DeltaVR3DModelGeneration/Assets/_PROJECT/Scripts/ModeGeneration/NPCs/CafeWaiterNPC.cs

using DG.Tweening;
using System;
using TMPro;
using UnityEngine;
using UnityEngine.XR.Interaction.Toolkit;

public class CafeWaiterNPC : NPCController
{
    public FMODWhisperBridge fmodWhisperBridge;
    public TextMeshProUGUI notepadText;
    public GameObject notepad;

    public Transform backRoom;
    public Transform plate;
    public float backRoomMovingTime = 5f;

    public int ignorePlayerCollisionLayer = 2;

    private Vector3 notepadOriginalRotation;
    private Vector3 notepadFlippedRotation;
    private Vector3 startingPosition;

    // states:
    // 0 - idle
    // 1 - waiting for player order
    // 2 - waiting for player to answer "is this correct?"
    // 3 - bringing food
    // 4 - waiting for player to go away
    private int state;
    private float lastPlayerVoiceUpdateTime;

    private void Awake()
    {
        state = 0;
        lastPlayerVoiceUpdateTime = Time.time;
        startingPosition = transform.position;
    }

    private void Start()
    {
        notepadOriginalRotation = notepad.transform.localRotation.eulerAngles;
        notepadFlippedRotation = notepadOriginalRotation + new Vector3(0, 180, 0);
        notepad.SetActive(false);
    }

    protected override void OnPlayerApproach()
    {
        if (state == 0)
        {
            SpeakVoiceLine(0);

            fmodWhisperBridge.OnWhisperSegmentUpdated += OnPlayerSpeechUpdate;
            fmodWhisperBridge.OnWhisperSegmentFinished += OnPlayerSpeechFinished;
            fmodWhisperBridge.ActivateRecording();

            notepad.SetActive(true);

            state = 1;
        }
    }

    protected override void OnPlayerLeave()
    {
        fmodWhisperBridge.OnWhisperSegmentUpdated -= OnPlayerSpeechUpdate;
        fmodWhisperBridge.OnWhisperSegmentFinished -= OnPlayerSpeechFinished;
        fmodWhisperBridge.DeactivateRecording();

        // If currently not bringing food and the player leaves, then go back to idle state
        if (state != 3)
        {
            state = 0;
        }
    }

    private void OnPlayerSpeechFinished(string playerText)
    {
        if (Time.time < lastPlayerVoiceUpdateTime + 0.5f)
        {
            Debug.Log("voiceline: player said '" + playerText + "' but not enough time passed");
            return;
        }
        if (string.IsNullOrEmpty(playerText) || playerText.Contains("BLANK_AUDIO"))
        {
            Debug.Log("player speech event but result empty");
            return;
        }
        lastPlayerVoiceUpdateTime = Time.time;
        Debug.Log("state = " + state.ToString() + ", voiceline: player text: " + playerText);

        if (state == 1)
        {
            // Show transcription and ask whether it is correct
            notepadText.text = playerText;
            notepad.transform.DOLocalRotate(notepadFlippedRotation, 0.5f);
            SpeakVoiceLine(1);
            state = 2;
        } else if (state == 2)
        {
            // Flip notepad back
            notepad.transform.DOLocalRotate(notepadOriginalRotation, 0.5f);
            // if player answered positively, bring food, otherwise ask again
            if (playerText.ToLower().Contains("yes"))
            {
                SpeakVoiceLine(2);
                Invoke("BringFood", 1f);
                state = 3;
            } else
            {
                SpeakVoiceLine(3);
                state = 1;
            }
        }
    }

    private void OnPlayerSpeechUpdate(string playerText)
    {
        if (string.IsNullOrEmpty(playerText) || playerText.Contains("BLANK_AUDIO"))
        {
            return;
        }
        // Only update notepad text when currently listening to player order
        if (state == 1)
        {
            notepadText.text = playerText;
        }
    }

    private async void BringFood()
    {
        notepad.SetActive(false);
        transform.DOMove(backRoom.position, backRoomMovingTime);

        byte[] imageBytes = await InvokeAiClient.Instance.GenerateImage(notepadText.text);
        Texture2D GeneratedTexture = ModelGenerationUtils.CreateTexture(imageBytes);
        Sprite sprite = ModelGenerationUtils.CreateSprite(GeneratedTexture);
        string encodedTexture = Convert.ToBase64String(GeneratedTexture.EncodeToJPG());
        byte[] encodedModel = await TrellisClient.Instance.GenerateModel(encodedTexture);
        GameObject spawnedObject = await ModelGenerationUtils.Instance.SpawnModel(encodedModel);

        // Come back
        transform.DOMove(startingPosition, backRoomMovingTime).OnComplete(() =>
        {
            spawnedObject.transform.position = plate.position + new Vector3(0, 1f, 0);
            InitializeSpawnedObject(spawnedObject);

            SpeakVoiceLine(4);

            state = 4;
        });
    }

    private void InitializeSpawnedObject(GameObject spawnedObject)
    {
        Rigidbody rigidbody = spawnedObject.AddComponent<Rigidbody>();

        //spawnedObject.AddComponent<NetworkObject>();
        //spawnedObject.AddComponent<NetworkTransform>();

        MeshCollider spawnedObjectCollider = spawnedObject.GetComponent<MeshCollider>();
        spawnedObjectCollider.convex = true;
        spawnedObject.layer = ignorePlayerCollisionLayer;

        spawnedObject.AddComponent<XRGrabInteractable>();
    }
}