1## 访问BCE国内节点启动命令
2BCE_APP_AK=your_ak \
3BCE_APP_SK=your_sk \
4nohup java -Djava.net.preferIPv4Stack=true -jar rtc-aiagent-userserver-version.jar \
5--server.address=0.0.0.0 > app.log 2>&1 &

1## 访问BCE香港节点启动命令
2BCE_APP_AK=your_ak \
3BCE_APP_SK=your_sk \
4BCE_ENDPOINT=http://rtc-aiagent-out.baidubce.com \
5nohup java -Djava.net.preferIPv4Stack=true -jar rtc-aiagent-userserver-version.jar \
6--server.address=0.0.0.0 > app.log 2>&1 &

1curl --location 'http://localhost:8936/api/v1/aiagent/generateAIAgentCall' \
2--header 'Content-Type: application/json' \
3--data '{
4    "config": "",
5    "app_id": "your_app_id"
6}'

1curl --location 'http://localhost:8936/api/v1/aiagent/stopAIAgentInstance' \
2--header 'Content-Type: application/json' \
3--data '{
4  "app_id": "your_app_id",
5  "ai_agent_instance_id": 2568137789341696
6}'

1# 查看占用端口的进程Id
2lsof -i:8936
3
4kill -9 [进程id]

1#include <stdio.h>
2#include <stdlib.h>
3#include "mongoose.h"
4
5#define debug printf
6
7struct FConnection
8{
9    char *url;
10};
11
12int main(int argc, char* argv[])
13{
14    struct mg_mgr mgr;        // Event manager
15    struct FConnection C;     // object
16
17    if (argc != 2) {
18        debug("Help: \n ./realtime 'ws://rtc-aiotgw.exp.bcelive.com/v1/realtime?a=APPID_GET_FROM_BAIDU&ak=AK&sk=SK'\n\n");
19        return -1;
20    }
21
22    // mg_log_set("4");
23    mg_mgr_init(&mgr);        // Initialise event manager
24
25    C.url = argv[1]; // "ws://rtc-aiotgw.exp.bcelive.com/v1/realtime?a=APPID_GET_FROM_BAIDU&ak=AK&sk=SK";
26    struct mg_connection * mg_c = mg_ws_connect(&mgr, C.url, &FConnection_cb, &C, NULL);
27
28    int cnt = 0;
29    char *question = "[T]:你是谁?";
30
31    while (1) {
32        mg_mgr_poll(&mgr, 50);
33
34		if (cnt++ % 200 == 40)  {
35            debug("tx: Question: %s\n", question);
36            mg_ws_send(mg_c, question, strlen(question), WEBSOCKET_OP_TEXT);
37        }
38    }
39
40    return 0;
41}
42
43
44static void FConnection_cb(struct mg_connection *c, int ev, void *ev_data, void *fn_data) {
45    struct FConnection* conn = (struct FConnection*) fn_data;
46    if (conn) {
47        if (ev == MG_EV_WS_OPEN) {
48            debug("MG_EV_WS_OPEN.\n");
49        } else if (ev == MG_EV_CONNECT) {
50            struct mg_str host = mg_url_host(conn->url);
51            debug("MG_EV_CONNECT: [%.*s]\n", host.len, host.ptr);
52            // If s_url is https://, tell client connection to use TLS
53            if (mg_url_is_ssl(conn->url)) {
54                struct mg_tls_opts opts = {.ca = "ca_for_tls", .srvname = host};
55                mg_tls_init(c, &opts);
56            }
57        } else if (ev == MG_EV_WS_MSG) {
58            struct mg_ws_message *wm = (struct mg_ws_message *) ev_data;
59            // debug("GOT ECHO REPLY: %d, [%.*s]\n", wm->data.len, (int)wm->data.len, wm->data.ptr);
60
61            switch (wm->flags & WEBSOCKET_FLAGS_MASK_OP) {
62                case WEBSOCKET_OP_TEXT:
63                    debug("rx: Text: %.*s\n", (int)wm->data.len, wm->data.ptr);
64                    if (strncmp("[E]:[LIC]:[MUST]", wm->data.ptr, 16) == 0) {
65                        char *lic = "[E]:[LIC]:[ACTIVE]:{\"devId\":\"DDD\",\"uId\":\"YYY\",\"licKey\":\"xxx\"}";
66                        debug("tx: do Active: %s\n", lic);
67                        mg_ws_send(c, lic, strlen(lic), WEBSOCKET_OP_TEXT);
68                    }
69                break;
70                case WEBSOCKET_OP_BINARY:
71                    debug("rx: Binary len: %d\n", (int)wm->data.len);
72                    // todo, binary data
73                break;
74            }
75        } else if (ev == MG_EV_CLOSE) {
76            debug("MG_EV_CLOSE, conn:%p\n", conn);
77        } else if (ev == MG_EV_ERROR) {
78            debug("MG_EV_ERROR, conn:%p\n", conn);
79        }
80
81        if (ev == MG_EV_ERROR || ev == MG_EV_CLOSE) {
82            debug("close done, conn:%p\n", conn);
83            //fn_data = NULL;  // Signal that we're done
84        }
85    }
86}

1# Connect online Server must with ak,sk
2# Get ak,sk from here: https://console.bce.baidu.com/iam/#/iam/accesslist
3cd src
4make linux
5./realtime 'ws://rtc-aiotgw.exp.bcelive.com/v1/realtime?a=APPID_GET_FROM_BAIDU&ak=AK&sk=SK'

1package com.baidu.rtc.realtime;
2
3import okhttp3.OkHttpClient;
4import okhttp3.Request;
5import okhttp3.Response;
6import okhttp3.WebSocket;
7import okhttp3.WebSocketListener;
8import okio.ByteString;
9
10import org.json.JSONObject;
11
12import java.nio.ByteBuffer;
13import java.util.concurrent.TimeUnit;
14import java.util.logging.Level;
15import java.util.logging.Logger;
16
17public class Realtime {
18    private static Logger logger = Logger.getLogger("REALTIME");
19    private static Realtime sInstance;
20    private static WebSocket globalWebSocket;
21
22    private volatile boolean isWebSocketClosed = false;
23
24    private Realtime() {
25        // Private constructor for singleton
26    }
27
28    public static Realtime getInstance() {
29        if (sInstance == null) {
30            synchronized(Realtime.class) {
31                if (sInstance == null) {
32                    sInstance = new Realtime();
33                }
34            }
35        }
36        return sInstance;
37    }
38
39    private void setupWebSocket(String url) {
40        OkHttpClient client = new OkHttpClient.Builder()
41                .connectTimeout(5, TimeUnit.SECONDS)
42                .build();
43
44        logger.info("Initializing WebSocket connection with URL: " + url);
45
46        try {
47            Request request = new Request.Builder().url(url).build();
48            globalWebSocket = client.newWebSocket(request, new WebSocketListenerImpl());
49            client.dispatcher().executorService().shutdown();
50        } catch (Exception e) {
51            logger.log(Level.SEVERE, "Failed to initialize WebSocket connection: " + e.getMessage(),
52                    e);
53            throw new RuntimeException("WebSocket initialization failed", e);
54        }
55    }
56
57    private void sendAudioFrame(ByteBuffer buffer) {
58        logger.fine("Sending audio frame, size: " + buffer.remaining());
59        globalWebSocket.send(ByteString.of(buffer));
60    }
61
62    public static void main(String []args){
63        Realtime rt = getInstance();
64
65        // https://cloud.baidu.com/doc/RTC/s/Jmakuvimy
66        // wss://rtc-aiotgw.exp.bcelive.com/v1/realtime?a=APPID&ak=AK&sk=SK&ac=raw16k
67        // OR wss://rtc-aiotgw.exp.bcelive.com/v1/realtime?a=APPID&id=YYYY&t=ZZZZ&ac=raw16k
68        rt.setupWebSocket("");
69    }
70
71    private class WebSocketListenerImpl extends WebSocketListener {
72        @Override
73        public void onOpen(WebSocket webSocket, Response response) {
74            super.onOpen(webSocket, response);
75            logger.info("WebSocket connection established.");
76            isWebSocketClosed = false;
77        }
78
79        @Override
80        public void onMessage(WebSocket webSocket, String msg) {
81            super.onMessage(webSocket, msg);
82            // 这里千万别阻塞，包括WebSocketListener其它回调
83            logger.info("WebSocket onMessage:" + msg);
84            if (msg.startsWith("[E]:[MEDIA]:[READY]")) {
85                webSocket.send("[T]:你是谁？");
86            } else if (msg.startsWith("[E]:[LIC]:[MUST]")) {
87                JSONObject req = new JSONObject();
88                req.put("devId", "DEVID1234567890");
89                req.put("uId", "1234567890");
90                req.put("licKey", "LICKEYXXXX");
91                webSocket.send("[E]:[LIC]:[ACTIVE]:" + req.toString());
92            }
93        }
94
95        @Override
96        public void onMessage(WebSocket webSocket, ByteString bytes) {
97            super.onMessage(webSocket, bytes);
98            logger.info("WebSocket onMessage. Binary len:" + bytes.size());
99        }
100
101        @Override
102        public void onClosing(WebSocket webSocket, int code, String reason) {
103            super.onClosing(webSocket, code, reason);
104            logger.info("WebSocket is closing: " + code + ", reason: " + reason);
105            isWebSocketClosed = true;
106        }
107
108        @Override
109        public void onClosed(WebSocket webSocket, int code, String reason) {
110            super.onClosed(webSocket, code, reason);
111            logger.info("WebSocket closed: " + code + ", reason: " + reason);
112            isWebSocketClosed = true;
113        }
114
115        @Override
116        public void onFailure(WebSocket webSocket, Throwable t, Response response) {
117            super.onFailure(webSocket, t, response);
118            logger.log(Level.SEVERE, "WebSocket failure: " + t.getMessage(), t);
119            isWebSocketClosed = true;
120        }
121    }
122}

1#!/usr/bin/env python
2"""Chat using realtime API and SoundDevice."""
3# -*- coding: utf-8 -*-
4
5import asyncio
6import json
7import sys
8import sounddevice as sd
9from websockets.asyncio.client import connect
10
11# =================配置区域=================
12# 百度 RTC 参数
13RATE = 16000
14CHANNELS = 1
15DTYPE = 'int16'  # 16位深度
16CHUNK = 3200  # 每次读写的数据块大小
17
18# 授权信息 (替换为真实Key)
19LIC_KEY = "lickey"
20DEV_ID = "myDev"
21USER_ID = "User01"
22
23speaking = False
24
25# ==========================================
26
27async def send_audio_loop(websocket, stream_in, loop):
28   """发送端：使用 sounddevice 读取麦克风"""
29   print(">>> 正在监听麦克风...")
30
31   def read_mic():
32       # sounddevice 的 read 是阻塞的，读取指定数量的帧
33       # 返回值是 (data, overflow_flag)
34       data, _ = stream_in.read(CHUNK)
35       return bytes(data)  # 转为 bytes 发送
36
37   try:
38       while True:
39           # 放到线程池运行，避免阻塞
40           pcm_bytes = await loop.run_in_executor(None, read_mic)
41           if speaking:
42               continue
43           await websocket.send(pcm_bytes)
44   except Exception as e:
45       print(f"发送循环停止: {e}")
46
47
48async def receive_loop(websocket, stream_out, loop):
49   """接收端"""
50   print(">>> 准备接收回复...")
51   
52   global speaking
53
54   def play_pcm(data):
55       stream_out.write(data)
56
57   try:
58       while True:
59           msg = await websocket.recv()
60
61           if isinstance(msg, str):
62               # 1. 鉴权
63               if msg.startswith("[E]:[LIC]:[MUST]"):
64                   print("[系统] 发送授权...")
65                   lic = f'[E]:[LIC]:[ACTIVE]:{{"devId":"{DEV_ID}","uId":"{USER_ID}","licKey":"{LIC_KEY}"}}'
66                   await websocket.send(lic)
67
68               # 2. 识别到的用户语音
69               elif msg.startswith("[Q]:"):
70                   if not msg.startswith("[Q]:[M]:"):
71                       print(f"--------\n问: {msg[4:]}")
72
73               # 3. AI 的回答文本
74               elif msg.startswith("[A]:"):
75                   print(f"AI答: {msg[4:]}")
76
77               # 4. 【关键】识别到的意图/Function Call
78               # 这里的格式通常是 [F]:[C]:{json数据}
79               elif msg.startswith("[F]:[C]:"):
80                   # 去掉前缀 [F]:[C]: 拿到后面的 JSON 字符串
81                   json_str = msg[8:]
82                   try:
83                       data = json.loads(json_str)
84                       # 打印原始数据方便调试
85                       print(f"\n★ 命中意图 (Function Call) ★")
86                       print(f"原始数据: {json_str}")
87
88                       # 解析具体字段 (根据你的截图配置)
89                       if "fname" in data and data["fname"] == "add_clock":
90                           params = data.get("param", {})
91                           time_val = params.get("time", "未知时间")
92                           date_val = params.get("date", "未知日期")
93                           print(f"指令: 添加闹钟")
94                           print(f" -> 时间: {time_val}")
95                           print(f" -> 日期: {date_val}")
96
97                           # 在这里可以写代码真的去设置电脑闹钟
98                           # ...
99
100                   except json.JSONDecodeError:
101                       print(f"收到FC指令但解析JSON失败: {json_str}")
102                       
103               elif msg.startswith('[E]:[TTS_BEGIN_SPEAKING]'):
104                   speaking = True
105               elif msg.startswith('[E]:[TTS_END_SPEAKING]'):
106                   speaking = False
107               else :
108                   print(f"RX: {msg}")
109
110           elif isinstance(msg, bytes):
111               if len(msg) > 0:
112                   await loop.run_in_executor(None, play_pcm, msg)
113
114   except Exception as e:
115       print(f"接收循环停止: {e}")
116
117async def run_chat(uri):
118   """Chat Main."""
119   print("正在打开音频设备 (SoundDevice)...")
120
121   # 打开原始输入流 (RawInputStream) 处理 bytes
122   # 打开原始输出流 (RawOutputStream) 处理 bytes
123   with sd.RawInputStream(samplerate=RATE, channels=CHANNELS, dtype=DTYPE, blocksize=CHUNK) as stream_in, \
124           sd.RawOutputStream(samplerate=RATE, channels=CHANNELS, dtype=DTYPE, blocksize=CHUNK) as stream_out:
125       print("音频设备就绪。")
126       loop = asyncio.get_running_loop()
127
128       async with connect(uri) as websocket:
129           print(">> 已连接到", uri)
130
131           send_task = asyncio.create_task(send_audio_loop(websocket, stream_in, loop))
132           recv_task = asyncio.create_task(receive_loop(websocket, stream_out, loop))
133
134           done, pending = await asyncio.wait([send_task, recv_task], return_when=asyncio.FIRST_COMPLETED)
135           for t in pending:
136               t.cancel()
137
138
139if __name__ == "__main__":
140   try:
141       uri = "ws://rtc-aiotgw.exp.bcelive.com/v1/realtime?a=APPID&ak=AK&sk=SK"
142       if len(sys.argv) > 1:
143           uri = sys.argv[1]
144
145       if sys.platform == 'win32':
146           asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
147
148       asyncio.run(run_chat(uri))
149   except KeyboardInterrupt:
150       print("\n已退出")
151   except Exception as e:
152       print(f"\n发生错误: {e}")

1package main
2
3import (
4	"context"
5	"flag"
6	"io"
7	"log"
8	"os"
9	"os/signal"
10	"strings"
11	"time"
12
13	"github.com/gorilla/websocket"
14)
15
16var uri string
17var ac string = "pcmu"
18var appid string
19var id string
20var token string
21
22func init() {
23   flag.StringVar(&appid, "a", "", "AppID")
24   flag.StringVar(&id, "id", "", "ID")
25   flag.StringVar(&token, "t", "", "token")
26   flag.StringVar(&uri, "s", "wss://rtc-aiotgw.exp.bcelive.com/v1/realtime", "ws service address")
27}
28
29type myListener struct {
30   wsCon                 *websocket.Conn
31   aw                    io.Writer
32   PubConnectedCtx       context.Context
33   PubConnectedCtxCancel context.CancelFunc
34   agentid               string
35}
36
37func (c *myListener) OnAudioData(audio []byte) {
38   if c.aw != nil {
39       c.aw.Write(audio)
40   }
41}
42
43func PrepareSendMedia(c *websocket.Conn, l *myListener, afn string) {
44   var (
45       audioFileName     = afn
46   )
47
48   _, err := os.Stat(audioFileName)
49   haveAudioFile := !os.IsNotExist(err)
50
51   if !haveAudioFile {
52       log.Println("Could not find `" + audioFileName)
53       return
54   }
55
56   l.PubConnectedCtx , l.PubConnectedCtxCancel = context.WithCancel(context.Background())
57
58   if haveAudioFile {
59       if strings.HasSuffix(afn, ".pcmu"){
60           go func() {
61               // Open a file and start reading
62               file, Err := os.Open(audioFileName)
63               if Err != nil {
64                   return
65               }
66
67               // Wait for connection established
68               <-l.PubConnectedCtx.Done()
69
70               fi, _ := os.Stat(audioFileName)
71
72               payload := make([]byte, fi.Size())
73               _, Err = io.ReadFull(file, payload)
74               if  Err != nil {
75                   return
76               }
77               var Pos = 0;
78
79               time.Sleep(1500 * time.Millisecond)
80               ticker := time.NewTicker(time.Millisecond * 20)
81               for ; true; <-ticker.C {
82
83                   c.WriteMessage(websocket.BinaryMessage, payload[Pos:Pos + 160])
84                   Pos += 160;
85
86                   if Pos + 160 > int(fi.Size()) {
87                       Pos = 0
88                       time.Sleep(8 * time.Second)
89                   }
90               }
91           }()
92           return
93       }
94   }
95}
96
97func main() {
98   flag.Parse()
99   log.SetFlags(log.LstdFlags | log.Lmicroseconds)
100
101   interrupt := make(chan os.Signal, 1)
102   signal.Notify(interrupt, os.Interrupt)
103
104   var splitChar = "?"
105   if strings.Contains(uri, "?") {
106       splitChar = "&"
107   }
108   if appid != "" {
109       uri +=  splitChar + "a=" + appid
110       splitChar = "&"
111   }
112   if id != "" {
113       uri +=  splitChar + "id=" + id
114       splitChar = "&"
115   }
116   if token != "" {
117       uri +=  splitChar + "t=" + token
118       splitChar = "&"
119   }
120
121   log.Println("Connecting uri:", uri)
122   c, _, err := websocket.DefaultDialer.Dial(uri, nil)
123   if err != nil {
124       log.Fatal("dial:", err)
125   }
126   defer c.Close()
127
128   done := make(chan struct{})
129
130   // ffplay -f mulaw -ar 8000 dump_audio.pcmu
131   af, _ := os.Create("dump_audio." + ac)
132   my := &myListener{wsCon: c, aw:af}
133   afn := "myaudio." + ac
134   PrepareSendMedia(c, my, afn)
135
136   go func() {
137       defer close(done)
138       for {
139           mt, message, err := c.ReadMessage()
140           if err != nil {
141               log.Println("read:", err)
142               return
143           }
144           if mt == websocket.TextMessage {
145               log.Printf("rx: %s", message)
146               if strings.HasPrefix(string(message), "[E]:[LIC]:[MUST]") {
147                   lic := `[E]:[LIC]:[ACTIVE]:{"devId":"DDD","uId":"YYY","licKey":"xxx"}`
148                   log.Printf("tx: do Active: %s\n", lic)
149                   c.WriteMessage(websocket.TextMessage, []byte(lic))
150               }
151           } else if mt == websocket.BinaryMessage {
152               my.OnAudioData(message)
153           }
154       }
155   }()
156
157   if my.PubConnectedCtxCancel != nil {
158       my.PubConnectedCtxCancel()
159   }
160
161   ticker := time.NewTicker(120 * time.Second)
162   defer ticker.Stop()
163
164   for {
165       select {
166       case <-done:
167           return
168       case t := <-ticker.C:
169           log.Println(t.String())
170           // c.WriteMessage(websocket.TextMessage, []byte(`[TTS]:你是谁？`))
171       case <-interrupt:
172           log.Println("interrupt")
173
174           // Cleanly close the connection by sending a close message and then
175           // waiting (with timeout) for the server to close the connection.
176           err := c.WriteMessage(websocket.CloseMessage, websocket.FormatCloseMessage(websocket.CloseNormalClosure, ""))
177           if err != nil {
178               log.Println("write close:", err)
179               return
180           }
181           select {
182           case <-done:
183           case <-time.After(time.Second):
184           }
185           return
186       }
187   }
188}

1# 调整增益接口
2static int create_rec_pipeline(void)
3{
4...
5    es7210_adc_set_gain(ES7210_INPUT_MIC3, GAIN_0DB);
6    es7210_adc_set_gain(ES7210_INPUT_MIC2 | ES7210_INPUT_MIC1, GAIN_33DB);  //调整增益接口函数，一般设置GAIN_33DB
7...
8}

1# 调整增益接口
2static int create_rec_pipeline(void)
3{
4...
5    const char *link_tag[] = { "aec", "raw"};
6    audio_pipeline_register(rec_res.rec_pipe, rec_res.aec_stream , link_tag[0]);  //打开AEC接口函数
7    audio_pipeline_register(rec_res.rec_pipe, rec_res.raw_stream , link_tag[1]);
8...
9}

1  - 在服务器上调用generateAIAgentCall接口，传入config参数`sceneRole:人设名称`。 人设名称就是控制台定义的人设名称。