BotSharp/src/Infrastructure/BotSharp.Core.Realtime/Services/RealtimeHub.cs

288 lines
11 KiB
C#
Raw Normal View History

2025-03-06 18:01:43 +00:00
namespace BotSharp.Core.Realtime.Services;
2025-02-07 22:40:57 +00:00
public class RealtimeHub : IRealtimeHub
{
private readonly IServiceProvider _services;
private readonly ILogger _logger;
2025-02-27 04:40:50 +00:00
2025-03-06 21:10:17 +00:00
private RealtimeHubConnection _conn;
public RealtimeHubConnection HubConn => _conn;
private IRealTimeCompletion _completer;
public IRealTimeCompletion Completer => _completer;
2025-02-07 22:40:57 +00:00
public RealtimeHub(IServiceProvider services, ILogger<RealtimeHub> logger)
{
_services = services;
_logger = logger;
}
public async Task Listen(WebSocket userWebSocket,
2025-03-06 21:10:17 +00:00
Action<string> onUserMessageReceived)
2025-02-07 22:40:57 +00:00
{
2025-03-03 21:30:06 +00:00
var buffer = new byte[1024 * 16];
2025-02-07 22:40:57 +00:00
WebSocketReceiveResult result;
2025-03-06 21:10:17 +00:00
2025-02-07 22:40:57 +00:00
do
{
result = await userWebSocket.ReceiveAsync(new ArraySegment<byte>(buffer), CancellationToken.None);
string receivedText = Encoding.UTF8.GetString(buffer, 0, result.Count);
2025-03-03 21:30:06 +00:00
2025-02-07 22:40:57 +00:00
if (string.IsNullOrEmpty(receivedText))
{
continue;
}
2025-03-06 21:10:17 +00:00
onUserMessageReceived(receivedText);
2025-02-09 23:31:46 +00:00
2025-03-06 21:10:17 +00:00
if (_conn.Event == "user_connected")
2025-02-07 22:40:57 +00:00
{
2025-03-06 21:10:17 +00:00
await ConnectToModel(userWebSocket);
2025-02-07 22:40:57 +00:00
}
2025-03-06 21:10:17 +00:00
else if (_conn.Event == "user_data_received")
2025-02-07 22:40:57 +00:00
{
2025-03-06 21:10:17 +00:00
await _completer.AppenAudioBuffer(_conn.Data);
2025-02-07 22:40:57 +00:00
}
2025-03-06 21:10:17 +00:00
else if (_conn.Event == "user_dtmf_received")
2025-03-05 22:57:34 +00:00
{
2025-03-06 21:10:17 +00:00
await HandleUserDtmfReceived();
2025-03-05 22:57:34 +00:00
}
2025-03-06 21:10:17 +00:00
else if (_conn.Event == "user_disconnected")
2025-02-07 22:40:57 +00:00
{
2025-03-06 21:10:17 +00:00
await _completer.Disconnect();
await HandleUserDisconnected();
2025-02-07 22:40:57 +00:00
}
} while (!result.CloseStatus.HasValue);
await userWebSocket.CloseAsync(result.CloseStatus.Value, result.CloseStatusDescription, CancellationToken.None);
}
2025-03-06 21:10:17 +00:00
private async Task ConnectToModel(WebSocket userWebSocket)
2025-02-07 22:40:57 +00:00
{
2025-02-09 23:31:46 +00:00
var hookProvider = _services.GetRequiredService<ConversationHookProvider>();
var convService = _services.GetRequiredService<IConversationService>();
2025-03-06 21:10:17 +00:00
convService.SetConversationId(_conn.ConversationId, []);
var conversation = await convService.GetConversation(_conn.ConversationId);
2025-02-11 23:27:07 +00:00
2025-02-09 23:31:46 +00:00
var agentService = _services.GetRequiredService<IAgentService>();
var agent = await agentService.LoadAgent(conversation.AgentId);
2025-03-06 21:10:17 +00:00
_conn.CurrentAgentId = agent.Id;
2025-02-11 23:27:07 +00:00
2025-03-03 21:30:06 +00:00
// Set model
var model = agent.LlmConfig.Model;
if (!model.Contains("-realtime-"))
{
var llmProviderService = _services.GetRequiredService<ILlmProviderService>();
2025-03-22 00:28:22 +00:00
model = llmProviderService.GetProviderModel("openai", "gpt-4o", realTime: true).Name;
2025-03-03 21:30:06 +00:00
}
2025-03-06 21:10:17 +00:00
_completer.SetModelName(model);
_conn.Model = model;
2025-03-03 21:30:06 +00:00
2025-02-09 23:31:46 +00:00
var routing = _services.GetRequiredService<IRoutingService>();
2025-02-28 18:44:59 +00:00
routing.Context.Push(agent.Id);
2025-03-12 19:03:38 +00:00
var storage = _services.GetRequiredService<IConversationStorage>();
2025-02-09 23:31:46 +00:00
var dialogs = convService.GetDialogHistory();
2025-02-28 03:19:20 +00:00
if (dialogs.Count == 0)
{
dialogs.Add(new RoleDialogModel(AgentRole.User, "Hi"));
2025-03-12 19:03:38 +00:00
storage.Append(_conn.ConversationId, dialogs.First());
2025-02-28 03:19:20 +00:00
}
2025-03-12 19:03:38 +00:00
2025-02-09 23:31:46 +00:00
routing.Context.SetDialogs(dialogs);
2025-03-17 22:09:02 +00:00
routing.Context.SetMessageId(_conn.ConversationId, dialogs.Last().MessageId);
2025-02-09 23:31:46 +00:00
var states = _services.GetRequiredService<IConversationStateService>();
2025-03-06 21:10:17 +00:00
await _completer.Connect(_conn,
2025-02-09 23:31:46 +00:00
onModelReady: async () =>
{
2025-03-15 02:06:40 +00:00
// Not TriggerModelInference, waiting for user utter.
2025-03-15 14:29:46 +00:00
var instruction = await _completer.UpdateSession(_conn);
2025-03-15 02:06:40 +00:00
// Trigger model inference if there is no audio file in the conversation
if (!states.ContainsState("init_audio_file"))
{
if (dialogs.LastOrDefault()?.Role == AgentRole.Assistant)
{
await _completer.TriggerModelInference($"Rephase your last response:\r\n{dialogs.LastOrDefault()?.Content}");
}
else
{
await _completer.TriggerModelInference("Reply based on the conversation context.");
}
2025-02-10 23:28:03 +00:00
}
2025-03-15 14:29:46 +00:00
else
{
2025-03-18 20:14:36 +00:00
// Append dialogs into model context
var history = "[CONVERSATION HISTORY]\r\n";
2025-03-15 14:29:46 +00:00
foreach (var message in dialogs)
{
2025-03-18 20:14:36 +00:00
history += $"{message.Role}: {message.Content}\r\n";
2025-03-15 14:29:46 +00:00
}
2025-03-18 20:14:36 +00:00
await _completer.TriggerModelInference($"{instruction}\r\n\r\n{history}\r\n\r\nAssist user without repeating your previous statement.");
2025-03-15 14:29:46 +00:00
}
2025-02-09 23:31:46 +00:00
},
2025-03-06 09:03:30 +00:00
onModelAudioDeltaReceived: async (audioDeltaData, itemId) =>
2025-02-09 23:31:46 +00:00
{
2025-03-06 21:10:17 +00:00
var data = _conn.OnModelMessageReceived(audioDeltaData);
2025-02-09 23:31:46 +00:00
await SendEventToUser(userWebSocket, data);
2025-03-05 18:29:28 +00:00
2025-03-06 09:03:30 +00:00
// If this is the first delta of a new response, set the start timestamp
2025-03-06 21:10:17 +00:00
if (!_conn.ResponseStartTimestamp.HasValue)
2025-03-05 18:29:28 +00:00
{
2025-03-06 21:10:17 +00:00
_conn.ResponseStartTimestamp = _conn.LatestMediaTimestamp;
_logger.LogDebug($"Setting start timestamp for new response: {_conn.ResponseStartTimestamp}ms");
2025-03-06 09:03:30 +00:00
}
// Record last assistant item ID for interruption handling
if (!string.IsNullOrEmpty(itemId))
{
2025-03-06 21:10:17 +00:00
_conn.LastAssistantItemId = itemId;
2025-03-05 18:29:28 +00:00
}
2025-03-06 09:03:30 +00:00
// Send mark messages to Media Streams so we know if and when AI response playback is finished
2025-03-06 21:10:17 +00:00
await SendMark(userWebSocket, _conn);
2025-02-09 23:31:46 +00:00
},
onModelAudioResponseDone: async () =>
{
2025-03-06 21:10:17 +00:00
var data = _conn.OnModelAudioResponseDone();
2025-02-09 23:31:46 +00:00
await SendEventToUser(userWebSocket, data);
},
onAudioTranscriptDone: async transcript =>
{
},
2025-02-11 23:27:07 +00:00
onModelResponseDone: async messages =>
2025-02-09 23:31:46 +00:00
{
foreach (var message in messages)
{
// Invoke function
2025-03-06 21:10:17 +00:00
if (message.MessageType == MessageTypeName.FunctionCall &&
!string.IsNullOrEmpty(message.FunctionName))
2025-02-09 23:31:46 +00:00
{
await routing.InvokeFunction(message.FunctionName, message);
}
2025-03-05 18:29:28 +00:00
else
{
2025-03-05 18:29:28 +00:00
// append output audio transcript to conversation
dialogs.Add(message);
2025-03-12 19:03:38 +00:00
storage.Append(_conn.ConversationId, message);
2025-03-05 18:29:28 +00:00
foreach (var hook in hookProvider.HooksOrderByPriority)
{
hook.SetAgent(agent)
.SetConversation(conversation);
2025-02-11 23:27:07 +00:00
2025-03-05 18:29:28 +00:00
await hook.OnResponseGenerated(message);
}
2025-02-11 23:27:07 +00:00
}
2025-02-09 23:31:46 +00:00
}
},
2025-02-11 23:27:07 +00:00
onConversationItemCreated: async response =>
{
},
onInputAudioTranscriptionCompleted: async message =>
{
2025-03-01 02:42:03 +00:00
// append input audio transcript to conversation
2025-02-11 23:27:07 +00:00
dialogs.Add(message);
2025-03-12 19:03:38 +00:00
storage.Append(_conn.ConversationId, message);
2025-03-17 22:09:02 +00:00
routing.Context.SetMessageId(_conn.ConversationId, message.MessageId);
2025-03-01 02:42:03 +00:00
foreach (var hook in hookProvider.HooksOrderByPriority)
{
hook.SetAgent(agent)
.SetConversation(conversation);
await hook.OnMessageReceived(message);
}
2025-02-11 23:27:07 +00:00
},
2025-02-09 23:31:46 +00:00
onUserInterrupted: async () =>
{
2025-03-05 18:29:28 +00:00
// Reset states
2025-03-06 21:10:17 +00:00
_conn.ResetResponseState();
2025-03-05 18:29:28 +00:00
2025-03-06 21:10:17 +00:00
var data = _conn.OnModelUserInterrupted();
2025-02-09 23:31:46 +00:00
await SendEventToUser(userWebSocket, data);
});
2025-02-07 22:40:57 +00:00
}
2025-03-06 09:03:30 +00:00
private async Task SendMark(WebSocket userWebSocket, RealtimeHubConnection conn)
{
if (!string.IsNullOrEmpty(conn.StreamId))
{
var markEvent = new
{
@event = "mark",
streamSid = conn.StreamId,
mark = new { name = "responsePart" }
};
await SendEventToUser(userWebSocket, markEvent);
conn.MarkQueue.Enqueue("responsePart");
}
}
2025-03-06 21:10:17 +00:00
private async Task HandleUserDtmfReceived()
2025-03-05 22:57:34 +00:00
{
var routing = _services.GetRequiredService<IRoutingService>();
var hookProvider = _services.GetRequiredService<ConversationHookProvider>();
var agentService = _services.GetRequiredService<IAgentService>();
2025-03-06 21:10:17 +00:00
var agent = await agentService.LoadAgent(_conn.CurrentAgentId);
2025-03-05 22:57:34 +00:00
var dialogs = routing.Context.GetDialogs();
var convService = _services.GetRequiredService<IConversationService>();
2025-03-06 21:10:17 +00:00
var conversation = await convService.GetConversation(_conn.ConversationId);
2025-03-05 22:57:34 +00:00
2025-03-06 21:10:17 +00:00
var message = new RoleDialogModel(AgentRole.User, _conn.Data)
2025-03-05 22:57:34 +00:00
{
CurrentAgentId = routing.Context.GetCurrentAgentId()
};
dialogs.Add(message);
2025-03-12 19:03:38 +00:00
var storage = _services.GetRequiredService<IConversationStorage>();
storage.Append(_conn.ConversationId, message);
2025-03-05 22:57:34 +00:00
foreach (var hook in hookProvider.HooksOrderByPriority)
{
hook.SetAgent(agent)
.SetConversation(conversation);
await hook.OnMessageReceived(message);
}
2025-03-06 21:10:17 +00:00
await _completer.InsertConversationItem(message);
2025-03-12 01:57:49 +00:00
var instruction = await _completer.UpdateSession(_conn);
await _completer.TriggerModelInference($"{instruction}\r\n\r\nReply based on the user input: {message.Content}");
2025-03-05 22:57:34 +00:00
}
2025-03-06 21:10:17 +00:00
private async Task HandleUserDisconnected()
2025-03-05 22:57:34 +00:00
{
2025-03-14 18:59:13 +00:00
2025-03-05 22:57:34 +00:00
}
2025-02-09 23:31:46 +00:00
private async Task SendEventToUser(WebSocket webSocket, object message)
2025-02-07 22:40:57 +00:00
{
var data = JsonSerializer.Serialize(message);
var buffer = Encoding.UTF8.GetBytes(data);
await webSocket.SendAsync(new ArraySegment<byte>(buffer), WebSocketMessageType.Text, true, CancellationToken.None);
}
2025-03-06 21:10:17 +00:00
public RealtimeHubConnection SetHubConnection(string conversationId)
{
_conn = new RealtimeHubConnection
{
ConversationId = conversationId
};
return _conn;
}
public IRealTimeCompletion SetCompleter(string provider)
{
_completer = _services.GetServices<IRealTimeCompletion>().First(x => x.Provider == provider);
return _completer;
}
2025-02-07 22:40:57 +00:00
}