = n + '';
while (s.length < 2) {
s = '0' + s;
}
return s;
};
return `${pad(m)}:${pad(s)}.${pad(hs)}`;
};
const connectConversation = async () => {
alert(123)
startTime.value = new Date().toISOString();
isConnected.value = true;
realtimeEvents.value = [];
items.value = client.value.conversation.getItems();
await wavRecorder.value.begin();
await wavStreamPlayer.value.connect();
await client.value.connect();
client.value.sendUserMessageContent([
{
type: 'input_text',
text: '你好,我是老阳!',
},
]);
if (client.value.getTurnDetectionType() === 'server_vad') {
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
}
};
const disconnectConversation = async () => {
isConnected.value = false;
realtimeEvents.value = [];
items.value = [];
memoryKv.value = {};
coords.value = { lat: 37.775593, lng: -122.418137 };
marker.value = null;
client.value.disconnect();
await wavRecorder.value.end();
await wavStreamPlayer.value.interrupt();
};
const deleteConversationItem = async (id) => {
client.value.deleteItem(id);
};
const startRecording = async () => {
isRecording.value = true;
const trackSampleOffset = await wavStreamPlayer.value.interrupt();
if (trackSampleOffset?.trackId) {
const { trackId, offset } = trackSampleOffset;
await client.value.cancelResponse(trackId, offset);
}
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
};
const stopRecording = async () => {
isRecording.value = false;
await wavRecorder.value.pause();
client.value.createResponse();
};
const changeTurnEndType = async (value) => {
if (value === 'none' && wavRecorder.value.getStatus() === 'recording') {
await wavRecorder.value.pause();
}
client.value.updateSession({
turn_detection: value === 'none' ? null : { type: 'server_vad' },
});
if (value === 'server_vad' && client.value.isConnected()) {
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
}
canPushToTalk.value = value === 'none';
};
const toggleEventDetails = (eventId) => {
if (expandedEvents[eventId]) {
delete expandedEvents[eventId];
} else {
expandedEvents[eventId] = true;
}
};
// Lifecycle hooks and watchers
onMounted(() => {
if (apiKey.value !== '') {
localStorage.setItem('tmp::voice_api_key', apiKey.value);
}
// Set up render loops for the visualization canvas
let isLoaded = true;
const render = () => {
if (isLoaded) {
if (clientCanvasRef.value) {
const canvas = clientCanvasRef.value;
if (!canvas.width || !canvas.height) {
canvas.width = canvas.offsetWidth;
canvas.height = canvas.offsetHeight;
}
const ctx = canvas.getContext('2d');
if (ctx) {
ctx.clearRect(0, 0, canvas.width, canvas.height);
const result = wavRecorder.value.recording
? wavRecorder.value.getFrequencies('voice')
: { values: new Float32Array([0]) };
WavRenderer.drawBars(canvas, ctx, result.values, '#0099ff', 10, 0, 8);
}
}
if (serverCanvasRef.value) {
const canvas = serverCanvasRef.value;
if (!canvas.width || !canvas.height) {
canvas.width = canvas.offsetWidth;
canvas.height = canvas.offsetHeight;
}
const ctx = canvas.getContext('2d');
if (ctx) {
ctx.clearRect(0, 0, canvas.width, canvas.height);
const result = wavStreamPlayer.value.analyser
? wavStreamPlayer.value.getFrequencies('voice')
: { values: new Float32Array([0]) };
WavRenderer.drawBars(canvas, ctx, result.values, '#009900', 10, 0, 8);
}
}
requestAnimationFrame(render);
}
};
render();
// Set up client event listeners
client.value.on('realtime.event', (realtimeEvent) => {
realtimeEvents.value = realtimeEvents.value.slice();
const lastEvent = realtimeEvents.value[realtimeEvents.value.length - 1];
if (lastEvent?.event.type === realtimeEvent.event.type) {
lastEvent.count = (lastEvent.count || 0) + 1;
realtimeEvents.value.splice(-1, 1, lastEvent);
} else {
realtimeEvents.value.push(realtimeEvent);
}
});
client.value.on('error', (event) => console.error(event));
client.value.on('conversation.interrupted', async () => {
const trackSampleOffset = await wavStreamPlayer.value.interrupt();
if (trackSampleOffset?.trackId) {
const { trackId, offset } = trackSampleOffset;
await client.value.cancelResponse(trackId, offset);
}
});
client.value.on('conversation.updated', async ({ item, delta }) => {
items.value = client.value.conversation.getItems();
if (delta?.audio) {
wavStreamPlayer.value.add16BitPCM(delta.audio, item.id);
}
if (item.status === 'completed' && item.formatted.audio?.length) {
const wavFile = await WavRecorder.decode(
item.formatted.audio,
24000,
24000
);
item.formatted.file = wavFile;
}
});
// Set up client instructions and tools
client.value.updateSession({ instructions: instructions });
client.value.updateSession({ input_audio_transcription: { model: 'whisper-1' } });
client.value.addTool(
{
name: 'set_memory',
description: 'Saves important data about the user into memory.',
parameters: {
type: 'object',
properties: {
key: {
type: 'string',
description:
'The key of the memory value. Always use lowercase and underscores, no other characters.',
},
value: {
type: 'string',
description: 'Value can be anything represented as a string',
},
},
required: ['key', 'value'],
},
},
async ({ key, value }) => {
memoryKv.value = { ...memoryKv.value, [key]: value };
return { ok: true };
}
);
client.value.addTool(
{
name: 'get_weather',
description:
'Retrieves the weather for a given lat, lng coordinate pair. Specify a label for the location.',
parameters: {
type: 'object',
properties: {
lat: {
type: 'number',
description: 'Latitude',
},
lng: {
type: 'number',
description: 'Longitude',
},
location: {
type: 'string',
description: 'Name of the location',
},
},
required: ['lat', 'lng', 'location'],
},
},
async ({ lat, lng, location }) => {
marker.value = { lat, lng, location };
coords.value = { lat, lng, location };
const result = await fetch(
`https://api.open-meteo.com/v1/forecast?latitude=${lat}&longitude=${lng}&current=temperature_2m,wind_speed_10m`
);
const json = await result.json();
const temperature = {
value: json.current.temperature_2m,
units: json.current_units.temperature_2m,
};
const wind_speed = {
value: json.current.wind_speed_10m,
units: json.current_units.wind_speed_10m,
};
marker.value = { lat, lng, location, temperature, wind_speed };
return json;
}
);
items.value = client.value.conversation.getItems();
});
onUnmounted(() => {
client.value.reset();
});
// Watchers
watch(realtimeEvents, () => {
if (eventsScrollRef.value) {
const eventsEl = eventsScrollRef.value;
eventsEl.scrollTop = eventsEl.scrollHeight;
}
});
watch(items, () => {
const conversationEls = document.querySelectorAll('[data-conversation-content]');
conversationEls.forEach((el) => {
el.scrollTop = el.scrollHeight;
});
});
</script>
<style scoped>
/* You can add your component-specific styles here */
/* If you're using SCSS, you might want to import your existing SCSS file */
/* @import './ConsolePage.scss'; */
</style>

View File

@ -1,471 +1,40 @@
<template> <template>
<div data-component="ConsolePage"> <div class="audio-chat-page">
<div class="content-top"> <el-button style="margin: 20px" type="primary" size="large" @click="connect()">开始语音对话</el-button>
<div class="content-title">
<img src="/openai-logomark.svg" alt="OpenAI Logo" />
<span>realtime console</span>
</div>
</div> <el-dialog v-model="showDialog" title="语音通话" >
<div class="content-main"> <realtime-conversation @close="showDialog = false" ref="conversationRef" :height="dialogHeight+'px'" />
<div class="content-logs"> </el-dialog>
<div class="content-block events">
<div class="visualization">
<div class="visualization-entry client">
<canvas ref="clientCanvasRef" />
</div>
<div class="visualization-entry server">
<canvas ref="serverCanvasRef" />
</div>
</div>
<div class="content-block-title">events</div>
<div class="content-block-body" ref="eventsScrollRef">
<template v-if="!realtimeEvents.length">
awaiting connection...
</template>
<template v-else>
<div v-for="(realtimeEvent, i) in realtimeEvents" :key="realtimeEvent.event.event_id" class="event">
<div class="event-timestamp">
{{ formatTime(realtimeEvent.time) }}
</div>
<div class="event-details">
<div
class="event-summary"
@click="toggleEventDetails(realtimeEvent.event.event_id)"
>
<div
:class="[
'event-source',
realtimeEvent.event.type === 'error'
? 'error'
: realtimeEvent.source,
]"
>
<component :is="realtimeEvent.source === 'client' ? ArrowUp : ArrowDown" />
<span>
{{ realtimeEvent.event.type === 'error'
? 'error!'
: realtimeEvent.source }}
</span>
</div>
<div class="event-type">
{{ realtimeEvent.event.type }}
{{ realtimeEvent.count ? `(${realtimeEvent.count})` : '' }}
</div>
</div>
<div
v-if="expandedEvents[realtimeEvent.event.event_id]"
class="event-payload"
>
{{ JSON.stringify(realtimeEvent.event, null, 2) }}
</div>
</div>
</div>
</template>
</div>
</div>
<div class="content-block conversation">
<div class="content-block-title">conversation</div>
<div class="content-block-body" data-conversation-content>
<template v-if="!items.length">
awaiting connection...
</template>
<template v-else>
<div
v-for="(conversationItem, i) in items"
:key="conversationItem.id"
class="conversation-item"
>
<div :class="['speaker', conversationItem.role || '']">
<div>
{{
(conversationItem.role || conversationItem.type).replaceAll(
'_',
' '
)
}}
</div>
<div class="close" @click="deleteConversationItem(conversationItem.id)">
<X />
</div>
</div>
<div class="speaker-content">
<!-- tool response -->
<div v-if="conversationItem.type === 'function_call_output'">
{{ conversationItem.formatted.output }}
</div>
<!-- tool call -->
<div v-if="conversationItem.formatted.tool">
{{ conversationItem.formatted.tool.name }}(
{{ conversationItem.formatted.tool.arguments }})
</div>
<div
v-if="
!conversationItem.formatted.tool &&
conversationItem.role === 'user'
"
>
{{
conversationItem.formatted.transcript ||
(conversationItem.formatted.audio?.length
? '(awaiting transcript)'
: conversationItem.formatted.text || '(item sent)')
}}
</div>
<div
v-if="
!conversationItem.formatted.tool &&
conversationItem.role === 'assistant'
"
>
{{
conversationItem.formatted.transcript ||
conversationItem.formatted.text ||
'(truncated)'
}}
</div>
<audio
v-if="conversationItem.formatted.file"
:src="conversationItem.formatted.file.url"
controls
/>
</div>
</div>
</template>
</div>
</div>
<div class="content-actions" style="position:absolute; top: 0; left: 0">
<el-button
:type="isConnected ? '' : 'primary'"
@click="connectConversation"
>
{{isConnected ? '断开连接' : '连接对话'}}
</el-button>
<el-button @mousedown="startRecording" @mouseup="stopRecording">开始讲话</el-button>
</div>
</div>
</div>
</div> </div>
</template> </template>
<script setup> <script setup>
import { ref, reactive, onMounted, onUnmounted, watch } from 'vue'; import {nextTick, ref} from 'vue';
import { RealtimeClient } from '@openai/realtime-api-beta'; import RealtimeConversation from "@/components/RealtimeConversation .vue";
import { WavRecorder, WavStreamPlayer } from '@/lib/wavtools/index.js';
import { instructions } from '@/utils/conversation_config.js';
import { WavRenderer } from '@/utils/wav_renderer';
// Constants const showDialog = ref(false);
const LOCAL_RELAY_SERVER_URL = process.env.REACT_APP_LOCAL_RELAY_SERVER_URL || ''; const dialogHeight = ref(window.innerHeight - 75);
const conversationRef = ref(null);
// Reactive state const connect = () => {
const apiKey = ref( showDialog.value = true;
LOCAL_RELAY_SERVER_URL nextTick(() => {
? '' conversationRef.value.connect()
: localStorage.getItem('tmp::voice_api_key') || prompt('OpenAI API Key') || '' })
); }
const wavRecorder = ref(new WavRecorder({ sampleRate: 24000 }));
const wavStreamPlayer = ref(new WavStreamPlayer({ sampleRate: 24000 }));
const client = ref(
new RealtimeClient({
url: "wss://api.geekai.pro/v1/realtime",
apiKey: "sk-Gc5cEzDzGQLIqxWA9d62089350F3454bB359C4A3Fa21B3E4",
dangerouslyAllowAPIKeyInBrowser: true,
})
);
const clientCanvasRef = ref(null);
const serverCanvasRef = ref(null);
const eventsScrollRef = ref(null);
const startTime = ref(new Date().toISOString());
const items = ref([]);
const realtimeEvents = ref([]);
const expandedEvents = reactive({});
const isConnected = ref(false);
const canPushToTalk = ref(true);
const isRecording = ref(false);
const memoryKv = ref({});
const coords = ref({ lat: 37.775593, lng: -122.418137 });
const marker = ref(null);
// Methods
const formatTime = (timestamp) => {
const t0 = new Date(startTime.value).valueOf();
const t1 = new Date(timestamp).valueOf();
const delta = t1 - t0;
const hs = Math.floor(delta / 10) % 100;
const s = Math.floor(delta / 1000) % 60;
const m = Math.floor(delta / 60_000) % 60;
const pad = (n) => {
let s = n + '';
while (s.length < 2) {
s = '0' + s;
}
return s;
};
return `${pad(m)}:${pad(s)}.${pad(hs)}`;
};
const connectConversation = async () => {
alert(123)
startTime.value = new Date().toISOString();
isConnected.value = true;
realtimeEvents.value = [];
items.value = client.value.conversation.getItems();
await wavRecorder.value.begin();
await wavStreamPlayer.value.connect();
await client.value.connect();
client.value.sendUserMessageContent([
{
type: 'input_text',
text: '你好,我是老阳!',
},
]);
if (client.value.getTurnDetectionType() === 'server_vad') {
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
}
};
const disconnectConversation = async () => {
isConnected.value = false;
realtimeEvents.value = [];
items.value = [];
memoryKv.value = {};
coords.value = { lat: 37.775593, lng: -122.418137 };
marker.value = null;
client.value.disconnect();
await wavRecorder.value.end();
await wavStreamPlayer.value.interrupt();
};
const deleteConversationItem = async (id) => {
client.value.deleteItem(id);
};
const startRecording = async () => {
isRecording.value = true;
const trackSampleOffset = await wavStreamPlayer.value.interrupt();
if (trackSampleOffset?.trackId) {
const { trackId, offset } = trackSampleOffset;
await client.value.cancelResponse(trackId, offset);
}
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
};
const stopRecording = async () => {
isRecording.value = false;
await wavRecorder.value.pause();
client.value.createResponse();
};
const changeTurnEndType = async (value) => {
if (value === 'none' && wavRecorder.value.getStatus() === 'recording') {
await wavRecorder.value.pause();
}
client.value.updateSession({
turn_detection: value === 'none' ? null : { type: 'server_vad' },
});
if (value === 'server_vad' && client.value.isConnected()) {
await wavRecorder.value.record((data) => client.value.appendInputAudio(data.mono));
}
canPushToTalk.value = value === 'none';
};
const toggleEventDetails = (eventId) => {
if (expandedEvents[eventId]) {
delete expandedEvents[eventId];
} else {
expandedEvents[eventId] = true;
}
};
// Lifecycle hooks and watchers
onMounted(() => {
if (apiKey.value !== '') {
localStorage.setItem('tmp::voice_api_key', apiKey.value);
}
// Set up render loops for the visualization canvas
let isLoaded = true;
const render = () => {
if (isLoaded) {
if (clientCanvasRef.value) {
const canvas = clientCanvasRef.value;
if (!canvas.width || !canvas.height) {
canvas.width = canvas.offsetWidth;
canvas.height = canvas.offsetHeight;
}
const ctx = canvas.getContext('2d');
if (ctx) {
ctx.clearRect(0, 0, canvas.width, canvas.height);
const result = wavRecorder.value.recording
? wavRecorder.value.getFrequencies('voice')
: { values: new Float32Array([0]) };
WavRenderer.drawBars(canvas, ctx, result.values, '#0099ff', 10, 0, 8);
}
}
if (serverCanvasRef.value) {
const canvas = serverCanvasRef.value;
if (!canvas.width || !canvas.height) {
canvas.width = canvas.offsetWidth;
canvas.height = canvas.offsetHeight;
}
const ctx = canvas.getContext('2d');
if (ctx) {
ctx.clearRect(0, 0, canvas.width, canvas.height);
const result = wavStreamPlayer.value.analyser
? wavStreamPlayer.value.getFrequencies('voice')
: { values: new Float32Array([0]) };
WavRenderer.drawBars(canvas, ctx, result.values, '#009900', 10, 0, 8);
}
}
requestAnimationFrame(render);
}
};
render();
// Set up client event listeners
client.value.on('realtime.event', (realtimeEvent) => {
realtimeEvents.value = realtimeEvents.value.slice();
const lastEvent = realtimeEvents.value[realtimeEvents.value.length - 1];
if (lastEvent?.event.type === realtimeEvent.event.type) {
lastEvent.count = (lastEvent.count || 0) + 1;
realtimeEvents.value.splice(-1, 1, lastEvent);
} else {
realtimeEvents.value.push(realtimeEvent);
}
});
client.value.on('error', (event) => console.error(event));
client.value.on('conversation.interrupted', async () => {
const trackSampleOffset = await wavStreamPlayer.value.interrupt();
if (trackSampleOffset?.trackId) {
const { trackId, offset } = trackSampleOffset;
await client.value.cancelResponse(trackId, offset);
}
});
client.value.on('conversation.updated', async ({ item, delta }) => {
items.value = client.value.conversation.getItems();
if (delta?.audio) {
wavStreamPlayer.value.add16BitPCM(delta.audio, item.id);
}
if (item.status === 'completed' && item.formatted.audio?.length) {
const wavFile = await WavRecorder.decode(
item.formatted.audio,
24000,
24000
);
item.formatted.file = wavFile;
}
});
// Set up client instructions and tools
client.value.updateSession({ instructions: instructions });
client.value.updateSession({ input_audio_transcription: { model: 'whisper-1' } });
client.value.addTool(
{
name: 'set_memory',
description: 'Saves important data about the user into memory.',
parameters: {
type: 'object',
properties: {
key: {
type: 'string',
description:
'The key of the memory value. Always use lowercase and underscores, no other characters.',
},
value: {
type: 'string',
description: 'Value can be anything represented as a string',
},
},
required: ['key', 'value'],
},
},
async ({ key, value }) => {
memoryKv.value = { ...memoryKv.value, [key]: value };
return { ok: true };
}
);
client.value.addTool(
{
name: 'get_weather',
description:
'Retrieves the weather for a given lat, lng coordinate pair. Specify a label for the location.',
parameters: {
type: 'object',
properties: {
lat: {
type: 'number',
description: 'Latitude',
},
lng: {
type: 'number',
description: 'Longitude',
},
location: {
type: 'string',
description: 'Name of the location',
},
},
required: ['lat', 'lng', 'location'],
},
},
async ({ lat, lng, location }) => {
marker.value = { lat, lng, location };
coords.value = { lat, lng, location };
const result = await fetch(
`https://api.open-meteo.com/v1/forecast?latitude=${lat}&longitude=${lng}&current=temperature_2m,wind_speed_10m`
);
const json = await result.json();
const temperature = {
value: json.current.temperature_2m,
units: json.current_units.temperature_2m,
};
const wind_speed = {
value: json.current.wind_speed_10m,
units: json.current_units.wind_speed_10m,
};
marker.value = { lat, lng, location, temperature, wind_speed };
return json;
}
);
items.value = client.value.conversation.getItems();
});
onUnmounted(() => {
client.value.reset();
});
// Watchers
watch(realtimeEvents, () => {
if (eventsScrollRef.value) {
const eventsEl = eventsScrollRef.value;
eventsEl.scrollTop = eventsEl.scrollHeight;
}
});
watch(items, () => {
const conversationEls = document.querySelectorAll('[data-conversation-content]');
conversationEls.forEach((el) => {
el.scrollTop = el.scrollHeight;
});
});
</script> </script>
<style scoped> <style scoped lang="stylus">
/* You can add your component-specific styles here */ .audio-chat-page {
/* If you're using SCSS, you might want to import your existing SCSS file */ display flex
/* @import './ConsolePage.scss'; */ flex-flow column
justify-content center
align-items center
}
canvas {
background-color: transparent;
}
</style> </style>

View File

@ -15,7 +15,7 @@
import {ref} from 'vue'; import {ref} from 'vue';
import { RealtimeClient } from '@openai/realtime-api-beta'; import { RealtimeClient } from '@openai/realtime-api-beta';
import Calling from "@/components/Calling.vue"; import Calling from "@/components/Calling.vue";
import Conversation from "@/components/Conversation .vue"; import Conversation from "@/components/RealtimeConversation .vue";
import {playPCM16} from "@/utils/wav_player"; import {playPCM16} from "@/utils/wav_player";
import {showMessageError} from "@/utils/dialog"; import {showMessageError} from "@/utils/dialog";