feat(core): notification transcript

Signed-off-by: rhunk <101876869+rhunk@users.noreply.github.com>

core/src/main/kotlin/me/rhunk/snapenhance/core/features/impl/experiments/BetterTranscript.kt

@@ -4,20 +4,47 @@ import me.rhunk.snapenhance.common.data.ContentType
 import me.rhunk.snapenhance.common.util.protobuf.ProtoEditor
 import me.rhunk.snapenhance.core.event.events.impl.BuildMessageEvent
 import me.rhunk.snapenhance.core.features.Feature
+import me.rhunk.snapenhance.core.util.dataBuilder
 import me.rhunk.snapenhance.core.util.hook.HookStage
 import me.rhunk.snapenhance.core.util.hook.hook
 import me.rhunk.snapenhance.core.util.ktx.getObjectFieldOrNull
 import me.rhunk.snapenhance.core.util.ktx.setObjectField
+import java.nio.ByteBuffer
 
 class BetterTranscript: Feature("Better Transcript") {
+    private val voiceML: Any by lazy {
+        findClass("com.snapchat.client.voiceml.IVoiceMLSDK").getMethod("create").invoke(null) ?: error("Failed to create IVoiceMLSDK instance")
+    }
+
+    private fun createAsrConfig(): Any? {
+        findClass("com.snapchat.client.voiceml.IConfigFactory").methods.first { it.name == "simpleAsrConfig" }.let { method ->
+            return method.invoke(null, method.parameterTypes[0].dataBuilder {
+                set("mSampleRate", 44100)
+                set("mLanguageModel", "en")
+                set("mUseCase", "VOICENOTESTRANSCRIPTION")
+                set("mAppVersion", "voice note transcript")
+                set("mUiLanguage", "en")
+                set("mAuthType", "SNAPTOKEN")
+                set("mEncoding", "AAC")
+            })
+        }
+    }
+
+    fun transcribe(audio: ByteBuffer): String? {
+        val transcribeMethod = voiceML.javaClass.methods.first { it.name == "asrTranscribe" }
+        val snapToken = context.database.getAccessTokens(context.database.myUserId)?.get("api-gateway") ?: error("Failed to get api-gateway token")
+
+        return transcribeMethod.invoke(voiceML, snapToken, createAsrConfig(), audio)
+            ?.let { asrResult ->
+                asrResult.getObjectFieldOrNull("mTranscription")?.toString()
+            }
+    }
+
     override fun init() {
         if (context.config.experimental.betterTranscript.globalState != true) return
 
         onNextActivityCreate {
             val config = context.config.experimental.betterTranscript
-            val preferredTranscriptionLang = config.preferredTranscriptionLang.getNullable()?.takeIf {
-                it.isNotBlank()
-            }
 
             if (config.forceTranscription.get()) {
                 context.event.subscribe(BuildMessageEvent::class, priority = 104) { event ->
@@ -33,7 +60,9 @@ class BetterTranscript: Feature("Better Transcript") {
             }
 
             findClass("com.snapchat.client.voiceml.IVoiceMLSDK\$CppProxy").hook("asrTranscribe", HookStage.BEFORE) { param ->
-                preferredTranscriptionLang?.lowercase()?.let {
+                config.preferredTranscriptionLang.getNullable()?.takeIf {
+                    it.isNotBlank()
+                }?.trim()?.lowercase()?.let {
                     val asrConfig = param.arg<Any>(1)
                     asrConfig.getObjectFieldOrNull("mBaseConfig")?.apply {
                         setObjectField("mLanguageModel", it)

core/src/main/kotlin/me/rhunk/snapenhance/core/features/impl/messaging/Notifications.kt

@@ -24,7 +24,9 @@ import me.rhunk.snapenhance.core.event.events.impl.SnapWidgetBroadcastReceiveEve
 import me.rhunk.snapenhance.core.features.Feature
 import me.rhunk.snapenhance.core.features.impl.FriendMutationObserver
 import me.rhunk.snapenhance.core.features.impl.downloader.MediaDownloader
+import me.rhunk.snapenhance.core.features.impl.downloader.decoder.AttachmentType
 import me.rhunk.snapenhance.core.features.impl.downloader.decoder.MessageDecoder
+import me.rhunk.snapenhance.core.features.impl.experiments.BetterTranscript
 import me.rhunk.snapenhance.core.features.impl.spying.StealthMode
 import me.rhunk.snapenhance.core.util.hook.HookStage
 import me.rhunk.snapenhance.core.util.hook.findRestrictedConstructor
@@ -34,6 +36,7 @@ import me.rhunk.snapenhance.core.util.ktx.setObjectField
 import me.rhunk.snapenhance.core.util.media.PreviewUtils
 import me.rhunk.snapenhance.core.wrapper.impl.Message
 import me.rhunk.snapenhance.core.wrapper.impl.SnapUUID
+import java.nio.ByteBuffer
 import kotlin.coroutines.suspendCoroutine
 
 class Notifications : Feature("Notifications") {
@@ -387,6 +390,29 @@ class Notifications : Feature("Notifications") {
                 }"
             }
 
+            if (contentType == ContentType.NOTE && context.config.experimental.betterTranscript.takeIf { it.globalState == true }?.notificationTranscript?.get() == true) {
+                MessageDecoder.decode(message.messageContent!!).firstOrNull { it.type == AttachmentType.NOTE  }?.also { media ->
+                    runCatching {
+                        media.openStream { mediaStream, length ->
+                            if (mediaStream == null || length > 25 * 1024 * 1024) {
+                                context.log.error("Failed to open media stream or media is too large")
+                                return@openStream
+                            }
+
+                            val text = context.feature(BetterTranscript::class).transcribe(
+                                ByteBuffer.allocateDirect(length.toInt()).apply {
+                                    put(mediaStream.readBytes())
+                                    rewind()
+                                }) ?: return@openStream
+                            serializedMessage = "\uD83C\uDFA4 $text"
+                            isChatMessage = true
+                        }
+                    }.onFailure {
+                        context.log.error("Failed to transcribe message", it)
+                    }
+                }
+            }
+
             if (isChatMessage || config.stackedMediaMessages.get()) {
                 setNotificationText(serializedMessage)
             } else {