Merge branch 'main' of https://github.com/kwaroran/RisuAI

server/node/server.cjs

@@ -3,12 +3,11 @@ const app = express();
 const path = require('path');
 const htmlparser = require('node-html-parser');
 const { existsSync, mkdirSync, readFileSync, writeFileSync } = require('fs');
-const bodyParser = require('body-parser');
 const fs = require('fs/promises')
 const crypto = require('crypto')
 app.use(express.static(path.join(process.cwd(), 'dist'), {index: false}));
-app.use(bodyParser.raw({ limit: 100000000 }));
+app.use(express.json({ limit: '50mb' }));
-app.use(bodyParser.json())
+app.use(express.raw({ type: 'application/octet-stream', limit: '50mb' }));
 const {pipeline} = require('stream/promises')
 
 let password = ''
@@ -90,6 +89,55 @@ const reverseProxyFunc = async (req, res, next) => {
     }
 }
 
+const reverseProxyFunc_get = async (req, res, next) => {
+    const urlParam = req.headers['risu-url'] ? decodeURIComponent(req.headers['risu-url']) : req.query.url;
+
+    if (!urlParam) {
+        res.status(400).send({
+            error:'URL has no param'
+        });
+        return;
+    }
+    const header = req.headers['risu-header'] ? JSON.parse(decodeURIComponent(req.headers['risu-header'])) : req.headers;
+    if(!header['x-forwarded-for']){
+        header['x-forwarded-for'] = req.ip
+    }
+    let originalResponse;
+    try {
+        // make request to original server
+        originalResponse = await fetch(urlParam, {
+            method: 'GET',
+            headers: header
+        });
+        // get response body as stream
+        const originalBody = originalResponse.body;
+        // get response headers
+        const head = new Headers(originalResponse.headers);
+        head.delete('content-security-policy');
+        head.delete('content-security-policy-report-only');
+        head.delete('clear-site-data');
+        head.delete('Cache-Control');
+        head.delete('Content-Encoding');
+        const headObj = {};
+        for (let [k, v] of head) {
+            headObj[k] = v;
+        }
+        // send response headers to client
+        res.header(headObj);
+        // send response status to client
+        res.status(originalResponse.status);
+        // send response body to client
+        await pipeline(originalResponse.body, res);
+    }
+    catch (err) {
+        next(err);
+        return;
+    }
+}
+
+app.get('/proxy', reverseProxyFunc_get);
+app.get('/proxy2', reverseProxyFunc_get);
+
 app.post('/proxy', reverseProxyFunc);
 app.post('/proxy2', reverseProxyFunc);
 

src/lib/Setting/Pages/LanguageSettings.svelte

@@ -117,6 +117,21 @@
         <TextAreaInput bind:value={$DataBase.translatorPrompt} placeholder={"You are a translator. translate the following html or text into {{slot}}. do not output anything other than the translation."}/>
     {/if}
 
+    {#if $DataBase.translatorType === 'google'}
+        <span class="text-textcolor mt-4">Translator Input Language</span>
+        <SelectInput className="mt-2 mb-4" bind:value={$DataBase.translatorInputLanguage}>
+            <OptionInput value="auto">Auto</OptionInput>
+            <OptionInput value="en">English</OptionInput>
+            <OptionInput value="zh">Chinese</OptionInput>
+            <OptionInput value="ja">Japanese</OptionInput>
+            <OptionInput value="ko">Korean</OptionInput>
+            <OptionInput value="fr">French</OptionInput>
+            <OptionInput value="es">Spanish</OptionInput>
+            <OptionInput value="de">German</OptionInput>
+            <OptionInput value="ru">Russian</OptionInput>
+        </SelectInput>
+    {/if}
+
 
     <div class="flex items-center mt-2">
         <Check bind:check={$DataBase.autoTranslate} name={language.autoTranslation}/>

src/lib/Setting/Pages/OtherBotSettings.svelte

@@ -13,6 +13,7 @@
     import Button from "src/lib/UI/GUI/Button.svelte";
     import { getCharImage } from "src/ts/characters";
     import Arcodion from "src/lib/UI/Arcodion.svelte";
+  import CheckInput from "src/lib/UI/GUI/CheckInput.svelte";
     $:{
         $DataBase.NAIImgConfig ??= {
             width: 512,
@@ -299,6 +300,9 @@
 
 {#if submenu === 1 || submenu === -1}
 <Arcodion name="TTS" styled disabled={submenu !== -1}>
+    <span class="text-textcolor mt-2">Auto Speech</span>
+    <CheckInput bind:check={$DataBase.ttsAutoSpeech}/>
+
     <span class="text-textcolor mt-2">ElevenLabs API key</span>
     <TextInput size="sm" marginBottom bind:value={$DataBase.elevenLabKey}/>
     

src/lib/SideBars/CharConfig.svelte

@@ -112,6 +112,13 @@
             if (!(currentChar.data as character).gptSoVitsConfig.use_prompt) {
                 (currentChar.data as character).gptSoVitsConfig.prompt = undefined
             }
+            if((currentChar.data as character).gptSoVitsConfig.use_auto_path){
+                (currentChar.data as character).gptSoVitsConfig.ref_audio_path = undefined;
+
+                (currentChar.data as character).gptSoVitsConfig.use_prompt = false;
+                (currentChar.data as character).gptSoVitsConfig.prompt = undefined;
+
+            }
         }
     })
 
@@ -160,7 +167,9 @@
     $: if (currentChar.data.ttsMode === 'gptsovits' && (currentChar.data as character).gptSoVitsConfig === undefined) {
         (currentChar.data as character).gptSoVitsConfig = {
             url: '',
-            ref_audio_path: 'C:/Users/user/Downloads/GPT-SoVITS-v2-240821',
+            use_auto_path: false,
+            ref_audio_path: '',
+            use_long_audio: false,
             ref_audio_data: {
                 fileName: '',
                 assetId: ''  
@@ -835,15 +844,24 @@
             <span class="text-textcolor">URL</span>
             <TextInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.url}/>
 
+            <span class="text-textcolor">Use Auto Path</span>
+            <Check bind:check={currentChar.data.gptSoVitsConfig.use_auto_path}/>
+
+            {#if !currentChar.data.gptSoVitsConfig.use_auto_path}
                 <span class="text-textcolor">Reference Audio Path (e.g. C:/Users/user/Downloads/GPT-SoVITS-v2-240821)</span>
                 <TextInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.ref_audio_path}/>
+            {/if}
+
+            <span class="text-textcolor">Use Long Audio</span>
+            <Check bind:check={currentChar.data.gptSoVitsConfig.use_long_audio}/>
 
             <span class="text-textcolor">Reference Audio Data (3~10s audio file)</span>
             <Button on:click={async () => {
                 const audio = await selectSingleFile([
                     'wav',
                     'ogg',
-                    'aac'
+                    'aac',
+                    'mp3'
                 ])
                 if(!audio){
                     return null
@@ -866,40 +884,42 @@
             </Button>
             <span class="text-textcolor">Text Language</span>
             <SelectInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.text_lang}>
-                <OptionInput value="auto">Auto</OptionInput>
+                <OptionInput value="auto">Multi-language Mixed</OptionInput>
-                <OptionInput value="auto_yue">Auto (Cantonese)</OptionInput>
+                <OptionInput value="auto_yue">Multi-language Mixed (Cantonese)</OptionInput>
                 <OptionInput value="en">English</OptionInput>
-                <OptionInput value="zh">Chinese</OptionInput>
+                <OptionInput value="zh">Chinese-English Mixed</OptionInput>
-                <OptionInput value="ja">Japanese</OptionInput>
+                <OptionInput value="ja">Japanese-English Mixed</OptionInput>
-                <OptionInput value="yue">Cantonese</OptionInput>
+                <OptionInput value="yue">Cantonese-English Mixed</OptionInput>
-                <OptionInput value="ko">Korean</OptionInput>
+                <OptionInput value="ko">Korean-English Mixed</OptionInput>
-                <OptionInput value="all_zh">All Chinese</OptionInput>
+                <OptionInput value="all_zh">Chinese</OptionInput>
-                <OptionInput value="all_ja">All Japanese</OptionInput>
+                <OptionInput value="all_ja">Japanese</OptionInput>
-                <OptionInput value="all_yue">All Cantonese</OptionInput>
+                <OptionInput value="all_yue">Cantonese</OptionInput>
-                <OptionInput value="all_ko">All Korean</OptionInput>
+                <OptionInput value="all_ko">Korean</OptionInput>
             </SelectInput>
 
+            {#if !currentChar.data.gptSoVitsConfig.use_long_audio}
                 <span class="text-textcolor">Use Reference Audio Script</span>
                 <Check bind:check={currentChar.data.gptSoVitsConfig.use_prompt}/>
+            {/if}
 
-            {#if currentChar.data.gptSoVitsConfig.use_prompt}
+            {#if currentChar.data.gptSoVitsConfig.use_prompt && !currentChar.data.gptSoVitsConfig.use_long_audio}
                 <span class="text-textcolor">Reference Audio Script</span>
                 <TextAreaInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.prompt}/>
             {/if}
 
             <span class="text-textcolor">Reference Audio Language</span>
             <SelectInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.prompt_lang}>
-                <OptionInput value="auto">Auto</OptionInput>
+                <OptionInput value="auto">Multi-language Mixed</OptionInput>
-                <OptionInput value="auto_yue">Auto (Cantonese)</OptionInput>
+                <OptionInput value="auto_yue">Multi-language Mixed (Cantonese)</OptionInput>
                 <OptionInput value="en">English</OptionInput>
-                <OptionInput value="zh">Chinese</OptionInput>
+                <OptionInput value="zh">Chinese-English Mixed</OptionInput>
-                <OptionInput value="ja">Japanese</OptionInput>
+                <OptionInput value="ja">Japanese-English Mixed</OptionInput>
-                <OptionInput value="yue">Cantonese</OptionInput>
+                <OptionInput value="yue">Cantonese-English Mixed</OptionInput>
-                <OptionInput value="ko">Korean</OptionInput>
+                <OptionInput value="ko">Korean-English Mixed</OptionInput>
-                <OptionInput value="all_zh">English And Chinese</OptionInput>
+                <OptionInput value="all_zh">Chinese</OptionInput>
-                <OptionInput value="all_ja">English And Japanese</OptionInput>
+                <OptionInput value="all_ja">Japanese</OptionInput>
-                <OptionInput value="all_yue">English And Cantonese</OptionInput>
+                <OptionInput value="all_yue">Cantonese</OptionInput>
-                <OptionInput value="all_ko">English And Korean</OptionInput>
+                <OptionInput value="all_ko">Korean</OptionInput>
             </SelectInput>
             <span class="text-textcolor">Top P</span>
             <SliderInput min={0.0} max={1.0} step={0.05} fixed={2} bind:value={currentChar.data.gptSoVitsConfig.top_p}/>

src/ts/process/index.ts

@@ -1181,8 +1181,10 @@ export async function sendChat(chatProcessIndex = -1,arg:{chatAdditonalTokens?:n
             db.characters[selectedChar].chats[selectedChat] = currentChat
             setDatabase(db)
         }
+        if(db.ttsAutoSpeech){
             await sayTTS(currentChar, result)
         }
+    }
     else{
         const msgs = (req.type === 'success') ? [['char',req.result]] as const 
                     : (req.type === 'multiline') ? req.result
@@ -1240,7 +1242,9 @@ export async function sendChat(chatProcessIndex = -1,arg:{chatAdditonalTokens?:n
                 mrerolls.push(result)
             }
             db.characters[selectedChar].reloadKeys += 1
+            if(db.ttsAutoSpeech){
                 await sayTTS(currentChar, result)
+            }
             setDatabase(db)
         }
 

src/ts/process/tts.ts

@@ -240,7 +240,7 @@ export async function sayTTS(character:character,text:string) {
                 const body = {
                     text: text,
                     text_lang: character.gptSoVitsConfig.text_lang,
-                    ref_audio_path: character.gptSoVitsConfig.ref_audio_path + '/public/audio/' + character.gptSoVitsConfig.ref_audio_data.fileName,
+                    ref_audio_path: undefined,
                     ref_audio_name: character.gptSoVitsConfig.ref_audio_data.fileName,
                     ref_audio_data: base64Audio,
                     prompt_text: undefined,
@@ -250,18 +250,41 @@ export async function sayTTS(character:character,text:string) {
                     speed_factor: character.gptSoVitsConfig.speed,
                     top_k: character.gptSoVitsConfig.top_k,
                     text_split_method: character.gptSoVitsConfig.text_split_method,
-                    parallel_infer: false,
+                    parallel_infer: true,
+                    // media_type: character.gptSoVitsConfig.ref_audio_data.fileName.split('.')[1],
+                    ref_free: character.gptSoVitsConfig.use_long_audio || !character.gptSoVitsConfig.use_prompt,
                 }
 
                 if (character.gptSoVitsConfig.use_prompt){
                     body.prompt_text = character.gptSoVitsConfig.prompt
                 }
+
+                if (character.gptSoVitsConfig.use_auto_path){
+                    console.log('auto')
+                    const path = await globalFetch(`${character.gptSoVitsConfig.url}/get_path`, {
+                        method: 'GET',
+                        headers: {
+                            'Content-Type': 'application/json'
+                        },
+                        rawResponse: false,
+
+                    })
+                    console.log(path)
+                    if(path.ok){
+                        body.ref_audio_path = path.data.message + '/public/audio/' + character.gptSoVitsConfig.ref_audio_data.fileName
+                    }
+                    else{
+                        throw new Error('Failed to Auto get path')
+                    }
+                } else {
+                    body.ref_audio_path = character.gptSoVitsConfig.ref_audio_path + '/public/audio/' + character.gptSoVitsConfig.ref_audio_data.fileName
+                }
                 console.log(body)
 
                 const response = await globalFetch(`${character.gptSoVitsConfig.url}/tts`, {
                     method: 'POST',
                     headers: {
-                        "Content-Type": "application/json",
+                        'Content-Type': 'application/json'
                     },
                     body: body,
                     rawResponse: true,

src/ts/storage/database.ts

@@ -430,7 +430,8 @@ export function setDatabase(data:Database){
     }
     data.hideApiKey ??= true
     data.unformatQuotes ??= false
-
+    data.ttsAutoSpeech ??= false
+    data.translatorInputLanguage ??= 'auto'
     changeLanguage(data.language)
     DataBase.set(data)
 }
@@ -510,6 +511,7 @@ export interface Database{
     NAII2I:boolean
     NAIREF:boolean
     NAIImgConfig:NAIImgConfig
+    ttsAutoSpeech?:boolean
     runpodKey:string
     promptPreprocess:boolean
     bias: [string, number][]
@@ -614,6 +616,7 @@ export interface Database{
     emotionProcesser:'submodel'|'embedding',
     showMenuChatList?:boolean,
     translatorType:'google'|'deepl'|'none'|'llm'|'deeplX',
+    translatorInputLanguage?:string
     NAIadventure?:boolean,
     NAIappendName?:boolean,
     deeplOptions:{
@@ -809,7 +812,9 @@ export interface character{
     }
     gptSoVitsConfig?:{
         url?:string
+        use_auto_path?:boolean
         ref_audio_path?:string
+        use_long_audio?:boolean
         ref_audio_data?: {
             fileName:string
             assetId:string

src/ts/translator/translator.ts

@@ -165,7 +165,7 @@ async function translateMain(text:string, arg:{from:string, to:string, host:stri
     }
 
 
-    const url = `https://${arg.host}/translate_a/single?client=gtx&dt=t&sl=auto&tl=${arg.to}&q=` + encodeURIComponent(text)
+    const url = `https://${arg.host}/translate_a/single?client=gtx&dt=t&sl=${db.translatorInputLanguage}&tl=${arg.to}&q=` + encodeURIComponent(text)