Update app.py

app.py

@@ -1,3 +1,5 @@
+# hal_bot.py
+
 import os
 import re
 import requests
@@ -7,11 +9,11 @@ from langchain_huggingface import HuggingFaceEndpoint
 from langchain_core.prompts import PromptTemplate
 from langchain_core.output_parsers import StrOutputParser
 from transformers import pipeline
-from langdetect import detect  # Ensure this package is installed
+from langdetect import detect
 
-# ✅ Check for GPU or Default to CPU
+# ✅ Device setup
 device = "cuda" if torch.cuda.is_available() else "cpu"
-print(f"✅ Using device: {device}")  # Debugging info
+print(f"✅ Using device: {device}")
 
 # ✅ Environment Variables
 HF_TOKEN = os.getenv("HF_TOKEN")
@@ -22,25 +24,25 @@ NASA_API_KEY = os.getenv("NASA_API_KEY")
 if NASA_API_KEY is None:
     raise ValueError("NASA_API_KEY is not set. Please add it to your environment variables.")
 
-# ✅ Set Up Streamlit
+# ✅ Streamlit Setup
 st.set_page_config(page_title="HAL - NASA ChatBot", page_icon="🚀")
 
-# ✅ Initialize Session State Variables (Ensuring Chat History Persists)
 if "chat_history" not in st.session_state:
     st.session_state.chat_history = [{"role": "assistant", "content": "Hello! How can I assist you today?"}]
 
-# ✅ Initialize Hugging Face Model (Explicitly Set to CPU/GPU)
-def get_llm_hf_inference(model_id="meta-llama/Llama-2-7b-chat-hf", max_new_tokens=800, temperature=0.3):
+# ✅ Switched to Flan-T5 Model
+MODEL_ID = "google/flan-t5-large"
+
+def get_llm_hf_inference(model_id=MODEL_ID, max_new_tokens=500, temperature=0.3):
     return HuggingFaceEndpoint(
         repo_id=model_id,
         max_new_tokens=max_new_tokens,
-        temperature=temperature,  # 🔥 Lowered temperature for more factual and structured responses
+        temperature=temperature,
         token=HF_TOKEN,
-        task="text-generation",
-        device=-1 if device == "cpu" else 0  # ✅ Force CPU (-1) or GPU (0)
+        task="text2text-generation",
+        device=-1 if device == "cpu" else 0
     )
 
-# ✅ Ensure English Responses
 def ensure_english(text):
     try:
         detected_lang = detect(text)
@@ -50,46 +52,41 @@ def ensure_english(text):
         return "⚠️ Language detection failed. Please ask your question again."
     return text
 
-# ✅ Main Response Function (Fixing Repetition & Context)
-def get_response(system_message, chat_history, user_text, max_new_tokens=800):
-    # ✅ Ensure conversation history is included correctly
+def get_response(system_message, chat_history, user_text, max_new_tokens=500):
     filtered_history = "\n".join(
-        f"{msg['role'].capitalize()}: {msg['content']}"
-        for msg in chat_history[-5:]  # ✅ Only keep the last 5 exchanges to prevent overflow
+        f"{msg['role'].capitalize()}: {msg['content']}" for msg in chat_history[-5:]
     )
- 
+
     prompt = PromptTemplate.from_template(
-        "[INST] You are a highly knowledgeable AI assistant. Answer concisely, avoid repetition, and structure responses well."
-        "\n\nCONTEXT:\n{chat_history}\n"
-        "\nLATEST USER INPUT:\nUser: {user_text}\n"
-        "\n[END CONTEXT]\n"
-        "Assistant:"
+        """
+        You are a helpful NASA AI assistant.
+        Answer concisely and clearly based on the conversation history and the user's latest message.
+
+        Conversation History:
+        {chat_history}
+
+        User: {user_text}
+        Assistant:
+        """
     )
 
-    # ✅ Invoke Hugging Face Model
-    hf = get_llm_hf_inference(max_new_tokens=max_new_tokens, temperature=0.3)  # 🔥 Lowered temperature
+    hf = get_llm_hf_inference(max_new_tokens=max_new_tokens, temperature=0.3)
     chat = prompt | hf.bind(skip_prompt=True) | StrOutputParser(output_key='content')
 
     response = chat.invoke(input=dict(system_message=system_message, user_text=user_text, chat_history=filtered_history))
-    
-    # Clean up the response - remove any "HAL:" prefix if present
-    response = response.split("HAL:")[-1].strip() if "HAL:" in response else response.strip()
+    response = response.strip()
     response = ensure_english(response)
 
     if not response:
         response = "I'm sorry, but I couldn't generate a response. Can you rephrase your question?"
 
-    # ✅ Update conversation history
     chat_history.append({'role': 'user', 'content': user_text})
     chat_history.append({'role': 'assistant', 'content': response})
 
-    # ✅ Keep only last 10 exchanges to prevent unnecessary repetition
     return response, chat_history[-10:]
 
-# ✅ Streamlit UI
 st.title("🚀 HAL - NASA AI Assistant")
 
-# ✅ Justify all chatbot responses
 st.markdown("""
     <style>
     .user-msg, .assistant-msg {
@@ -107,22 +104,19 @@ st.markdown("""
     </style>
 """, unsafe_allow_html=True)
 
-# ✅ Chat UI
 user_input = st.chat_input("Type your message here...")
 
 if user_input:
-    # Get response and update chat history
     response, st.session_state.chat_history = get_response(
         system_message="You are a helpful AI assistant.",
         user_text=user_input,
         chat_history=st.session_state.chat_history
     )
 
-# ✅ Display chat history (ONLY display from history, not separately)
 st.markdown("<div class='container'>", unsafe_allow_html=True)
 for message in st.session_state.chat_history:
     if message["role"] == "user":
         st.markdown(f"<div class='user-msg'><strong>You:</strong> {message['content']}</div>", unsafe_allow_html=True)
     else:
         st.markdown(f"<div class='assistant-msg'><strong>HAL:</strong> {message['content']}</div>", unsafe_allow_html=True)
-st.markdown("</div>", unsafe_allow_html=True)
+st.markdown("</div>", unsafe_allow_html=True)