azure-communication-callautomation-java

Azure Communication 呼叫自动化（Java）

构建服务端呼叫自动化工作流，包括 IVR 系统、呼叫路由、录音和 AI 驱动的交互。

集成边界

示例是针对指定 SDK 版本的代码片段，并非完整的 IVR。编译前，请核对该版本中的构造函数和类型。外呼和呼叫转接会产生费用，且需要使用获批的接收方；录音需要适当告知并征得同意，同时妥善保护录音数据。对 Webhook 投递进行身份验证，验证事件架构，去重事件 ID 和操作 ID，并避免记录 DTMF 或账户数据。异步请求已受理，并不表示播放或识别已成功：请处理失败和断开连接事件。

安装

<dependency>
    <groupId>com.azure</groupId>
    <artifactId>azure-communication-callautomation</artifactId>
    <version>1.6.0</version>
</dependency>

创建客户端

import com.azure.communication.callautomation.CallAutomationClient;
import com.azure.communication.callautomation.CallAutomationClientBuilder;
import com.azure.identity.DefaultAzureCredentialBuilder;

// With DefaultAzureCredential
CallAutomationClient client = new CallAutomationClientBuilder()
    .endpoint("https://<resource>.communication.azure.com")
    .credential(new DefaultAzureCredentialBuilder().build())
    .buildClient();

// With connection string
CallAutomationClient client = new CallAutomationClientBuilder()
    .connectionString("<connection-string>")
    .buildClient();

核心概念

类	用途
CallAutomationClient	发起呼叫、接听或拒绝来电、重定向呼叫
CallConnection	已建立呼叫中的操作（添加参与者、终止呼叫）
CallMedia	媒体操作（播放音频、识别 DTMF/语音）
CallRecording	开始、停止或暂停录音
CallAutomationEventParser	解析来自 ACS 的 Webhook 事件

创建外呼

import com.azure.communication.callautomation.models.*;
import com.azure.communication.common.CommunicationUserIdentifier;
import com.azure.communication.common.PhoneNumberIdentifier;

// Call to PSTN number
PhoneNumberIdentifier target = new PhoneNumberIdentifier("+14255551234");
PhoneNumberIdentifier caller = new PhoneNumberIdentifier("+14255550100");

CreateCallOptions options = new CreateCallOptions(
    new CommunicationUserIdentifier("<user-id>"),  // Source
    List.of(target))                                // Targets
    .setSourceCallerId(caller)
    .setCallbackUrl("https://your-app.com/api/callbacks");

CreateCallResult result = client.createCall(options);
String callConnectionId = result.getCallConnectionProperties().getCallConnectionId();

接听来电

// From Event Grid webhook - IncomingCall event
String incomingCallContext = "<incoming-call-context-from-event>";

AnswerCallOptions options = new AnswerCallOptions(
    incomingCallContext,
    "https://your-app.com/api/callbacks");

AnswerCallResult result = client.answerCall(options);
CallConnection callConnection = result.getCallConnection();

播放音频（文本转语音）

CallConnection callConnection = client.getCallConnection(callConnectionId);
CallMedia callMedia = callConnection.getCallMedia();

// Play text-to-speech
TextSource textSource = new TextSource()
    .setText("Welcome to Contoso. Press 1 for sales, 2 for support.")
    .setVoiceName("en-US-JennyNeural");

PlayOptions playOptions = new PlayOptions(
    List.of(textSource),
    List.of(new CommunicationUserIdentifier("<target-user>")));

callMedia.play(playOptions);

// Play audio file
FileSource fileSource = new FileSource()
    .setUrl("https://storage.blob.core.windows.net/audio/greeting.wav");

callMedia.play(new PlayOptions(List.of(fileSource), List.of(target)));

识别 DTMF 输入

// Recognize DTMF tones
DtmfTone stopTones = DtmfTone.POUND;

CallMediaRecognizeDtmfOptions recognizeOptions = new CallMediaRecognizeDtmfOptions(
    new CommunicationUserIdentifier("<target-user>"),
    5)  // Max tones to collect
    .setInterToneTimeout(Duration.ofSeconds(5))
    .setStopTones(List.of(stopTones))
    .setInitialSilenceTimeout(Duration.ofSeconds(15))
    .setPlayPrompt(new TextSource().setText("Enter your account number followed by pound."));

callMedia.startRecognizing(recognizeOptions);

识别语音

// Speech recognition with AI
CallMediaRecognizeSpeechOptions speechOptions = new CallMediaRecognizeSpeechOptions(
    new CommunicationUserIdentifier("<target-user>"))
    .setEndSilenceTimeout(Duration.ofSeconds(2))
    .setSpeechLanguage("en-US")
    .setPlayPrompt(new TextSource().setText("How can I help you today?"));

callMedia.startRecognizing(speechOptions);

呼叫录音

CallRecording callRecording = client.getCallRecording();

// Start recording
StartRecordingOptions recordingOptions = new StartRecordingOptions(
    new ServerCallLocator("<server-call-id>"))
    .setRecordingChannel(RecordingChannel.MIXED)
    .setRecordingContent(RecordingContent.AUDIO_VIDEO)
    .setRecordingFormat(RecordingFormat.MP4);

RecordingStateResult recordingResult = callRecording.start(recordingOptions);
String recordingId = recordingResult.getRecordingId();

// Pause/resume/stop
callRecording.pause(recordingId);
callRecording.resume(recordingId);
callRecording.stop(recordingId);

// Download recording (after RecordingFileStatusUpdated event)
callRecording.downloadTo(recordingUrl, Paths.get("recording.mp4"));

向呼叫中添加参与者

CallConnection callConnection = client.getCallConnection(callConnectionId);

CommunicationUserIdentifier participant = new CommunicationUserIdentifier("<user-id>");
AddParticipantOptions addOptions = new AddParticipantOptions(participant)
    .setInvitationTimeout(Duration.ofSeconds(30));

AddParticipantResult result = callConnection.addParticipant(addOptions);

转接呼叫

// Blind transfer
PhoneNumberIdentifier transferTarget = new PhoneNumberIdentifier("+14255559999");
TransferCallToParticipantResult result = callConnection.transferCallToParticipant(transferTarget);

处理事件（Webhook）

import com.azure.communication.callautomation.CallAutomationEventParser;
import com.azure.communication.callautomation.models.events.*;

// In your webhook endpoint
public void handleCallback(String requestBody) {
    List<CallAutomationEventBase> events = CallAutomationEventParser.parseEvents(requestBody);
    
    for (CallAutomationEventBase event : events) {
        if (event instanceof CallConnected) {
            CallConnected connected = (CallConnected) event;
            System.out.println("Call connected: " + connected.getCallConnectionId());
        } else if (event instanceof RecognizeCompleted) {
            RecognizeCompleted recognized = (RecognizeCompleted) event;
            // Handle DTMF or speech recognition result
            if (!(recognized.getRecognizeResult() instanceof DtmfResult)) {
                // Handle speech results in a separate branch.
                continue;
            }
            DtmfResult dtmfResult = (DtmfResult) recognized.getRecognizeResult();
            String tones = dtmfResult.getTones().stream()
                .map(DtmfTone::toString)
                .collect(Collectors.joining());
            System.out.println("DTMF received: " + tones);
        } else if (event instanceof PlayCompleted) {
            System.out.println("Audio playback completed");
        } else if (event instanceof CallDisconnected) {
            System.out.println("Call ended");
        }
    }
}

挂断呼叫

// Hang up for all participants
callConnection.hangUp(true);

// Hang up only this leg
callConnection.hangUp(false);

错误处理

import com.azure.core.exception.HttpResponseException;

try {
    client.answerCall(options);
} catch (HttpResponseException e) {
    if (e.getResponse().getStatusCode() == 404) {
        System.out.println("Call not found or already ended");
    } else if (e.getResponse().getStatusCode() == 400) {
        System.out.println("Invalid request: " + e.getMessage());
    }
}

环境变量

AZURE_COMMUNICATION_ENDPOINT=https://<resource>.communication.azure.com
AZURE_COMMUNICATION_CONNECTION_STRING=endpoint=https://...;accesskey=...
CALLBACK_BASE_URL=https://your-app.com/api/callbacks

触发短语

• “call automation Java”、“IVR Java”、“interactive voice response”
• “call recording Java”、“DTMF recognition Java”
• “text to speech call”、“speech recognition call”
• “answer incoming call”、“transfer call Java”
• “Azure Communication Services call automation”

适用场景

此技能适用于执行概述中所述的工作流或操作。

限制

• 仅当任务明确符合上述范围时，才使用此技能。
• 不要将输出视为环境特定验证、测试或专家审查的替代品。
• 如果缺少必要的输入、权限、安全边界或成功标准，请停止并请求澄清。