Outbound call phone_recording_url

Haiping Chen · Haiping Chen · commit addb1b1bba10 · 2025-03-14T15:08:34.000-05:00
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/Controllers/TwilioRecordController.cs b/src/Plugins/BotSharp.Plugin.Twilio/Controllers/TwilioRecordController.cs
@@ -0,0 +1,44 @@
+using BotSharp.Core.Infrastructures;
+using BotSharp.Plugin.Twilio.Interfaces;
+using BotSharp.Plugin.Twilio.Models;
+using Microsoft.AspNetCore.Http;
+using Microsoft.AspNetCore.Mvc;
+
+namespace BotSharp.Plugin.Twilio.Controllers;
+
+public class TwilioRecordController : TwilioController
+{
+    private readonly TwilioSetting _settings;
+    private readonly IServiceProvider _services;
+    private readonly ILogger _logger;
+
+    public TwilioRecordController(TwilioSetting settings, IServiceProvider services, IHttpContextAccessor context, ILogger<TwilioRecordController> logger)
+    {
+        _settings = settings;
+        _services = services;
+        _logger = logger;
+    }
+
+    [ValidateRequest]
+    [HttpPost("twilio/record/status")]
+    public async Task<ActionResult> PhoneRecordingStatus(ConversationalVoiceRequest request)
+    {
+        if (request.RecordingStatus == "completed")
+        {
+            _logger.LogInformation($"Recording completed for {request.CallSid}, the record URL is {request.RecordingUrl}");
+
+            // Set the recording URL to the conversation state
+            var convService = _services.GetRequiredService<IConversationService>();
+            convService.SetConversationId(request.ConversationId, new List<MessageState>
+            {
+                new("phone_recording_url", request.RecordingUrl)
+            });
+            convService.SaveStates();
+
+            // recording completed
+            await HookEmitter.Emit<ITwilioCallStatusHook>(_services, x => x.OnRecordingCompleted(request));
+        }
+
+        return Ok();
+    }
+}
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/Controllers/TwilioStreamController.cs b/src/Plugins/BotSharp.Plugin.Twilio/Controllers/TwilioStreamController.cs
@@ -48,6 +48,7 @@ public async Task<TwiMLResult> InitiateStreamConversation(ConversationalVoiceReq
 
         var instruction = new ConversationalVoiceResponse
         {
+            ConversationId = request.ConversationId,
             SpeechPaths = [],
             ActionOnEmptyResult = true
         };
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/Controllers/TwilioVoiceController.cs b/src/Plugins/BotSharp.Plugin.Twilio/Controllers/TwilioVoiceController.cs
@@ -52,6 +52,7 @@ public async Task<TwiMLResult> InitiateConversation(ConversationalVoiceRequest r
         VoiceResponse response = null;
         var instruction = new ConversationalVoiceResponse
         {
+            ConversationId = request.ConversationId,
             SpeechPaths = ["twilio/welcome.mp3"],
             ActionOnEmptyResult = true
         };
@@ -170,6 +171,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>
             {
                 var instruction = new ConversationalVoiceResponse
                 {
+                    ConversationId = request.ConversationId,
                     SpeechPaths = new List<string>(),
                     CallbackPath = $"twilio/voice/receive/{request.SeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}&attempts={++request.Attempts}",
                     ActionOnEmptyResult = true
@@ -273,6 +275,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>
 
                 var instruction = new ConversationalVoiceResponse
                 {
+                    ConversationId = request.ConversationId,
                     SpeechPaths = speechPaths,
                     CallbackPath = $"twilio/voice/reply/{request.SeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}&AIResponseWaitTime={++request.AIResponseWaitTime}",
                     ActionOnEmptyResult = true
@@ -312,6 +315,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>
 
                 var instruction = new ConversationalVoiceResponse
                 {
+                    ConversationId = request.ConversationId,
                     SpeechPaths = instructions,
                     CallbackPath = $"twilio/voice/reply/{request.SeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}&AIResponseWaitTime={++request.AIResponseWaitTime}",
                     ActionOnEmptyResult = true
@@ -358,6 +362,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>
             {
                 var instruction = new ConversationalVoiceResponse
                 {
+                    ConversationId = request.ConversationId,
                     SpeechPaths = [$"twilio/voice/speeches/{request.ConversationId}/{reply.SpeechFileName}"],
                     CallbackPath = $"twilio/voice/receive/{nextSeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}",
                     ActionOnEmptyResult = true,
@@ -383,8 +388,19 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>
     [HttpPost("twilio/voice/init-outbound-call")]
     public TwiMLResult InitiateOutboundCall(ConversationalVoiceRequest request)
     {
+        VoiceResponse response = default!;
+        if (request.AnsweredBy == "machine_start" &&
+            request.Direction == "outbound-api" &&
+            request.InitAudioFile != null)
+        {
+            response = new VoiceResponse();
+            response.Play(new Uri(request.InitAudioFile));
+            return TwiML(response);
+        }
+
         var instruction = new ConversationalVoiceResponse
         {
+            ConversationId = request.ConversationId,
             ActionOnEmptyResult = true,
             CallbackPath = $"twilio/voice/receive/1?conversation-id={request.ConversationId}",
         };
@@ -396,7 +412,7 @@ public TwiMLResult InitiateOutboundCall(ConversationalVoiceRequest request)
         }
 
         var twilio = _services.GetRequiredService<TwilioService>();
-        var response = twilio.ReturnNoninterruptedInstructions(instruction);
+        response = twilio.ReturnNoninterruptedInstructions(instruction);
         return TwiML(response);
     }
 
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/Interfaces/ITwilioCallStatusHook.cs b/src/Plugins/BotSharp.Plugin.Twilio/Interfaces/ITwilioCallStatusHook.cs
@@ -7,4 +7,5 @@ public interface ITwilioCallStatusHook
 {
     Task OnVoicemailLeft(ConversationalVoiceRequest request);
     Task OnUserDisconnected(ConversationalVoiceRequest request);
+    Task OnRecordingCompleted(ConversationalVoiceRequest request);
 }
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/Models/ConversationalVoiceResponse.cs b/src/Plugins/BotSharp.Plugin.Twilio/Models/ConversationalVoiceResponse.cs
@@ -2,6 +2,7 @@ namespace BotSharp.Plugin.Twilio.Models;
 
 public class ConversationalVoiceResponse
 {
+    public string ConversationId { get; set; } = null!;
     public List<string> SpeechPaths { get; set; } = [];
     public string CallbackPath { get; set; } 
     public bool ActionOnEmptyResult { get; set; }
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/OutboundPhoneCallHandler/Functions/OutboundPhoneCallFn.cs b/src/Plugins/BotSharp.Plugin.Twilio/OutboundPhoneCallHandler/Functions/OutboundPhoneCallFn.cs
@@ -63,6 +63,7 @@ public async Task<bool> Execute(RoleDialogModel message)
 
         var processUrl = $"{_twilioSetting.CallbackHost}/twilio";
         var statusUrl = $"{_twilioSetting.CallbackHost}/twilio/voice/status?conversation-id={newConversationId}";
+        var recordingStatusUrl = $"{_twilioSetting.CallbackHost}/twilio/recording/status?conversation-id={newConversationId}";
 
         // Generate initial assistant audio
         string initAudioUrl = null;
@@ -102,7 +103,9 @@ public async Task<bool> Execute(RoleDialogModel message)
             from: new PhoneNumber(_twilioSetting.PhoneNumber),
             statusCallback: new Uri(statusUrl),
             // https://www.twilio.com/docs/voice/answering-machine-detection
-            machineDetection: _twilioSetting.MachineDetection);
+            machineDetection: _twilioSetting.MachineDetection,
+            record: _twilioSetting.RecordingEnabled,
+            recordingStatusCallback: $"{_twilioSetting.CallbackHost}/twilio/record/status?conversation-id={newConversationId}");
 
         var convService = _services.GetRequiredService<IConversationService>();
         var routing = _services.GetRequiredService<IRoutingContext>();
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/Services/TwilioService.cs b/src/Plugins/BotSharp.Plugin.Twilio/Services/TwilioService.cs
@@ -1,7 +1,6 @@
 using BotSharp.Abstraction.Utilities;
 using BotSharp.Plugin.Twilio.Models;
 using Twilio.Jwt.AccessToken;
-using Twilio.TwiML.Messaging;
 using Token = Twilio.Jwt.AccessToken.Token;
 
 namespace BotSharp.Plugin.Twilio.Services;
@@ -101,13 +100,13 @@ public VoiceResponse ReturnInstructions(ConversationalVoiceResponse conversation
         return response;
     }
 
-    public VoiceResponse ReturnNoninterruptedInstructions(ConversationalVoiceResponse conversationalVoiceResponse)
+    public VoiceResponse ReturnNoninterruptedInstructions(ConversationalVoiceResponse voiceResponse)
     {
         var response = new VoiceResponse();
-        response.Pause(2);
-        if (conversationalVoiceResponse.SpeechPaths != null && conversationalVoiceResponse.SpeechPaths.Any())
+
+        if (voiceResponse.SpeechPaths != null && voiceResponse.SpeechPaths.Any())
         {
-            foreach (var speechPath in conversationalVoiceResponse.SpeechPaths)
+            foreach (var speechPath in voiceResponse.SpeechPaths)
             {
                 if (speechPath.StartsWith(_settings.CallbackHost))
                 {
@@ -119,21 +118,23 @@ public VoiceResponse ReturnNoninterruptedInstructions(ConversationalVoiceRespons
                 }
             }
         }
+
         var gather = new Gather()
         {
             Input = new List<Gather.InputEnum>()
             {
                 Gather.InputEnum.Speech,
                 Gather.InputEnum.Dtmf
             },
-            Action = new Uri($"{_settings.CallbackHost}/{conversationalVoiceResponse.CallbackPath}"),
+            Action = new Uri($"{_settings.CallbackHost}/{voiceResponse.CallbackPath}"),
             Enhanced = true,
             SpeechModel = Gather.SpeechModelEnum.PhoneCall,
             SpeechTimeout = "auto", // conversationalVoiceResponse.Timeout > 0 ? conversationalVoiceResponse.Timeout.ToString() : "3",
-            Timeout = conversationalVoiceResponse.Timeout > 0 ? conversationalVoiceResponse.Timeout : 3,
-            ActionOnEmptyResult = conversationalVoiceResponse.ActionOnEmptyResult
+            Timeout = voiceResponse.Timeout > 0 ? voiceResponse.Timeout : 3,
+            ActionOnEmptyResult = voiceResponse.ActionOnEmptyResult,
         };
         response.Append(gather);
+
         return response;
     }
 
@@ -192,6 +193,7 @@ public VoiceResponse HoldOn(int interval, string message = null)
     public VoiceResponse ReturnBidirectionalMediaStreamsInstructions(string conversationId, ConversationalVoiceResponse conversationalVoiceResponse)
     {
         var response = new VoiceResponse();
+
         if (conversationalVoiceResponse.SpeechPaths != null && conversationalVoiceResponse.SpeechPaths.Any())
         {
             foreach (var speechPath in conversationalVoiceResponse.SpeechPaths)
diff --git a/src/Plugins/BotSharp.Plugin.Twilio/Settings/TwilioSetting.cs b/src/Plugins/BotSharp.Plugin.Twilio/Settings/TwilioSetting.cs
@@ -33,5 +33,4 @@ public class TwilioSetting
     public string? MachineDetection { get; set; }
 
     public bool RecordingEnabled { get; set; } = false;
-    public bool RecordingTranscribe { get; set; } = false;
 }

Original file line number	Diff line number	Diff line change
`@@ -48,6 +48,7 @@ public async Task<TwiMLResult> InitiateStreamConversation(ConversationalVoiceReq`
`48`	`48`
`49`	`49`	`var instruction = new ConversationalVoiceResponse`
`50`	`50`	`{`
	`51`	`+ ConversationId = request.ConversationId,`
`51`	`52`	`SpeechPaths = [],`
`52`	`53`	`ActionOnEmptyResult = true`
`53`	`54`	`};`
Original file line number	Diff line number	Diff line change
`@@ -52,6 +52,7 @@ public async Task<TwiMLResult> InitiateConversation(ConversationalVoiceRequest r`
`52`	`52`	`VoiceResponse response = null;`
`53`	`53`	`var instruction = new ConversationalVoiceResponse`
`54`	`54`	`{`
	`55`	`+ ConversationId = request.ConversationId,`
`55`	`56`	`SpeechPaths = ["twilio/welcome.mp3"],`
`56`	`57`	`ActionOnEmptyResult = true`
`57`	`58`	`};`
`@@ -170,6 +171,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>`
`170`	`171`	`{`
`171`	`172`	`var instruction = new ConversationalVoiceResponse`
`172`	`173`	`{`
	`174`	`+ ConversationId = request.ConversationId,`
`173`	`175`	`SpeechPaths = new List<string>(),`
`174`	`176`	`CallbackPath = $"twilio/voice/receive/{request.SeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}&attempts={++request.Attempts}",`
`175`	`177`	`ActionOnEmptyResult = true`
`@@ -273,6 +275,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>`
`273`	`275`
`274`	`276`	`var instruction = new ConversationalVoiceResponse`
`275`	`277`	`{`
	`278`	`+ ConversationId = request.ConversationId,`
`276`	`279`	`SpeechPaths = speechPaths,`
`277`	`280`	`CallbackPath = $"twilio/voice/reply/{request.SeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}&AIResponseWaitTime={++request.AIResponseWaitTime}",`
`278`	`281`	`ActionOnEmptyResult = true`
`@@ -312,6 +315,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>`
`312`	`315`
`313`	`316`	`var instruction = new ConversationalVoiceResponse`
`314`	`317`	`{`
	`318`	`+ ConversationId = request.ConversationId,`
`315`	`319`	`SpeechPaths = instructions,`
`316`	`320`	`CallbackPath = $"twilio/voice/reply/{request.SeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}&AIResponseWaitTime={++request.AIResponseWaitTime}",`
`317`	`321`	`ActionOnEmptyResult = true`
`@@ -358,6 +362,7 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>`
`358`	`362`	`{`
`359`	`363`	`var instruction = new ConversationalVoiceResponse`
`360`	`364`	`{`
	`365`	`+ ConversationId = request.ConversationId,`
`361`	`366`	`SpeechPaths = [$"twilio/voice/speeches/{request.ConversationId}/{reply.SpeechFileName}"],`
`362`	`367`	`CallbackPath = $"twilio/voice/receive/{nextSeqNum}?conversation-id={request.ConversationId}&{GenerateStatesParameter(request.States)}",`
`363`	`368`	`ActionOnEmptyResult = true,`
`@@ -383,8 +388,19 @@ await HookEmitter.Emit<ITwilioSessionHook>(_services, async hook =>`
`383`	`388`	`[HttpPost("twilio/voice/init-outbound-call")]`
`384`	`389`	`public TwiMLResult InitiateOutboundCall(ConversationalVoiceRequest request)`
`385`	`390`	`{`
	`391`	`+ VoiceResponse response = default!;`
	`392`	`+ if (request.AnsweredBy == "machine_start" &&`
	`393`	`+ request.Direction == "outbound-api" &&`
	`394`	`+ request.InitAudioFile != null)`
	`395`	`+ {`
	`396`	`+ response = new VoiceResponse();`
	`397`	`+ response.Play(new Uri(request.InitAudioFile));`
	`398`	`+ return TwiML(response);`
	`399`	`+ }`
	`400`	`+`
`386`	`401`	`var instruction = new ConversationalVoiceResponse`
`387`	`402`	`{`
	`403`	`+ ConversationId = request.ConversationId,`
`388`	`404`	`ActionOnEmptyResult = true,`
`389`	`405`	`CallbackPath = $"twilio/voice/receive/1?conversation-id={request.ConversationId}",`
`390`	`406`	`};`
`@@ -396,7 +412,7 @@ public TwiMLResult InitiateOutboundCall(ConversationalVoiceRequest request)`
`396`	`412`	`}`
`397`	`413`
`398`	`414`	`var twilio = _services.GetRequiredService<TwilioService>();`
`399`		`- var response = twilio.ReturnNoninterruptedInstructions(instruction);`
	`415`	`+ response = twilio.ReturnNoninterruptedInstructions(instruction);`
`400`	`416`	`return TwiML(response);`
`401`	`417`	`}`
`402`	`418`
Original file line number	Diff line number	Diff line change
`@@ -7,4 +7,5 @@ public interface ITwilioCallStatusHook`
`7`	`7`	`{`
`8`	`8`	`Task OnVoicemailLeft(ConversationalVoiceRequest request);`
`9`	`9`	`Task OnUserDisconnected(ConversationalVoiceRequest request);`
	`10`	`+ Task OnRecordingCompleted(ConversationalVoiceRequest request);`
`10`	`11`	`}`
Original file line number	Diff line number	Diff line change
`@@ -2,6 +2,7 @@ namespace BotSharp.Plugin.Twilio.Models;`
`2`	`2`
`3`	`3`	`public class ConversationalVoiceResponse`
`4`	`4`	`{`
	`5`	`+ public string ConversationId { get; set; } = null!;`
`5`	`6`	`public List<string> SpeechPaths { get; set; } = [];`
`6`	`7`	`public string CallbackPath { get; set; }`
`7`	`8`	`public bool ActionOnEmptyResult { get; set; }`
Original file line number	Diff line number	Diff line change
`@@ -1,7 +1,6 @@`
`1`	`1`	`using BotSharp.Abstraction.Utilities;`
`2`	`2`	`using BotSharp.Plugin.Twilio.Models;`
`3`	`3`	`using Twilio.Jwt.AccessToken;`
`4`		`-using Twilio.TwiML.Messaging;`
`5`	`4`	`using Token = Twilio.Jwt.AccessToken.Token;`
`6`	`5`
`7`	`6`	`namespace BotSharp.Plugin.Twilio.Services;`
`@@ -101,13 +100,13 @@ public VoiceResponse ReturnInstructions(ConversationalVoiceResponse conversation`
`101`	`100`	`return response;`
`102`	`101`	`}`
`103`	`102`
`104`		`- public VoiceResponse ReturnNoninterruptedInstructions(ConversationalVoiceResponse conversationalVoiceResponse)`
	`103`	`+ public VoiceResponse ReturnNoninterruptedInstructions(ConversationalVoiceResponse voiceResponse)`
`105`	`104`	`{`
`106`	`105`	`var response = new VoiceResponse();`
`107`		`- response.Pause(2);`
`108`		`- if (conversationalVoiceResponse.SpeechPaths != null && conversationalVoiceResponse.SpeechPaths.Any())`
	`106`	`+`
	`107`	`+ if (voiceResponse.SpeechPaths != null && voiceResponse.SpeechPaths.Any())`
`109`	`108`	`{`
`110`		`- foreach (var speechPath in conversationalVoiceResponse.SpeechPaths)`
	`109`	`+ foreach (var speechPath in voiceResponse.SpeechPaths)`
`111`	`110`	`{`
`112`	`111`	`if (speechPath.StartsWith(_settings.CallbackHost))`
`113`	`112`	`{`
`@@ -119,21 +118,23 @@ public VoiceResponse ReturnNoninterruptedInstructions(ConversationalVoiceRespons`
`119`	`118`	`}`
`120`	`119`	`}`
`121`	`120`	`}`
	`121`	`+`
`122`	`122`	`var gather = new Gather()`
`123`	`123`	`{`
`124`	`124`	`Input = new List<Gather.InputEnum>()`
`125`	`125`	`{`
`126`	`126`	`Gather.InputEnum.Speech,`
`127`	`127`	`Gather.InputEnum.Dtmf`
`128`	`128`	`},`
`129`		`- Action = new Uri($"{_settings.CallbackHost}/{conversationalVoiceResponse.CallbackPath}"),`
	`129`	`+ Action = new Uri($"{_settings.CallbackHost}/{voiceResponse.CallbackPath}"),`
`130`	`130`	`Enhanced = true,`
`131`	`131`	`SpeechModel = Gather.SpeechModelEnum.PhoneCall,`
`132`	`132`	`SpeechTimeout = "auto", // conversationalVoiceResponse.Timeout > 0 ? conversationalVoiceResponse.Timeout.ToString() : "3",`
`133`		`- Timeout = conversationalVoiceResponse.Timeout > 0 ? conversationalVoiceResponse.Timeout : 3,`
`134`		`- ActionOnEmptyResult = conversationalVoiceResponse.ActionOnEmptyResult`
	`133`	`+ Timeout = voiceResponse.Timeout > 0 ? voiceResponse.Timeout : 3,`
	`134`	`+ ActionOnEmptyResult = voiceResponse.ActionOnEmptyResult,`
`135`	`135`	`};`
`136`	`136`	`response.Append(gather);`
	`137`	`+`
`137`	`138`	`return response;`
`138`	`139`	`}`
`139`	`140`
`@@ -192,6 +193,7 @@ public VoiceResponse HoldOn(int interval, string message = null)`
`192`	`193`	`public VoiceResponse ReturnBidirectionalMediaStreamsInstructions(string conversationId, ConversationalVoiceResponse conversationalVoiceResponse)`
`193`	`194`	`{`
`194`	`195`	`var response = new VoiceResponse();`
	`196`	`+`
`195`	`197`	`if (conversationalVoiceResponse.SpeechPaths != null && conversationalVoiceResponse.SpeechPaths.Any())`
`196`	`198`	`{`
`197`	`199`	`foreach (var speechPath in conversationalVoiceResponse.SpeechPaths)`
Original file line number	Diff line number	Diff line change
`@@ -33,5 +33,4 @@ public class TwilioSetting`
`33`	`33`	`public string? MachineDetection { get; set; }`
`34`	`34`
`35`	`35`	`public bool RecordingEnabled { get; set; } = false;`
`36`		`- public bool RecordingTranscribe { get; set; } = false;`
`37`	`36`	`}`