feat: merge audio

.gitignore

@@ -133,3 +133,4 @@ dist
 config.toml
 recordings
 .idea
+prisma/data.db*

package.json

@@ -4,7 +4,8 @@
   "description": "",
   "main": "dist/index.js",
   "scripts": {
-    "dev": "tsx src/index.ts"
+    "dev": "tsx src/index.ts",
+    "merge": "tsx src/merge.ts"
   },
   "keywords": [],
   "author": "",
@@ -23,10 +24,12 @@
     "@discordjs/rest": "^1.7.1",
     "@discordjs/voice": "^0.16.0",
     "@discordjs/ws": "^0.8.3",
+    "@prisma/client": "4.16.1",
     "discord-api-types": "^0.37.46",
     "libsodium-wrappers": "^0.7.11",
     "node-crc": "1.3.2",
     "prism-media": "2.0.0-alpha.0",
+    "prisma": "^4.16.1",
     "toml": "^3.0.0"
   }
 }

pnpm-lock.yaml

@@ -23,6 +23,9 @@ dependencies:
   '@discordjs/ws':
     specifier: ^0.8.3
     version: 0.8.3
+  '@prisma/client':
+    specifier: 4.16.1
+    version: 4.16.1(prisma@4.16.1)
   discord-api-types:
     specifier: ^0.37.46
     version: 0.37.46
@@ -35,6 +38,9 @@ dependencies:
   prism-media:
     specifier: 2.0.0-alpha.0
     version: 2.0.0-alpha.0
+  prisma:
+    specifier: ^4.16.1
+    version: 4.16.1
   toml:
     specifier: ^3.0.0
     version: 3.0.0
@@ -379,6 +385,29 @@ packages:
     dev: true
     optional: true
 
+  /@prisma/client@4.16.1(prisma@4.16.1):
+    resolution: {integrity: sha512-CoDHu7Bt+NuDo40ijoeHP79EHtECsPBTy3yte5Yo3op8TqXt/kV0OT5OrsWewKvQGKFMHhYQ+ePed3zzjYdGAw==}
+    engines: {node: '>=14.17'}
+    requiresBuild: true
+    peerDependencies:
+      prisma: '*'
+    peerDependenciesMeta:
+      prisma:
+        optional: true
+    dependencies:
+      '@prisma/engines-version': 4.16.0-66.b20ead4d3ab9e78ac112966e242ded703f4a052c
+      prisma: 4.16.1
+    dev: false
+
+  /@prisma/engines-version@4.16.0-66.b20ead4d3ab9e78ac112966e242ded703f4a052c:
+    resolution: {integrity: sha512-tMWAF/qF00fbUH1HB4Yjmz6bjh7fzkb7Y3NRoUfMlHu6V+O45MGvqwYxqwBjn1BIUXkl3r04W351D4qdJjrgvA==}
+    dev: false
+
+  /@prisma/engines@4.16.1:
+    resolution: {integrity: sha512-gpZG0kGGxfemgvK/LghHdBIz+crHkZjzszja94xp4oytpsXrgt/Ice82MvPsWMleVIniKuARrowtsIsim0PFJQ==}
+    requiresBuild: true
+    dev: false
+
   /@sapphire/async-queue@1.5.0:
     resolution: {integrity: sha512-JkLdIsP8fPAdh9ZZjrbHWR/+mZj0wvKS5ICibcLrRI1j84UmLMshx5n9QmL8b95d4onJ2xxiyugTgSAX7AalmA==}
     engines: {node: '>=v14.0.0', npm: '>=7.0.0'}
@@ -803,6 +832,15 @@ packages:
       duplex-child-process: 1.0.1
     dev: false
 
+  /prisma@4.16.1:
+    resolution: {integrity: sha512-C2Xm7yxHxjFjjscBEW4tmoraPHH/Vyu/A0XABdbaFtoiOZARsxvOM7rwc2iZ0qVxbh0bGBGBWZUSXO/52/nHBQ==}
+    engines: {node: '>=14.17'}
+    hasBin: true
+    requiresBuild: true
+    dependencies:
+      '@prisma/engines': 4.16.1
+    dev: false
+
   /readable-stream@3.6.2:
     resolution: {integrity: sha512-9u/sniCrY3D5WdsERHzHE4G2YCXqoG5FTHUiCC4SIbr6XcLZBY05ya9EKjYek9O5xOAwjGq+1JdGBAS7Q9ScoA==}
     engines: {node: '>= 6'}

prisma/migrations/20230628145216_init/migration.sql

@@ -0,0 +1,15 @@
+-- CreateTable
+CREATE TABLE "Session" (
+    "id" TEXT NOT NULL PRIMARY KEY,
+    "createdAt" DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    "channel" TEXT NOT NULL
+);
+
+-- CreateTable
+CREATE TABLE "File" (
+    "id" TEXT NOT NULL PRIMARY KEY,
+    "filename" TEXT NOT NULL,
+    "timestamp" BIGINT NOT NULL,
+    "sessionId" TEXT,
+    CONSTRAINT "File_sessionId_fkey" FOREIGN KEY ("sessionId") REFERENCES "Session" ("id") ON DELETE SET NULL ON UPDATE CASCADE
+);

prisma/migrations/20230628154831_wow/migration.sql

@@ -0,0 +1,20 @@
+/*
+  Warnings:
+
+  - Made the column `sessionId` on table `File` required. This step will fail if there are existing NULL values in that column.
+
+*/
+-- RedefineTables
+PRAGMA foreign_keys=OFF;
+CREATE TABLE "new_File" (
+    "id" TEXT NOT NULL PRIMARY KEY,
+    "filename" TEXT NOT NULL,
+    "timestamp" BIGINT NOT NULL,
+    "sessionId" TEXT NOT NULL,
+    CONSTRAINT "File_sessionId_fkey" FOREIGN KEY ("sessionId") REFERENCES "Session" ("id") ON DELETE CASCADE ON UPDATE CASCADE
+);
+INSERT INTO "new_File" ("filename", "id", "sessionId", "timestamp") SELECT "filename", "id", "sessionId", "timestamp" FROM "File";
+DROP TABLE "File";
+ALTER TABLE "new_File" RENAME TO "File";
+PRAGMA foreign_key_check;
+PRAGMA foreign_keys=ON;

prisma/migrations/migration_lock.toml

@@ -0,0 +1,3 @@
+# Please do not edit this file manually
+# It should be added in your version-control system (i.e. Git)
+provider = "sqlite"

prisma/schema.prisma

@@ -0,0 +1,27 @@
+// This is your Prisma schema file,
+// learn more about it in the docs: https://pris.ly/d/prisma-schema
+
+generator client {
+  provider = "prisma-client-js"
+}
+
+datasource db {
+  provider = "sqlite"
+  url      = "file:data.db"
+}
+
+model Session {
+  id         String      @id @default(uuid())
+  createdAt  DateTime    @default(now())
+  channel    String
+  files File[]
+}
+
+model File {
+  id String @id @default(uuid())
+  filename String
+  timestamp BigInt
+
+  session   Session @relation(fields: [sessionId], references: [id], onDelete: Cascade)
+  sessionId String
+}

src/api.ts

@@ -4,6 +4,7 @@ import { Client, GatewayIntentBits } from "@discordjs/core"
 import { config } from "./config.js"
 import { VoiceStateManager } from "./utils/wrapper/VoiceState.js"
 import { RecordingSessionManager } from "./utils/sessions.js"
+import { PrismaClient } from "@prisma/client"
 
 export const rest = new REST({ version: "10" }).setToken(config.token)
 
@@ -20,3 +21,5 @@ export const client = new Client({ rest, gateway })
 export const voiceStateManager = new VoiceStateManager()
 
 export const sessionManager = new RecordingSessionManager()
+
+export const db = new PrismaClient()

src/commands/start.ts

@@ -34,7 +34,7 @@ export const startCommand = async ({
 			content: "먼저 음성 채널에 들어가주세요!",
 		})
 
-	sessionManager.addSession(data.guild_id, vc, shardId)
+	await sessionManager.addSession(data.guild_id, vc, data.channel_id, shardId)
 
 	await api.interactions.reply(data.id, data.token, {
 		content: "냥냥",

src/index.ts

@@ -13,6 +13,7 @@ import { startCommand } from "./commands/start.js"
 import { config } from "./config.js"
 import { syncCommands } from "./utils/commands.js"
 import { VoiceState } from "./utils/wrapper/VoiceState.js"
+import { isMe } from "./utils/appid.js"
 
 process.on("uncaughtException", (error) => console.error(error))
 process.on("unhandledRejection", (error) => console.error(error))
@@ -82,20 +83,41 @@ client.on(GatewayDispatchEvents.InteractionCreate, async (args) => {
 	}
 })
 
-client.on(GatewayDispatchEvents.VoiceStateUpdate, async (args) => {
+client.on(GatewayDispatchEvents.VoiceStateUpdate, async ({ data, api }) => {
-	if (args.data.guild_id) {
+	if (data.guild_id) {
-		const adapter = sessionManager.adapters.get(args.data.guild_id)
+		const adapter = sessionManager.adapters.get(data.guild_id)
-		if (adapter) adapter.onVoiceStateUpdate(args.data)
+		if (adapter) adapter.onVoiceStateUpdate(data)
-		const guild = voiceStateManager.guilds.get(args.data.guild_id)
+		const guild = voiceStateManager.guilds.get(data.guild_id)
 		if (guild) {
-			const state = guild.get(args.data.user_id)
+			const state = guild.get(data.user_id)
 			if (state) {
-				state.update(args.data)
+				state.update(data)
 			} else {
-				guild.set(
+				guild.set(data.user_id, new VoiceState(data, data.guild_id))
-					args.data.user_id,
+			}
-					new VoiceState(args.data, args.data.guild_id)
+
-				)
+			const session = sessionManager.sessionMap.get(data.guild_id)
+
+			if (session) {
+				let counter = 0
+				let hasMe = false
+
+				for (const state of guild.values()) {
+					if (state.channelId !== session.voiceChannelId) continue
+					if (isMe(state.id)) {
+						hasMe = true
+						continue
+					}
+					counter++
+				}
+
+				if (counter === 0 || !hasMe) {
+					session.vc.destroy()
+					sessionManager.removeSession(data.guild_id)
+					await api.channels.createMessage(session.textChannelId, {
+						content: `${session.id}`,
+					})
+				}
 			}
 		}
 	}

src/merge.ts

@@ -0,0 +1,12 @@
+import { db } from "./api.js"
+import { mergeAudio } from "./utils/merge.js"
+
+const id = process.argv[2]
+
+const session = await db.session.findUnique({ where: { id } })
+
+if (!session) {
+	throw new Error("session not found")
+}
+
+await mergeAudio(id)

src/utils/appid.ts

@@ -1,3 +1,8 @@
+import { config } from "../config.js"
+
 export const getAppIdFromToken = (token: string) => {
 	return atob(token.split(".")[0])
 }
+
+export const isMe = (userId: string) =>
+	userId === getAppIdFromToken(config.token)

src/utils/merge.ts

@@ -0,0 +1,65 @@
+import path from "path"
+import { db } from "../api.js"
+import { spawn } from "child_process"
+import { existsSync, mkdirSync } from "fs"
+
+export const mergeAudio = async (id: string) => {
+	const finalDir = "recordings/final"
+	const sourceDir = path.resolve("recordings", id)
+
+	if (!existsSync(finalDir)) {
+		mkdirSync(finalDir, { recursive: true })
+	}
+
+	const files = await db.file.findMany({
+		where: {
+			session: {
+				id,
+			},
+		},
+	})
+
+	let inputs: string[] = []
+
+	let filter: string[] = []
+	let trackNames: string[] = []
+
+	let counter = 0
+
+	for (const file of files) {
+		const filePath = path.join(sourceDir, file.filename)
+		const index = counter++
+		const trackName = `[c${index}]`
+
+		inputs.push("-i", filePath)
+
+		filter.push(
+			`[${index}:a] adelay=${file.timestamp}|${file.timestamp} ${trackName}`
+		)
+
+		trackNames.push(trackName)
+	}
+
+	filter.push(
+		`${trackNames.join(" ")} amix=inputs=${
+			trackNames.length
+		}:normalize=0 [out]`
+	)
+
+	const args = [
+		...inputs,
+		"-filter_complex",
+		filter.join(";"),
+		"-map",
+		"[out]",
+		"-async",
+		"1",
+		"-y",
+		path.join(finalDir, `${id}.ogg`),
+	]
+
+	const p = spawn("ffmpeg", args).on("exit", (code) => {
+		console.log("ffmpeg exited with code", code)
+	})
+	p.stderr.pipe(process.stderr)
+}

src/utils/sessions.ts

@@ -1,5 +1,5 @@
 import { Collection } from "@discordjs/collection"
-import { gateway, voiceStateManager } from "../api.js"
+import { db, gateway, voiceStateManager } from "../api.js"
 import {
 	AudioReceiveStream,
 	DiscordGatewayAdapterLibraryMethods,
@@ -11,9 +11,9 @@ import { Snowflake } from "discord-api-types/v10"
 import * as prism from "prism-media"
 import { pipeline } from "stream"
 import { createWriteStream, existsSync, mkdirSync } from "fs"
-import { getAppIdFromToken } from "./appid.js"
-import { config } from "../config.js"
 import path from "path"
+import { isMe } from "./appid.js"
+import { exec, spawn } from "child_process"
 
 export class RecordingSessionManager {
 	sessionMap = new Collection<Snowflake, RecordingSession>()
@@ -23,11 +23,16 @@ export class RecordingSessionManager {
 		return this.sessionMap.has(guild)
 	}
 
-	addSession(guild: Snowflake, channel: Snowflake, shardId: number) {
+	async addSession(
+		guild: Snowflake,
+		voiceChannel: Snowflake,
+		textChannel: Snowflake,
+		shardId: number
+	) {
 		const session = new RecordingSession(
 			joinVoiceChannel({
 				guildId: guild,
-				channelId: channel,
+				channelId: voiceChannel,
 				adapterCreator: (methods) => {
 					this.adapters.set(guild, methods)
 					return {
@@ -43,22 +48,48 @@ export class RecordingSessionManager {
 				selfDeaf: false,
 				selfMute: true,
 				debug: true,
-			})
+			}),
+			voiceChannel,
+			textChannel
 		)
 
+		await session.init()
+
 		this.sessionMap.set(guild, session)
 	}
+
+	removeSession(guild: Snowflake) {
+		const session = this.sessionMap.get(guild)
+		if (session) session.shutdown()
+		this.sessionMap.delete(guild)
+	}
 }
 
 export class RecordingSession {
-	id: string
 	streams = new Collection<Snowflake, UserStream>()
 	startedAt = Date.now()
+	id!: string
+
+	constructor(
+		public vc: VoiceConnection,
+		public voiceChannelId: Snowflake,
+		public textChannelId: Snowflake
+	) {}
+
+	async init() {
+		const result = await db.session.create({
+			data: {
+				id: this.id,
+				channel: this.voiceChannelId,
+			},
+		})
+		this.id = result.id
 
-	constructor(public vc: VoiceConnection) {
 		this.id = Date.now().toString()
-		vc.receiver.speaking.on("start", (userId) => this.addReceiver(userId))
+		this.receiver.speaking.on("start", (userId) => this.addReceiver(userId))
-		vc.receiver.speaking.on("end", (userId) => this.removeReceiver(userId))
+		this.receiver.speaking.on("end", (userId) =>
+			this.removeReceiver(userId)
+		)
 	}
 
 	get receiver() {
@@ -66,11 +97,9 @@ export class RecordingSession {
 	}
 
 	addReceiver(userId: string) {
-		if (userId === getAppIdFromToken(config.token)) return
+		if (isMe(userId)) return
 		if (this.streams.has(userId)) return
 
-		console.log("Starting record:", userId)
-
 		const receiver = this.receiver.subscribe(userId)
 
 		this.streams.set(userId, new UserStream(this, receiver, userId))
@@ -80,11 +109,12 @@ export class RecordingSession {
 		const stream = this.streams.get(userId)
 
 		if (stream) {
-			console.log("Stopping record:", userId)
 			stream.subscription.push(null)
 			this.streams.delete(userId)
 		}
 	}
+
+	async shutdown() {}
 }
 
 class UserStream {
@@ -105,7 +135,9 @@ class UserStream {
 			},
 		})
 
-		const dir = path.resolve("recordings", session.id, userId)
+		const dir = path.resolve("recordings", session.id)
+		const timestamp = Date.now() - session.startedAt
+		const file = `${userId}-${timestamp}.ogg`
 
 		if (!existsSync(dir)) {
 			mkdirSync(dir, {
@@ -113,20 +145,36 @@ class UserStream {
 			})
 		}
 
-		const filename = path.join(dir, `${Date.now() - session.startedAt}.ogg`)
+		const filename = path.join(dir, file)
+
+		console.log(`Recording ${filename} started`)
 
 		const outputStream = createWriteStream(filename)
 
-		pipeline(subscription, this.oggStream, outputStream, (err) => {
+		pipeline(subscription, this.oggStream, outputStream, async (err) => {
 			outputStream.end()
 
 			if (err) {
-				console.log(`Recording ${userId} ended with error: ${err}`)
+				console.log(`Recording ${filename} ended with error: ${err}`)
 				subscription.destroy(err)
 				return
 			}
 
-			console.log(`Recording ${userId} ended`)
+			await db.file
+				.create({
+					data: {
+						session: {
+							connect: {
+								id: session.id,
+							},
+						},
+						filename: file,
+						timestamp,
+					},
+				})
+				.catch(console.error)
+
+			console.log(`Recording ${filename} ended`)
 		})
 	}
 }