uwonbot 1.1.6 → 1.1.8

package/bin/uwonbot.js

@@ -14,7 +14,7 @@ showBanner();
 program
     .name('uwonbot')
     .description('Uwonbot AI Assistant — Your AI controls your computer')
-    .version('1.1.6');
+    .version('1.1.7');
 
 program
     .command('login')

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "uwonbot",
-  "version": "1.1.6",
+  "version": "1.1.8",
   "description": "Uwonbot AI Assistant CLI — Your AI controls your computer",
   "main": "src/index.js",
   "bin": {

package/src/agent.js

@@ -323,36 +323,26 @@ async function handleCommand(msg) {
 }
 
 async function openWebAssistant(assistantId) {
-    const url = `${WEB_APP_URL}/assistant/live?id=${assistantId}`;
+    const chatUrl = `${WEB_APP_URL}/assistant/live?id=${assistantId}`;
     try {
         if (platform === 'darwin') {
-            await execAsync(`open -na "Google Chrome" --args --new-window "${url}" 2>/dev/null || open -na "Safari" --args "${url}" 2>/dev/null || open "${url}"`);
+            await execAsync(`open -na "Google Chrome" --args --new-window --window-size=380,520 --window-position=1000,200 "${chatUrl}" 2>/dev/null || open "${chatUrl}"`);
         } else if (platform === 'win32') {
-            await execAsync(`start chrome --new-window "${url}" 2>nul || start "" "${url}"`);
+            await execAsync(`start chrome --new-window --window-size=380,520 --window-position=1000,200 "${chatUrl}" 2>nul || start "" "${chatUrl}"`);
         } else {
-            await execAsync(`google-chrome --new-window "${url}" 2>/dev/null || xdg-open "${url}"`);
+            await execAsync(`google-chrome --new-window "${chatUrl}" 2>/dev/null || xdg-open "${chatUrl}"`);
         }
         return true;
     } catch {
-        try { await execAsync(`open "${url}"`); return true; } catch { return false; }
+        try { await execAsync(`open "${chatUrl}"`); return true; } catch { return false; }
     }
 }
 
 async function activateAllAssistants(assistants) {
     const opened = [];
     for (const a of assistants) {
-        const mode = a.activationMode || 'web';
-        if (mode === 'terminal') {
-            await openTerminalWithChat(a.name);
-            opened.push({ name: a.name, mode: 'terminal' });
-        } else if (mode === 'web') {
-            await openWebAssistant(a.id);
-            opened.push({ name: a.name, mode: 'web' });
-        } else {
-            await openTerminalWithChat(a.name);
-            await openWebAssistant(a.id);
-            opened.push({ name: a.name, mode: 'both' });
-        }
+        await openWebAssistant(a.id);
+        opened.push({ name: a.name, mode: 'web' });
         await new Promise(r => setTimeout(r, 500));
     }
     return opened;
@@ -449,17 +439,23 @@ export async function startAgent(port = 9876, options = {}) {
             console.log(chalk.bold.cyan('  👏 박수 감지! 비서 활성화 중...'));
 
             if (userAssistants.length === 0) {
-                console.log(chalk.gray('  → 기본 Uwonbot 실행'));
-                openTerminalWithChat('Uwonbot');
+                console.log(chalk.gray('  → 기본 Uwonbot 웹 실행'));
+                try {
+                    if (platform === 'darwin') {
+                        await execAsync(`open "${WEB_APP_URL}/assistant"`);
+                    } else if (platform === 'win32') {
+                        await execAsync(`start "" "${WEB_APP_URL}/assistant"`);
+                    } else {
+                        await execAsync(`xdg-open "${WEB_APP_URL}/assistant"`);
+                    }
+                } catch {}
                 return;
             }
 
             if (userAssistants.length === 1) {
                 const a = userAssistants[0];
-                const mode = a.activationMode || 'web';
-                console.log(chalk.green(`  → ${a.name} 실행 (${mode})`));
-                if (mode === 'terminal' || mode === 'both') await openTerminalWithChat(a.name);
-                if (mode === 'web' || mode === 'both') await openWebAssistant(a.id);
+                console.log(chalk.green(`  → ${a.name} 웹 실행`));
+                await openWebAssistant(a.id);
                 return;
             }
 
@@ -488,7 +484,11 @@ export async function startAgent(port = 9876, options = {}) {
         console.log(chalk.green(`  ✓ Client connected from ${origin}`));
 
         ws.on('message', async (data) => {
-            const result = await handleCommand(data.toString());
+            const raw = data.toString();
+            let _reqId;
+            try { _reqId = JSON.parse(raw)._reqId; } catch {}
+            const result = await handleCommand(raw);
+            if (_reqId) result._reqId = _reqId;
             ws.send(JSON.stringify(result));
         });
 
@@ -508,10 +508,14 @@ export async function startAgent(port = 9876, options = {}) {
                 const retry = new WebSocketServer({ port });
                 retry.on('connection', (ws, req) => {
                     ws.on('message', async (data) => {
-                        const result = await handleCommand(data.toString());
+                        const raw = data.toString();
+                        let _reqId;
+                        try { _reqId = JSON.parse(raw)._reqId; } catch {}
+                        const result = await handleCommand(raw);
+                        if (_reqId) result._reqId = _reqId;
                         ws.send(JSON.stringify(result));
                     });
-                    ws.send(JSON.stringify({ type: 'welcome', agent: 'uwonbot', version: '1.1.4', uid }));
+                    ws.send(JSON.stringify({ type: 'welcome', agent: 'uwonbot', version: '1.1.8', uid }));
                 });
                 console.log(chalk.green(`  ✓ 재시도 성공 — ws://localhost:${port}`));
             } catch {

package/src/chat.js

@@ -7,7 +7,8 @@ import open from 'open';
 import { getConfig } from './config.js';
 import { sendToBrain } from './brain.js';
 import { showMiniBar } from './banner.js';
-import { printOrb, animateOrb } from './terminalOrb.js';
+// terminalOrb is used by agent.js for display; chat.js shows text-only
+import { speak } from './terminalTTS.js';
 import VoiceInput from './voiceInput.js';
 import {
     fetchAssistants,
@@ -177,19 +178,19 @@ export async function startChat(assistantName, assistant, initialCommand, option
 
     const colorsArr = assistant.colors;
     const orbColorHex = (Array.isArray(colorsArr) && colorsArr[0]) || assistant.orbColor || brainColor;
-    const orbRgb = hexToRgb(orbColorHex);
 
     console.clear();
-    printOrb({
-        radius: 10,
-        color: orbRgb,
-        label: assistant.name.toUpperCase(),
-        status: '준비됨',
-    });
+
+    const c = chalk.hex(brainColor);
+    console.log('');
+    console.log(c('  ●') + chalk.bold.white(` ${assistant.name}`));
+    console.log(chalk.gray(`  ${brainLabel} | ${assistant.voiceLang || 'ko-KR'}`));
+    console.log('');
 
     await bootSequence(assistant, brainLabel, brainColor);
 
-    const voiceMode = options.voice || false;
+    const hasApiKey = !!(assistant.apiKey || process.env.GEMINI_API_KEY);
+    const voiceMode = options.voice !== undefined ? options.voice : hasApiKey;
 
     console.log('');
     console.log(chalk.gray('  ─────────────────────────────────────────'));
@@ -221,22 +222,27 @@ export async function startChat(assistantName, assistant, initialCommand, option
             const ok = await voiceInput.start({
                 onListening: () => {
                     if (!processingVoice) {
-                        process.stdout.write(chalk.gray('  🎙 듣고 있습니다...') + '\r');
+                        process.stdout.write('\x1b[2K\r' + chalk.gray('  🎙 듣고 있습니다... (말하세요)'));
                     }
                 },
                 onSpeechStart: () => {
                     process.stdout.write('\x1b[2K\r');
-                    process.stdout.write(chalk.cyan('  🔴 음성 감지 중...') + '\r');
                 },
                 onSpeechEnd: () => {
-                    process.stdout.write('\x1b[2K\r');
-                    process.stdout.write(chalk.gray('  ⏳ 음성 인식 중...') + '\r');
+                    process.stdout.write('\x1b[2K\r' + chalk.yellow('  ⏳ 음성 인식 중...'));
+                },
+                onAmplitude: (amp, isSpeaking) => {
+                    if (processingVoice) return;
+                    if (isSpeaking) {
+                        const bar = renderVoiceBar(amp);
+                        process.stdout.write('\x1b[2K\r' + chalk.cyan('  🔴 ') + chalk.hex(brainColor)(bar) + chalk.gray(' 말하는 중...'));
+                    }
                 },
                 onTranscript: async (text) => {
                     if (processingVoice) return;
                     processingVoice = true;
                     process.stdout.write('\x1b[2K\r');
-                    console.log(chalk.hex(brainColor)(`  You (voice) > `) + chalk.white(text));
+                    console.log(chalk.hex(brainColor)(`  🗣 You > `) + chalk.white(text));
                     rl.pause();
                     await processMessage(text, messages, assistant, brainColor);
                     processingVoice = false;
@@ -285,13 +291,18 @@ export async function startChat(assistantName, assistant, initialCommand, option
                     voiceInput = new VoiceInput(apiKey);
                     await voiceInput.start({
                         onListening: () => {},
-                        onSpeechStart: () => process.stdout.write(chalk.cyan('\r  🔴 음성 감지 중...') + '\r'),
-                        onSpeechEnd: () => process.stdout.write(chalk.gray('\r  ⏳ 인식 중...') + '\r'),
+                        onSpeechStart: () => {},
+                        onSpeechEnd: () => process.stdout.write('\x1b[2K\r' + chalk.yellow('  ⏳ 인식 중...')),
+                        onAmplitude: (amp, isSpeaking) => {
+                            if (processingVoice || !isSpeaking) return;
+                            const bar = renderVoiceBar(amp);
+                            process.stdout.write('\x1b[2K\r' + chalk.cyan('  🔴 ') + chalk.hex(brainColor)(bar));
+                        },
                         onTranscript: async (text) => {
                             if (processingVoice) return;
                             processingVoice = true;
                             process.stdout.write('\x1b[2K\r');
-                            console.log(chalk.hex(brainColor)(`  You (voice) > `) + chalk.white(text));
+                            console.log(chalk.hex(brainColor)(`  🗣 You > `) + chalk.white(text));
                             rl.pause();
                             await processMessage(text, messages, assistant, brainColor);
                             processingVoice = false;
@@ -330,6 +341,13 @@ export async function startChat(assistantName, assistant, initialCommand, option
     });
 }
 
+function renderVoiceBar(amplitude) {
+    const barLen = Math.round(amplitude * 40);
+    const bar = '█'.repeat(Math.min(barLen, 40));
+    const empty = '░'.repeat(40 - Math.min(barLen, 40));
+    return bar + empty;
+}
+
 async function processMessage(input, messages, assistant, brainColor) {
     const spinner = ora({
         text: chalk.gray('Thinking...'),
@@ -337,6 +355,9 @@ async function processMessage(input, messages, assistant, brainColor) {
         color: 'blue',
     }).start();
 
+    const vLang = assistant.voiceLang || 'ko-KR';
+    const vGender = assistant.voiceGender || assistant.voiceStyle || 'male';
+
     try {
         const reply = await sendToBrain(assistant, messages, input);
         spinner.stop();
@@ -348,12 +369,20 @@ async function processMessage(input, messages, assistant, brainColor) {
             console.log(chalk.white(`  ${line}`));
         }
         console.log('');
+
+        speak(reply, { lang: vLang, gender: vGender }).catch(() => {});
     } catch (err) {
         spinner.stop();
-        console.log(chalk.red(`\n  ❌ Error: ${err.message}\n`));
-        if (err.message.includes('API key') || err.message.includes('api key')) {
-            console.log(chalk.yellow('  Make sure your assistant has a valid API key configured.'));
-            console.log(chalk.gray('  Update it at: https://chartapp-653e1.web.app/assistant/create\n'));
+        const rawMsg = err.message || '';
+        let friendlyMsg = rawMsg;
+        if (rawMsg.includes('API_KEY_INVALID') || rawMsg.includes('API key not valid')) {
+            friendlyMsg = 'API 키가 유효하지 않습니다. 비서 설정에서 올바른 API 키를 확인해주세요.';
+        } else if (rawMsg.includes('No API key')) {
+            friendlyMsg = 'API 키가 설정되지 않았습니다. 웹에서 비서의 API 키를 설정해주세요.';
+        }
+        console.log(chalk.red(`\n  ❌ ${friendlyMsg}\n`));
+        if (rawMsg.includes('API key') || rawMsg.includes('api key') || rawMsg.includes('No API key')) {
+            console.log(chalk.gray('  API 키 설정: https://chartapp-653e1.web.app/assistant/create\n'));
         }
     }
 }

package/src/terminalTTS.js

@@ -0,0 +1,99 @@
+import { exec } from 'child_process';
+import { promisify } from 'util';
+import { writeFileSync, unlinkSync, existsSync } from 'fs';
+import { join } from 'path';
+import { tmpdir } from 'os';
+import fetch from 'node-fetch';
+
+const execAsync = promisify(exec);
+const platform = process.platform;
+
+const GOOGLE_TTS_FUNCTION = 'https://us-central1-chartapp-653e1.cloudfunctions.net/googleTTS';
+
+async function speakWithGoogleTTS(text, lang = 'ko-KR', gender = 'male') {
+    try {
+        const res = await fetch(GOOGLE_TTS_FUNCTION, {
+            method: 'POST',
+            headers: { 'Content-Type': 'application/json' },
+            body: JSON.stringify({ text, lang, gender }),
+        });
+        if (!res.ok) throw new Error(`TTS API ${res.status}`);
+        const data = await res.json();
+        if (!data.audio) throw new Error('No audio');
+
+        const tmpFile = join(tmpdir(), `uwonbot_tts_${Date.now()}.mp3`);
+        writeFileSync(tmpFile, Buffer.from(data.audio, 'base64'));
+
+        if (platform === 'darwin') {
+            await execAsync(`afplay "${tmpFile}"`);
+        } else if (platform === 'win32') {
+            await execAsync(`powershell -c "(New-Object Media.SoundPlayer '${tmpFile}').PlaySync()"`);
+        } else {
+            try {
+                await execAsync(`mpg123 "${tmpFile}" 2>/dev/null || ffplay -nodisp -autoexit "${tmpFile}" 2>/dev/null || aplay "${tmpFile}" 2>/dev/null`);
+            } catch {}
+        }
+
+        try { unlinkSync(tmpFile); } catch {}
+        return true;
+    } catch {
+        return false;
+    }
+}
+
+async function speakWithSay(text, lang = 'ko-KR', gender = 'male') {
+    if (platform !== 'darwin') return false;
+
+    let voice;
+    if (lang.startsWith('ko')) {
+        voice = gender === 'female' ? 'Yuna' : 'Jian';
+    } else {
+        voice = gender === 'female' ? 'Samantha' : 'Daniel';
+    }
+
+    try {
+        await execAsync(`say -v "${voice}" "${text.replace(/"/g, '\\"')}"`);
+        return true;
+    } catch {
+        try {
+            await execAsync(`say "${text.replace(/"/g, '\\"')}"`);
+            return true;
+        } catch { return false; }
+    }
+}
+
+async function speakWithEspeak(text, lang = 'ko-KR') {
+    if (platform === 'darwin') return false;
+    const espeakLang = lang.startsWith('ko') ? 'ko' : 'en';
+    try {
+        await execAsync(`espeak -v ${espeakLang} "${text.replace(/"/g, '\\"')}" 2>/dev/null`);
+        return true;
+    } catch { return false; }
+}
+
+export async function speak(text, options = {}) {
+    const lang = options.lang || 'ko-KR';
+    const gender = options.gender || 'male';
+
+    if (!text || !text.trim()) return;
+
+    const ok = await speakWithGoogleTTS(text, lang, gender);
+    if (ok) return;
+
+    if (platform === 'darwin') {
+        await speakWithSay(text, lang, gender);
+        return;
+    }
+
+    await speakWithEspeak(text, lang);
+}
+
+export async function speakStreaming(text, options = {}) {
+    const sentences = text.match(/[^.!?。！？\n]+[.!?。！？\n]?/g) || [text];
+    for (const s of sentences) {
+        const t = s.trim();
+        if (t) await speak(t, options);
+    }
+}
+
+export default { speak, speakStreaming };

package/src/voiceInput.js

@@ -78,11 +78,12 @@ export default class VoiceInput {
         this.onSpeechEnd = null;
     }
 
-    async start({ onTranscript, onListening, onSpeechStart, onSpeechEnd }) {
+    async start({ onTranscript, onListening, onSpeechStart, onSpeechEnd, onAmplitude }) {
         this.onTranscript = onTranscript;
         this.onListening = onListening;
         this.onSpeechStart = onSpeechStart;
         this.onSpeechEnd = onSpeechEnd;
+        this.onAmplitude = onAmplitude;
 
         let micModule;
         try {
@@ -120,10 +121,14 @@ export default class VoiceInput {
 
         this.onListening?.();
 
+        let speechDuration = 0;
+
         stream.on('data', (buf) => {
             if (!this.running) return;
             const amp = getAmplitude(buf);
 
+            this.onAmplitude?.(amp, isSpeaking);
+
             if (amp > SILENCE_THRESHOLD) {
                 if (!isSpeaking) {
                     isSpeaking = true;
@@ -133,6 +138,7 @@ export default class VoiceInput {
                 }
                 silenceStart = null;
                 speechChunks.push(Buffer.from(buf));
+                speechDuration = Date.now() - speechStart;
             } else if (isSpeaking) {
                 speechChunks.push(Buffer.from(buf));
                 if (!silenceStart) silenceStart = Date.now();
@@ -141,6 +147,7 @@ export default class VoiceInput {
                     const duration = Date.now() - speechStart;
                     isSpeaking = false;
                     silenceStart = null;
+                    speechDuration = 0;
                     this.onSpeechEnd?.();
 
                     if (duration >= MIN_SPEECH_MS && speechChunks.length > 0) {