UP-Viagg-io/Viagg-io/Assets/afca/ViaggioAI/Scripts/AIServices/ConversationService.cs

using System;
using System.Threading.Tasks;
using UnityEngine;


#region Enums

public enum EConversationState
{
    NotStarted = 0,
    Running = 10,
    OutroText = 20,
    Ended = 30,
}

public enum EBotState
{
    Idle = 0,
    Thinking = 10,
    Speaking = 20,
}

#endregion

public class ConversationService : MonoBehaviour
{

    #region Inspector Properties

    [Header("Config Values")]
    [SerializeField]
    private bool debugModeIsActive;

    [Header("Scene Objects")]
    [SerializeField]
    private OpenAIServices openAIServices_Conversation;

    #endregion

    #region Public Properties

    #region ConversationState

    private EConversationState _conversationState = EConversationState.NotStarted;

    public EConversationState ConversationState
    {
        get { return this._conversationState; }
        set
        {
            if (value != this._conversationState)
            {
                this.logIfInDebugMode($"ConversationState changed, new value= {value}");

                this._conversationState = value;

                this.OnConversationStateChangedEvent?.Invoke(this, value);
            }
        }
    }

    public event EventHandler<EConversationState> OnConversationStateChangedEvent;

    #endregion

    #region BotState

    private EBotState _botState = EBotState.Idle;

    public EBotState BotState
    {
        get { return this._botState; }
        set
        {
            if (value != this._botState)
            {
                this.logIfInDebugMode($"BotState changed, new value= {value}");

                this._botState = value;

                this.OnBotStateChangedEvent?.Invoke(this, value);
            }
        }
    }

    public event EventHandler<EBotState> OnBotStateChangedEvent;

    #endregion

    #endregion

    #region Private Properties

    private ViaggioAIManager vaim { get { return ViaggioAIManager.Instance; } }

    private string languageCode;
    private string instructions;
    private string context;
    private int maxRoundTrips;
    private float maxSilenceBeforeTimeOut;
    private string outroText;

    private int currentRoundTrips;
    private float currentSilenceDuratation;

    #endregion

    #region Framework Functions

    void OnEnable()
    {
        this.subscribeToEvents();
    }

    void Update()
    {
        this.checkForTimeout();
    }

    void OnDisable()
    {
        this.unsubscribeFromEvents();
    }

    #endregion

    #region Events

    private async void subscribeToEvents()
    {
        while (vaim == null)
        {
            await System.Threading.Tasks.Task.Delay(10);
        }

        vaim.SpeechRecognitionService.OnPartialTranscriptionChangedEvent += this.onPartialTranscriptionChanged;
        vaim.SpeechRecognitionService.OnFullTranscriptionChangedEvent += this.onFullTranscriptionChangedEvent;

        vaim.SpeechSynthesizerService.OnSpeechOutputEndedEvent += this.onSpeechOutputEnded;

        this.openAIServices_Conversation.OnLastBotErrorChangedEvent += this.onLastBotErrorChanged;
        this.openAIServices_Conversation.OnLastBotReplyChangedEvent += this.onLastBotReplyChanged;
    }

    private void unsubscribeFromEvents()
    {
        if (vaim?.SpeechRecognitionService != null)
        {
            vaim.SpeechRecognitionService.OnPartialTranscriptionChangedEvent -= this.onPartialTranscriptionChanged;
            vaim.SpeechRecognitionService.OnFullTranscriptionChangedEvent -= this.onFullTranscriptionChangedEvent;
        }

        if (vaim?.SpeechSynthesizerService != null)
        {
            vaim.SpeechSynthesizerService.OnSpeechOutputEndedEvent -= this.onSpeechOutputEnded;
        }

        if (this?.openAIServices_Conversation != null)
        {
            this.openAIServices_Conversation.OnLastBotErrorChangedEvent -= this.onLastBotErrorChanged;
            this.openAIServices_Conversation.OnLastBotReplyChangedEvent -= this.onLastBotReplyChanged;
        }
    }

    private void onLastBotErrorChanged(object sender, string botError)
    {
        if (vaim.ViaggioAIState != EViaggioAIState.Conversation)
        {
            return;
        }
        this.OnConversationErrorEvent?.Invoke(this, $"Conversation ended because of BotError: {botError}");

        this.ConversationState = EConversationState.Ended;
        vaim.StopConversation();
    }

    private void onLastBotReplyChanged(object sender, string botReply)
    {
        if (vaim.ViaggioAIState != EViaggioAIState.Conversation)
        {
            return;
        }
        if (this.ConversationState != EConversationState.Running)
        {
            ViaggioAIManager.Instance.LogIfInDebugMode($"Got Bot Reply in wrong conversation state ({this.ConversationState}). Ignoring");
            return;
        }

        this.handleBotReply(botReply);
    }

    private void onPartialTranscriptionChanged(object sender, string e)
    {
        if (vaim.ViaggioAIState != EViaggioAIState.Conversation)
        {
            return;
        }

        this.currentSilenceDuratation = 0;
    }

    private void onFullTranscriptionChangedEvent(object sender, string userInput)
    {
        if (vaim.ViaggioAIState != EViaggioAIState.Conversation)
        {
            return;
        }
        if (this.ConversationState != EConversationState.Running)
        {
            ViaggioAIManager.Instance.LogIfInDebugMode($"Got User Input in wrong conversation state ({this.ConversationState}). Ignoring");
            return;
        }

        this.currentSilenceDuratation = 0;

        this.handleUserInput(userInput);
    }

    private void onSpeechOutputEnded(object sender, bool e)
    {
        if (this.BotState != EBotState.Speaking)
        {
            return;
        }

        this.BotState = EBotState.Idle;
    }

    #endregion

    #region Public Events

    public event EventHandler<string> OnConversationInitiatedEvent;
    public event EventHandler<string> OnConversationErrorEvent;
    public event EventHandler<string> OnConversationEndedEvent;

    #endregion

    #region Public Functions

    public async void StartConversation(string _languageCode, string _instructions, string _context, int _maxRoundtrips, float _maxSilenceBeforeTimeOut, string _outroText)
    {
        this.logIfInDebugMode($"Starting new conversation");

        this.languageCode = _languageCode;
        this.instructions = _instructions;
        this.context = _context;
        this.maxRoundTrips = _maxRoundtrips;
        this.maxSilenceBeforeTimeOut = _maxSilenceBeforeTimeOut;
        this.outroText = _outroText;

        string fullInitialPrompt = this.instructions + this.context;

        await this.sendTextToBot(fullInitialPrompt);

        this.ConversationState = EConversationState.Running;
        this.OnConversationInitiatedEvent?.Invoke(this, $"Conversation started");

        this.currentRoundTrips = 0;
        this.currentSilenceDuratation = 0;
    }

    public void StopConversation()
    {
        vaim.SpeechRecognitionService.StopListening();
        this.openAIServices_Conversation.Stop();

        this.ConversationState = EConversationState.NotStarted;
    }

    #endregion

    #region Private Functions

    private async Task sendTextToBot(string newText)
    {
        while (!this.openAIServices_Conversation.IsInitialized)
        {
            await Task.Delay(200);

            if (this == null)
            {
                return;
            }

            this.logIfInDebugMode($"openAIServices_Conversation not initialized yet, wait before sending text...");
        }

        this.openAIServices_Conversation.SendTextToBot(newText);

        this.BotState = EBotState.Thinking;

        this.currentRoundTrips++;

        this.logIfInDebugMode($"Sent text to bot: {newText}. Current Round Trips = {this.currentRoundTrips}");
    }

    private void handleBotReply(string botReply)
    {
        if (vaim.ViaggioAIState != EViaggioAIState.Conversation)
        {
            return;
        }

        string textToSynthesize = botReply;

        if (this.currentRoundTrips == this.maxRoundTrips)
        {
            this.logIfInDebugMode($"Got bot reply {botReply} and reached max roundtrips {this.maxRoundTrips}");

            textToSynthesize += this.outroText;
            this.ConversationState = EConversationState.Ended;

            this.stopConversation(false);
        }
        else
        {
            this.logIfInDebugMode($"Got bot reply {botReply} and started listening again.");
            vaim.SpeechRecognitionService.StartListeningOnceAsync();
        }

        vaim.SpeechSynthesizerService.Synthesize(textToSynthesize, this.languageCode);

        this.BotState = EBotState.Speaking;
    }

    private async void handleUserInput(string userInput)
    {
        if (vaim.SpeechSynthesizerService.SpeechSynthesizerState == ESpeechSynthesizerState.Speaking)
        {
            await vaim.SpeechSynthesizerService.Stop();
        }

        this.sendTextToBot(userInput);
    }

    private void checkForTimeout()
    {
        if (this.ConversationState != EConversationState.Running)
        {
            return;
        }

        if (this.BotState == EBotState.Speaking || this.BotState == EBotState.Thinking)
        {
            // Can't timeout while bot is speaking or thinking
            return;
        }

        this.currentSilenceDuratation += Time.deltaTime;
        Debug.Log($"Increased currentSilenceDuratation ({this.currentSilenceDuratation}), ConvState ={this.ConversationState}, BotState={this.BotState}");

        if (this.currentSilenceDuratation > this.maxSilenceBeforeTimeOut)
        {
            this.logIfInDebugMode("Conversation timed out -> Stopping");

            this.stopConversation(true);
        }
    }

    private void stopConversation(bool timeOut)
    {
        this.logIfInDebugMode("Stopping conversation");

        this.OnConversationEndedEvent?.Invoke(this, $"Conversation stopped");

        if (timeOut)
        {
            vaim.SpeechSynthesizerService.Synthesize(this.outroText, this.languageCode);
        }
        
        this.ConversationState = EConversationState.Ended;

        vaim.StopConversation();
    }

    private void logIfInDebugMode(string message)
    {
        if (!this.debugModeIsActive)
        {
            return;
        }

        Debug.Log($"(ViaggioAI ConversationService) => {message}");
    }

    #endregion

}