Update app_qwen_tts.py

app_qwen_tts.py

@@ -98,18 +98,46 @@ def answer_question(question):
 # =====================================================
 # TTS (BASE64 → WAV)
 # =====================================================
-def generate_audio(text):
-    payload = {"text": text, "language_id": "en", "mode": "Speak 🗣️"}
+def generate_audio(text: str):
+    payload = {
+        "text": text,
+        "language_id": "en",
+        "mode": "Speak 🗣️"
+    }
+
     r = requests.post(TTS_API_URL, json=payload, timeout=None)
-    audio_b64 = r.json()["audio"]
 
-    audio_bytes = base64.b64decode(audio_b64)
-    wav_path = "/tmp/output.wav"
+    # ---- Case 1: API returns raw WAV bytes ----
+    if r.headers.get("content-type", "").startswith("audio"):
+        wav_path = "/tmp/output.wav"
+        with open(wav_path, "wb") as f:
+            f.write(r.content)
+        return wav_path
+
+    # ---- Case 2: API returns JSON ----
+    data = r.json()
+
+    # Try known keys safely
+    audio_b64 = (
+        data.get("audio")
+        or data.get("audio_base64")
+        or data.get("wav")
+    )
+
+    if audio_b64:
+        wav_bytes = base64.b64decode(audio_b64)
+        wav_path = "/tmp/output.wav"
+        with open(wav_path, "wb") as f:
+            f.write(wav_bytes)
+        return wav_path
+
+    # ---- Case 3: API returns a file path ----
+    if "path" in data and os.path.exists(data["path"]):
+        return data["path"]
 
-    with open(wav_path, "wb") as f:
-        f.write(audio_bytes)
+    # ---- Otherwise ----
+    raise RuntimeError(f"TTS API response invalid: {data}")
 
-    return wav_path
 
 # =====================================================
 # MAIN HANDLER