feat: Piper Hindi TTS + backend translate/tts endpoints, drop direct LibreTranslate from frontend

app.js

@@ -95,7 +95,7 @@ window.speakText = function(text) {
         .replace(/🎯|💻|🔍|⚠️|🚀|💡/g, "")
         .replace(/\[([^\]]+)\]\([^)]+\)/g, "$1")
         .trim();
-    // Translate sentence by sentence to reduce lag
+    // Non-English: translate via backend then speak via Piper TTS
     if (window.ttsLang !== 'en-US' && window.ttsLang !== 'en-GB') {
         const langCode = window.ttsLang.split('-')[0];
         const sentences = clean.match(/[^.!?]+[.!?]+/g) || [clean];
@@ -103,24 +103,26 @@ window.speakText = function(text) {
         function translateAndSpeakNext() {
             if (i >= sentences.length || !window.ttsEnabled) return;
             const sentence = sentences[i].trim();
-            fetch('http://192.168.217.160:5000/translate', {
+            fetch('/translate', {
                 method: 'POST',
                 headers: {'Content-Type': 'application/json'},
-                body: JSON.stringify({q: sentence, source: 'en', target: langCode})
+                body: JSON.stringify({text: sentence, target: langCode})
             })
                 .then(r => r.json())
                 .then(data => {
                     const translated = data.translatedText || sentence;
-                    const utt = new SpeechSynthesisUtterance(translated);
-                    utt.lang = window.ttsLang;
-                    const voices = window.speechSynthesis.getVoices();
-                    const match = voices.find(v => v.lang === window.ttsLang);
-                    if (match) utt.voice = match;
-                    utt.rate = 0.95;
-                    utt.pitch = 1.0;
-                    utt.volume = 1.0;
-                    utt.onend = () => { i++; translateAndSpeakNext(); };
-                    window.speechSynthesis.speak(utt);
+                    return fetch('/tts', {
+                        method: 'POST',
+                        headers: {'Content-Type': 'application/json'},
+                        body: JSON.stringify({text: translated, lang: langCode})
+                    });
+                })
+                .then(r => r.blob())
+                .then(blob => {
+                    const url = URL.createObjectURL(blob);
+                    const audio = new Audio(url);
+                    audio.onended = () => { URL.revokeObjectURL(url); i++; translateAndSpeakNext(); };
+                    audio.play();
                 })
                 .catch(() => {
                     const utt = new SpeechSynthesisUtterance(sentence);

deepshell-backend/deepshell/__main__.py

@@ -371,6 +371,88 @@ async def run_agent_stream(req: ChatRequest, request: Request):
         }
     )
 
+# ============================================
+# Translation + TTS Endpoints
+# ============================================
+
+import subprocess
+import httpx
+
+class TranslateRequest(BaseModel):
+    text: str
+    target: str = "hi"
+
+class TTSRequest(BaseModel):
+    text: str
+    lang: str = "hi"
+
+LIBRETRANSLATE_URL = os.getenv("LIBRETRANSLATE_URL", "http://localhost:5000/translate")
+PIPER_BINARY = os.getenv("PIPER_BINARY", str(Path.home() / "piper/piper"))
+PIPER_VOICE_DIR = os.getenv("PIPER_VOICE_DIR", str(Path.home() / "piper/voices"))
+
+VOICE_MAP = {
+    "hi": "hi_IN-rohan-medium.onnx",
+    "ta": "ta_IN-cmu_indic-medium.onnx",
+    "te": "te_IN-cmu_indic-medium.onnx",
+    "ar": "ar_JO-kareem-medium.onnx",
+}
+
+@app.post("/translate")
+async def translate_text(req: TranslateRequest):
+    """Translate text via LibreTranslate."""
+    try:
+        async with httpx.AsyncClient(timeout=10.0) as client:
+            resp = await client.post(
+                LIBRETRANSLATE_URL,
+                json={"q": req.text, "source": "en", "target": req.target}
+            )
+            data = resp.json()
+            return {"translatedText": data.get("translatedText", req.text)}
+    except Exception as e:
+        return {"translatedText": req.text, "error": str(e)}
+
+@app.post("/tts")
+async def text_to_speech(req: TTSRequest):
+    """Convert text to speech via Piper, return WAV audio."""
+    voice_file = VOICE_MAP.get(req.lang)
+    if not voice_file:
+        return JSONResponse({"error": f"No voice for lang: {req.lang}"}, status_code=400)
+
+    model_path = os.path.join(PIPER_VOICE_DIR, voice_file)
+    if not os.path.exists(model_path):
+        return JSONResponse({"error": f"Voice model not found: {model_path}"}, status_code=404)
+
+    if not os.path.exists(PIPER_BINARY):
+        return JSONResponse({"error": f"Piper binary not found: {PIPER_BINARY}"}, status_code=404)
+
+    try:
+        import tempfile
+        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp:
+            tmp_path = tmp.name
+
+        proc = await asyncio.create_subprocess_exec(
+            PIPER_BINARY,
+            "--model", model_path,
+            "--output_file", tmp_path,
+            stdin=asyncio.subprocess.PIPE,
+            stdout=asyncio.subprocess.DEVNULL,
+            stderr=asyncio.subprocess.DEVNULL,
+            env={**os.environ, "LD_LIBRARY_PATH": str(Path(PIPER_BINARY).parent)}
+        )
+        await asyncio.wait_for(proc.communicate(input=req.text.encode()), timeout=30)
+        with open(tmp_path, "rb") as f:
+            wav_data = f.read()
+        os.unlink(tmp_path)
+        return StreamingResponse(
+            iter([wav_data]),
+            media_type="audio/wav",
+            headers={"Content-Disposition": "inline; filename=tts.wav"}
+        )
+    except asyncio.TimeoutError:
+        return JSONResponse({"error": "TTS generation timed out"}, status_code=504)
+    except Exception as e:
+        return JSONResponse({"error": str(e)}, status_code=500)
+
 # ============================================
 # COMMAND EXECUTION - DISABLED FOR FREE VERSION
 # ============================================