Update app.py

app.py

@@ -104,15 +104,19 @@ def generate_answer(query: str) -> str:
         f"Question: {query}\n\n"
         f"Answer:"
     )
-    # Use max_new_tokens to generate additional tokens beyond the prompt.
     response = generator(prompt, max_new_tokens=100, do_sample=True, temperature=0.7)
-    return response[0]["generated_text"]
+    generated_text = response[0]["generated_text"]
+    # Return only the text after the "Answer:" delimiter
+    if "Answer:" in generated_text:
+        answer = generated_text.split("Answer:", 1)[1].strip()
+    else:
+        answer = generated_text.strip()
+    return answer
 
 # ====================================================
 # 2. Speech-to-Text and Text-to-Speech Functions
 # ====================================================
 
-# Force Whisper to load on CPU explicitly
 stt_model = whisper.load_model("base", device="cpu")
 
 def speech_to_text(audio_array: np.ndarray, sample_rate: int = 16000) -> str:
@@ -193,7 +197,6 @@ class RAGVoiceHandler(AsyncStreamHandler):
 # 4. Voice Streaming Setup & FastAPI Endpoints
 # ====================================================
 
-# Supply a dummy (but valid) RTC configuration with a TURN server entry
 rtc_config = {
     "iceServers": [
         {"urls": "stun:stun.l.google.com:19302"},
@@ -229,7 +232,6 @@ async def input_hook(body: InputData):
 async def webrtc_offer(offer: dict):
     return await stream.handle_offer(offer)
 
-# Added /chat endpoint for text-based queries (fallback)
 @app.post("/chat")
 async def chat_endpoint(payload: dict):
     question = payload.get("question", "")