Add TTS

discord/.env.example

@@ -4,8 +4,7 @@ CLIENT="123456789012345678"
 GUILD="123456789012345678"
 ADMIN="123456789012345678"
 
-LLM_HOST="127.0.0.1"
+LLM_HOST="http://127.0.0.1:8000"
-LLM_PORT=8000
 LLM_TOKEN="dfsl;kjsdl;kfja"
 REPLY_CHANCE=0.2
 

discord/bot.ts

@@ -32,6 +32,7 @@ import {
     openDb,
     reactionEmojis,
     recordReaction,
+    requestTTSResponse,
     sync
 } from './util';
 import 'dotenv/config';
@@ -166,11 +167,16 @@ async function onNewMessage(message: Message)
 
 async function fetchMotd()
 {
-    const res = await fetch(process.env.MOTD_HREF);
+    try {
-    const xml = await res.text();
+	    const res = await fetch(process.env.MOTD_HREF);
-    const parser = new JSDOM(xml);
+	    const xml = await res.text();
-    const doc = parser.window.document;
+	    const parser = new JSDOM(xml);
-    return doc.querySelector(process.env.MOTD_QUERY).textContent;
+	    const doc = parser.window.document;
+	    const el = doc.querySelector(process.env.MOTD_QUERY);
+	    return el ? el.textContent : null;
+    } catch (err) {
+	    logWarn('[bot] Failed to fetch MOTD; is the booru down?');
+    }
 }
 
 async function requestRVCResponse(src: Attachment): Promise<Blob>
@@ -189,7 +195,7 @@ async function requestRVCResponse(src: Attachment): Promise<Blob>
     const fd = new FormData();
     fd.append('file', fs.readFileSync(tmpFileName), 'voice-message.ogg');
 
-    const rvcEndpoint = `http://${process.env.LLM_HOST}:${process.env.LLM_PORT}/rvc?${queryParams.toString()}`;
+    const rvcEndpoint = `${process.env.LLM_HOST}/rvc?${queryParams.toString()}`;
     logInfo(`[bot] Requesting RVC response for ${src.id}`);
     const res = await fetch(rvcEndpoint, {
         method: 'POST',
@@ -206,7 +212,7 @@ async function requestLLMResponse(messages)
     for (const field of Object.keys(config["llmconf"].llmSettings)) {
         queryParams.append(field, config["llmconf"].llmSettings[field]);
     }
-    const llmEndpoint = `http://${process.env.LLM_HOST}:${process.env.LLM_PORT}/?${queryParams.toString()}`;
+    const llmEndpoint = `${process.env.LLM_HOST}/?${queryParams.toString()}`;
     const messageList = messages.map((m: Message) => ({
         role: m.author.bot ? "assistant" : "user",
         content: m.cleanContent,
@@ -244,8 +250,18 @@ async function scheduleRandomMessage(firstTime = false)
             return;
         }
         const randomMessage = await fetchMotd();
-        await channel.send(randomMessage);
+	if (randomMessage) {
-        logInfo(`[bot] Sent MOTD: ${randomMessage}`);
+            const audio = await requestTTSResponse(randomMessage);
+            const audioBuf = await audio.arrayBuffer();
+            const audioFile = new AttachmentBuilder(Buffer.from(audioBuf)).setName('mikuified.wav');
+            await channel.send({
+		content: randomMessage,
+                files: [audioFile]
+            });
+	    logInfo(`[bot] Sent MOTD: ${randomMessage}`);
+	} else {
+	    logWarn(`[bot] Could not fetch MOTD.`);
+	}
     }
     // wait between 2-8 hours
     const timeoutMins = Math.random() * 360 + 120;

discord/commands/tts/tts.ts

@@ -0,0 +1,43 @@
+import {
+    AttachmentBuilder,
+    ChatInputCommandInteraction,
+    SlashCommandBuilder
+} from 'discord.js';
+import 'dotenv/config';
+import { logError, logInfo, logWarn } from '../../../logging';
+import { requestTTSResponse } from '../../util';
+
+const config = {
+    ttsSettings: {
+        pitch_change_oct: 1,
+	pitch_change_sem: 0
+    }
+};
+
+async function ttsCommand(interaction: ChatInputCommandInteraction)
+{
+    const text = interaction.options.getString('text');
+    await interaction.reply(`generating audio for "${text}"...`);
+    try {
+	    const audio = await requestTTSResponse(text);
+	    const audioBuf = await audio.arrayBuffer();
+	    const audioFile = new AttachmentBuilder(Buffer.from(audioBuf)).setName('mikuified.wav');
+	    await interaction.editReply({
+		files: [audioFile]
+	    });
+    } catch (err) {
+	await interaction.editReply(`Error: ${err}`);
+        logError(`Error while generating TTS: ${err}`);
+    }
+}
+
+export = {
+    data: new SlashCommandBuilder()
+        .setName('tts')
+        .setDescription('Read text in Miku\'s voice')
+        .addStringOption(
+            opt => opt.setName('text').setDescription('Text').setRequired(true)
+        ),
+    execute: ttsCommand,
+    config: config
+};

discord/util.ts

@@ -9,6 +9,7 @@ import { get as httpGet } from 'https';
 import { Database, open } from 'sqlite';
 import { Database as Database3 } from 'sqlite3';
 import 'dotenv/config';
+import fetch from 'node-fetch';
 import { logError, logInfo, logWarn } from '../logging';
 import { ScoreboardMessageRow } from '../models';
 
@@ -163,4 +164,19 @@ async function sync(guilds: GuildManager) {
     }
 }
 
-export { db, clearDb, openDb, reactionEmojis, recordReaction, sync };
+async function requestTTSResponse(txt: string): Promise<Blob>
+{
+    const queryParams = new URLSearchParams();
+    queryParams.append("token", process.env.LLM_TOKEN);
+    queryParams.append("text", txt);
+
+    const ttsEndpoint = `${process.env.LLM_HOST}/tts?${queryParams.toString()}`;
+    logInfo(`[bot] Requesting TTS response for "${txt}"`);
+    const res = await fetch(ttsEndpoint, {
+        method: 'POST'
+    });
+    const resContents = await res.blob();
+    return resContents;
+}
+
+export { db, clearDb, openDb, reactionEmojis, recordReaction, requestTTSResponse, sync };