Update app.py

app.py

@@ -18,7 +18,6 @@ def load_spacy_model():
     try:
         return spacy.load("en_core_web_sm")
     except OSError:
-        st.warning("Downloading spaCy model en_core_web_sm... This may take a moment.")
         subprocess.run(["python", "-m", "spacy", "download", "en_core_web_sm"], check=True)
         return spacy.load("en_core_web_sm")
 
@@ -110,9 +109,9 @@ def generate_follow_up(user_text):
     Generates two variant follow-up questions and randomly selects one.
     """
     prompt_text = (
-        f"Based on the user's question: '{user_text}', generate two concise, friendly follow-up questions that invite further discussion. "
-        "For example, one might be 'Would you like to know more about the six types of quarks?' and another 'Would you like to explore another aspect of quantum physics?'. "
-        "Answer exclusively in English, and do not include extra commentary."
+        f"Based on the user's question: '{user_text}', generate two concise, friendly follow-up questions "
+        "that invite further discussion. For example, one might be 'Would you like to know more about the six types of quarks?' "
+        "and another 'Would you like to explore another aspect of quantum physics?'. Do not include extra commentary."
     )
     hf = get_llm_hf_inference(max_new_tokens=80, temperature=0.9)
     output = hf.invoke(input=prompt_text).strip()
@@ -124,17 +123,15 @@ def generate_follow_up(user_text):
 
 def get_response(system_message, chat_history, user_text, max_new_tokens=1024):
     """
-    Generates HAL's detailed, in-depth response and a follow-up question.
-    It incorporates sentiment analysis, additional NLP context, and style instructions.
+    Generates HAL's detailed, in-depth answer and a follow-up question.
+    Incorporates sentiment analysis, additional NLP context, and style instructions.
     """
     sentiment = analyze_sentiment(user_text)
     action = predict_action(user_text)
     
-    # Extract extra context (e.g., named entities)
     context_info = extract_context(user_text)
     context_clause = f" The key topics here are: {context_info}." if context_info else ""
     
-    # Extract style instruction if provided.
     style_instruction = ""
     lower_text = user_text.lower()
     if "in the voice of" in lower_text or "speaking as" in lower_text:
@@ -143,7 +140,6 @@ def get_response(system_message, chat_history, user_text, max_new_tokens=1024):
             style_instruction = match.group(2).strip().capitalize()
             style_instruction = f" Please respond in the voice of {style_instruction}."
     
-    # Force output in English.
     language_clause = " Answer exclusively in English."
     
     if action == "nasa_info":
@@ -174,15 +170,9 @@ def get_response(system_message, chat_history, user_text, max_new_tokens=1024):
         )
     )
     
-    st.write("DEBUG: Prompt sent to model:")
-    st.write(prompt.format(system_message=system_message, chat_history=filtered_history, user_text=user_text))
-    
     chat = prompt | hf.bind(skip_prompt=True) | StrOutputParser(output_key='content')
     raw_output = chat.invoke(input=dict(system_message=system_message, user_text=user_text, chat_history=filtered_history))
-    st.write("DEBUG: Raw model output:")
-    st.write(raw_output)
-    
-    response = raw_output  # Using the full raw output without splitting
+    response = raw_output.split("HAL:")[-1].strip()
     if not response:
         response = "Certainly, here is an in-depth explanation: [Fallback explanation]."
     
@@ -196,8 +186,6 @@ def get_response(system_message, chat_history, user_text, max_new_tokens=1024):
     follow_up = generate_follow_up(user_text)
     chat_history.append({'role': 'assistant', 'content': follow_up})
     
-    st.write("DEBUG: Generated follow-up question:", follow_up)
-    
     return response, follow_up, chat_history, None
 
 # --- Chat UI ---