PSC_SharkTable/Plugins/BasemashVoiceCommander/Source/BasemashVoiceCommander/Public/VoiceCaptureComponent.h

// Copyright (c) Basemash. All Rights Reserved.

#pragma once

#include "AudioCaptureCore.h"
#include "Components/ActorComponent.h"
#include "CoreMinimal.h"
#include "HAL/ThreadSafeBool.h"

#include "VoiceCaptureComponent.generated.h"

class UPSVoiceAudioConsumer;
class IPixelStreaming2AudioSink;
class UVoiceCommandSubsystem;

/**
 * Audio source selection for UVoiceCaptureComponent.
 * LocalMic       - Capture from the default OS microphone via Audio::FAudioCapture.
 * PixelStreaming - Capture from a remote browser peer microphone via PS2 audio consumer.
 */
UENUM(BlueprintType)
enum class EVoiceAudioSource : uint8
{
    LocalMic       UMETA(DisplayName = "Local Microphone"),
    PixelStreaming UMETA(DisplayName = "Pixel Streaming (Browser Mic)")
};

/** Broadcast after successful transcription. Fired only on the component that initiated the request. */
DECLARE_DYNAMIC_MULTICAST_DELEGATE_OneParam(FOnVoiceCaptureTranscription, const FString&, Text);
/** Broadcast after LLM has parsed a command. Fired only on the component that initiated the request. */
DECLARE_DYNAMIC_MULTICAST_DELEGATE_TwoParams(
    FOnVoiceCaptureCommand, const FString&, Action, const FString&, ParamsJson);
/** Broadcast on any error during the voice pipeline. Fired only on the component that initiated the request. */
DECLARE_DYNAMIC_MULTICAST_DELEGATE_OneParam(FOnVoiceCaptureError, const FString&, ErrorMessage);

/**
 * UVoiceCaptureComponent
 *
 * Drop on any actor to give it voice-capture abilities. Handles:
 *   - Local mic or Pixel Streaming browser mic audio capture
 *   - Start/stop recording lifecycle with per-component event isolation
 *   - Stereo-to-mono mixdown, silence detection, WAV conversion
 *   - Round-trip to UVoiceCommandSubsystem (Groq STT + LLM)
 *
 * Transcription, parsed-command, and error events are delivered ONLY to the component
 * that initiated the request (via FVoiceRequestCallbacks) — multiple components can
 * coexist without cross-talk.
 */
UCLASS(ClassGroup = ("VoiceCommander"), meta = (BlueprintSpawnableComponent, DisplayName = "Voice Capture"))
class BASEMASHVOICECOMMANDER_API UVoiceCaptureComponent : public UActorComponent
{
    GENERATED_BODY()

public:
    UVoiceCaptureComponent();

    /** Which audio source to use at runtime. Change before BeginPlay for initial setup. */
    UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "Voice Commander")
    EVoiceAudioSource AudioSource = EVoiceAudioSource::LocalMic;

    /**
	 * Runtime-only per-component API key. Set via SetApiKey() or Blueprint at runtime.
	 * NOT serialized to asset — use Project Settings for a persistent default.
	 * If non-empty, takes priority over Subsystem runtime key and Settings default.
	 */
    UPROPERTY(Transient, BlueprintReadWrite, Category = "Voice Commander", meta = (PasswordField = true))
    FString ApiKeyOverride;

    /** Blueprint helper to set the per-component API key at runtime. */
    UFUNCTION(BlueprintCallable, Category = "Voice Commander")
    void SetApiKey(const FString& Key) { ApiKeyOverride = Key; }

    /** Silence threshold (max absolute float sample value). Below this, request is aborted with an error event. */
    UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "Voice Commander|Audio")
    float SilenceThreshold = 0.001f;

    /** If true, saves debug_recording.wav to ProjectSavedDir on every send. */
    UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "Voice Commander|Debug")
    bool bSaveDebugRecording = false;

    /** Begin capturing audio into the internal buffer. No-op if already recording. */
    UFUNCTION(BlueprintCallable, Category = "Voice Commander")
    void StartRecording();

    /** Stop capturing, perform silence check + WAV encode, dispatch to the subsystem. */
    UFUNCTION(BlueprintCallable, Category = "Voice Commander")
    void StopRecordingAndSend();

    /** True between StartRecording and StopRecordingAndSend. */
    [[nodiscard]] UFUNCTION(BlueprintPure, Category = "Voice Commander")
    bool IsRecording() const { return bIsRecording; }

    /**
	 * True if the configured audio source is ready to capture.
	 * - LocalMic: underlying Audio::FAudioCapture stream is open.
	 * - PixelStreaming: PS2 consumer is registered on an audio sink.
	 */
    [[nodiscard]] UFUNCTION(BlueprintPure, Category = "Voice Commander")
    bool IsAudioSourceReady() const;

    /** Events fired after Groq round-trip — scoped to THIS component only. */
    UPROPERTY(BlueprintAssignable, Category = "Voice Commander")
    FOnVoiceCaptureTranscription OnTranscription;

    UPROPERTY(BlueprintAssignable, Category = "Voice Commander")
    FOnVoiceCaptureCommand OnCommandParsed;

    UPROPERTY(BlueprintAssignable, Category = "Voice Commander")
    FOnVoiceCaptureError OnError;

protected:
    virtual void BeginPlay() override;
    virtual void EndPlay(const EEndPlayReason::Type EndPlayReason) override;

    // --- Local mic state ---
    Audio::FAudioCapture AudioCapture;
    int32                CaptureSampleRate  = 0;
    int32                CaptureNumChannels = 0;
    TArray<float>        RecordedPCMData;
    FCriticalSection     RecordingCriticalSection;

    // --- Pixel Streaming state ---
    UPROPERTY()
    UPSVoiceAudioConsumer* PSAudioConsumer = nullptr;

    TWeakPtr<IPixelStreaming2AudioSink> PSAudioSink;

    // Game-thread dominant, but read from the stale OnRemovedDelegate lambda path too.
    FThreadSafeBool bPSAudioRegistered = false;
    FTimerHandle    PSRetryTimerHandle;
    FString         CurrentAudioPlayerId;

    /** Handle for OnAudioTrackOpenNative so we can cleanly unsubscribe in EndPlay. */
    FDelegateHandle PSAudioTrackOpenHandle;

    // --- Recording state ---
    // Written on the game thread (Start/Stop), read on the audio thread (LocalMic capture callback).
    // FThreadSafeBool keeps the cross-thread read/write formally race-free without expanding the critical section.
    FThreadSafeBool bIsRecording = false;

    // --- Lifecycle ---
    void InitLocalMic();
    void CleanupLocalMic();
    void InitPixelStreamingAudio();
    void PSRetryRegistration();
    void CleanupPixelStreamingAudio();
    void SwitchAudioToPlayer(const FString& StreamerId, const FString& PlayerId);

    // --- Utility ---
    TArray<uint8> ConvertPCMToWav(const TArray<float>& PCMData, int32 SampleRate, int32 NumChannels) const;

    // --- Subsystem callback glue ---
    // Subsystem invokes these via FVoiceRequestCallbacks when this component is the request origin.
    // Each one simply re-broadcasts the corresponding multicast so listeners on this component only see events for their own requests.
    UFUNCTION()
    void HandleTranscription(const FString& Text);

    UFUNCTION()
    void HandleCommandParsed(const FString& Action, const FString& ParamsJson);

    UFUNCTION()
    void HandleError(const FString& ErrorMessage);
};