Update app.py

app.py

@@ -1,326 +1,327 @@
-import os
-import logging
-import gradio as gr
-import re
-
-#from langchain_chroma import Chroma
-#from langchain_huggingface import HuggingFaceEmbeddings
-
-from langchain_community.vectorstores import Chroma
-from langchain_community.embeddings import HuggingFaceEmbeddings
-
-
-from groq import Groq
-
-from transformers import pipeline
-
-
-import requests  # for calling SerpAPI
-# Add your SerpAPI key here
-SERPAPI_API_KEY = "48125364c49a1952f1c8fdadf0d22e0d5bc3d195a5a98ad64d7e935aad503efa"
-
-# Load zero-shot classification model
-classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
-
-# Suppress noisy logs
-os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
-logging.getLogger('tensorflow').setLevel(logging.ERROR)
-logging.getLogger('torch').setLevel(logging.ERROR)
-
-# Config
-CHROMA_DIR = "chroma_country_info"
-GROQ_MODEL = "llama-3.3-70b-versatile"  # You can also try: mixtral-8x7b or gemma-7b-it
-
-
-
-
-# Load Groq client
-client = Groq(api_key="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")
-
-# Load vector DB
-print("🔍 Loading Chroma Vector DB...")
-embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
-vectordb = Chroma(persist_directory=CHROMA_DIR, embedding_function=embeddings)
-
-# Memory for user-provided factual information (in-session only)
-user_memory = []
-
-
-
-
-def is_question(text):
-    
-    QUESTION_STARTERS = [
-    "who", "what", "when", "where", "why", "which", "whose", "whom", "how",
-    "is", "are", "am", "was", "were",
-    "do", "does", "did",
-    "have", "has", "had",
-    "can", "could", "will", "would", "shall", "should",
-    "may", "might", "must",
-    "didn't", "isn't", "aren't", "wasn't", "weren't",
-    "won't", "shouldn't", "couldn't", "wouldn't"
-    ]
-
-    text_clean = text.strip().lower()
-    
-    if text_clean.endswith('?'):
-        return True
-    
-    # Rule 2: Starts with known question word
-    words = text_clean.split()
-    if words and words[0] in QUESTION_STARTERS:
-        return True
-    
-    
-    return False
-
-def is_fact(text):
-
-    if is_question(text):
-        return False
-    
-    candidate_labels = ["fact", "opinion", "speculation"]
-    result = classifier(text, candidate_labels)
-    
-    label = result['labels'][0]
-    score = result['scores'][0]
-
-    # Rule: High-confidence fact => keep as fact
-    if label == "fact" and score > 0.75:
-        final_label = "fact"
-
-    # Rule: If label is opinion or speculation, but sentence sounds assertive => custom fact
-    elif label in ["opinion", "speculation"] and score < 0.9:
-        final_label = "custom fact"
-
-    # Rule: Low-confidence fact => custom fact
-    elif label == "fact" and score <= 0.75:
-        final_label = "custom fact"
-
-    else:
-        final_label = "custom fact"
-
-    # check if there is "fact in final_label and return True
-    
-    is_fact = "fact" in final_label
-
-    return is_fact
-
-def clear_user_memory():
-    user_memory.clear()
-    return [], []  # Clear chat and state too (optional)
-
-
-def generate_answer(user_input):
-    print("📚 generate_answer:\n", user_input)
-
-    # Handle memory commands
-    if "what do you remember" in user_input.lower():
-        if not user_memory:
-            return "I don't remember anything yet."
-        return "Here's what I remember:\n" + "\n".join(f"- {fact}" for fact in user_memory)
-
-    if "forget everything" in user_input.lower():
-        user_memory.clear()
-        return "Okay, I’ve forgotten everything you told me."
-
-    if not user_input.strip():
-        return "Please enter a question."
-
-    # Store statements as memory
-    Isitquestion=True
-    if is_fact(user_input):
-        Isitquestion=False
-        if user_input not in user_memory:
-            user_memory.append(user_input)
-            print("✅ Added to memory:", user_input)
-
-
-
-    # Retrieve context from Chroma
-    docs = vectordb.similarity_search(user_input, k=3)
-    context = "\n\n".join([doc.page_content for doc in docs]) if docs else ""
-    memory_context = "\n".join(user_memory)
-
-    system_prompt = (
-        "You are a helpful AI assistant. Use ONLY the context and memory provided below. "
-        "If the answer is not in the context or memory, respond with: 'I don't know based on the context.'\n\n"
-        f"Context from documents:\n{context if context else 'None'}\n\n"
-        f"Memory from conversation:\n{memory_context if memory_context else 'None'}"
-    )
-
-    system_prompt = (
-        "You are a helpful AI assistant. "
-        "You MUST NOT use any knowledge from your pretraining. "
-        "Only use the information provided in the context or memory below. "
-        "If the information is not found in either, always reply with:\n"
-        "'I don't know based on the context.'\n\n"
-        f"Context from documents:\n{context if context else 'None'}\n\n"
-        f"Memory from conversation:\n{memory_context if memory_context else 'None'}"
-    )
-
-    system_prompt = (
-        "You are a helpful AI assistant.\n"
-        "You MUST NOT use any knowledge from your pretraining.\n"
-        "Use ONLY the information in the context or memory below.\n"
-        "ONLY include facts from memory or context **if they directly answer or support the user's input**.\n"
-        "If the information is not found, reply with:\n"
-        "'I don't know based on the context.'\n\n"
-        f"Context from documents:\n{context if context else 'None'}\n\n"
-        f"Memory from conversation:\n{memory_context if memory_context else 'None'}"
-    )
-
-    system_prompt = (
-        "You are a helpful AI assistant.\n"
-        "You must NEVER use your own knowledge or make any assumptions.\n"
-        "Only respond using the information provided in the CONTEXT and MEMORY sections below.\n"
-        "If the answer is not found there, you MUST reply with:\n"
-        "'I don't know based on the context.'\n"
-        "Do not guess or calculate anything that is not already mentioned.\n"
-        "Do not try to verify or correct any user-provided statements.\n\n"
-        f"CONTEXT:\n{context if context else 'None'}\n\n"
-        f"MEMORY:\n{memory_context if memory_context else 'None'}"
-    )
-
-
-    messages = [
-        {"role": "system", "content": system_prompt},
-        {"role": "user", "content": user_input}
-    ]
-
-    chat_completion = client.chat.completions.create(
-        model=GROQ_MODEL,
-        messages=messages
-    )
-
-    final_answer = chat_completion.choices[0].message.content.strip()
-    print("🤖 Model Response:", final_answer)
-
-
-    if not Isitquestion and final_answer.lower() == "i don't know based on the context." :
-        return "Got it"
-
-    return final_answer
-
-
-def search_google(query):
-    url = f"https://serpapi.com/search.json?q={query}&engine=google&api_key={SERPAPI_API_KEY}"
-    resp = requests.get(url)
-    data = resp.json()
-
-    if "answer_box" in data:
-        answer_box = data["answer_box"]
-        if "answer" in answer_box and answer_box["answer"]:
-            return answer_box["answer"]
-        if "snippet" in answer_box and answer_box["snippet"]:
-            return answer_box["snippet"]
-    
-    snippets = []
-    for i, result in enumerate(data.get("organic_results", [])[:3], 1):
-        snippet = result.get("snippet", "")
-        if snippet:
-            snippets.append(f"{i}. {snippet}")
-
-    if snippets:
-        return "\n\n".join(snippets)
-
-    return "No snippet found."
-
-
-
-def search_bing(query):
-    # Bing search via SerpAPI
-    url = f"https://serpapi.com/search.json?q={query}&engine=bing&api_key={SERPAPI_API_KEY}"
-    resp = requests.get(url)
-    data = resp.json()
-    try:
-        snippet = data['organic_results'][0].get('snippet', 'No snippet found.')
-    except Exception:
-        snippet = "No snippet found."
-    return snippet
-
-
-
-# Gradio UI
-def chat_interface(message, history):
-    if not message.strip():
-        return history, history, "" , gr.update(visible=False), gr.update(visible=False)  # hide buttons on empty input
-
-    reply = generate_answer(message)
-
-    messagegui = f"You said: {message}"
-    history.append({"role": "user", "content": messagegui})
-    history.append({"role": "assistant", "content": reply})
-
-    # Show buttons only if reply is exactly "I don't know based on the context."
-    show_buttons = reply.lower() == "i don't know based on the context."
-
-    return (
-        history,
-        history,
-        "",  # clear input box
-        gr.update(visible=show_buttons),
-        gr.update(visible=show_buttons),
-    )
-
-
-
-def google_search_button_click(history, state):
-    # Find last user message to search
-    last_user_message = ""
-    for msg in reversed(history):
-        if msg["role"] == "user":
-            # Strip "You said: " prefix to get original query
-            last_user_message = msg["content"].replace("You said: ", "", 1)
-            break
-
-    snippet = search_google(last_user_message)
-    history.append({"role": "assistant", "content": f"Google search snippet:\n{snippet}"})
-    # After search, keep buttons hidden until next unknown answer
-    return history, history, gr.update(visible=True), gr.update(visible=True)
-
-
-def bing_search_button_click(history, state):
-    last_user_message = ""
-    for msg in reversed(history):
-        if msg["role"] == "user":
-            last_user_message = msg["content"].replace("You said: ", "", 1)
-            break
-
-    snippet = search_bing(last_user_message)
-    history.append({"role": "assistant", "content": f"Bing search snippet:\n{snippet}"})
-    return history, history, gr.update(visible=True), gr.update(visible=True)
-
-
-# Launch UI
-with gr.Blocks() as demo:
-    gr.Markdown("## 🌍 Chatbot : countries and their capital city and population)")
-    chatbot = gr.Chatbot(label="Chat History", type="messages")
-
-    with gr.Row():
-        # Add buttons for search, initially hidden
-        google_btn = gr.Button("Search on Google", visible=False)
-        bing_btn = gr.Button("Search on Bing", visible=False)
-
-    msg = gr.Textbox(label="Your message", placeholder="Ask or tell me anything about countries and capital city", lines=2)
-
-    with gr.Row():
-        submit_btn = gr.Button("Submit")
-        clear = gr.Button("Clear chat")
-        clear_memory = gr.Button("Clear My Short Memory")  # 👈 New button
-
-
-
-
-    state = gr.State([])
-
-    submit_btn.click(chat_interface, [msg, state], [chatbot, state, msg, google_btn, bing_btn])
-    clear.click(lambda: ([], []), None, [chatbot, state])
-    clear_memory.click(clear_user_memory, None, [chatbot, state])
-
-    google_btn.click(google_search_button_click, [state, state], [chatbot, state, google_btn, bing_btn])
-    bing_btn.click(bing_search_button_click, [state, state], [chatbot, state, google_btn, bing_btn])
-
-
-demo.launch(server_name="0.0.0.0", server_port=7860, share=False)
-
+import os
+import logging
+import gradio as gr
+import re
+
+#from langchain_chroma import Chroma
+#from langchain_huggingface import HuggingFaceEmbeddings
+
+from langchain_community.vectorstores import Chroma
+from langchain_community.embeddings import HuggingFaceEmbeddings
+
+
+from groq import Groq
+
+from transformers import pipeline
+
+
+import requests  # for calling SerpAPI
+# Add your SerpAPI key here
+SERPAPI_API_KEY = "48125364c49a1952f1c8fdadf0d22e0d5bc3d195a5a98ad64d7e935aad503efa"
+
+# Load zero-shot classification model
+classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
+
+# Suppress noisy logs
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
+logging.getLogger('tensorflow').setLevel(logging.ERROR)
+logging.getLogger('torch').setLevel(logging.ERROR)
+
+# Config
+CHROMA_DIR = "chroma_country_info"
+GROQ_MODEL = "llama-3.3-70b-versatile"  # You can also try: mixtral-8x7b or gemma-7b-it
+
+
+
+
+# Load Groq client
+client = Groq(api_key="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")
+
+# Load vector DB
+print("🔍 Loading Chroma Vector DB...")
+embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+vectordb = Chroma(persist_directory=CHROMA_DIR, embedding_function=embeddings)
+
+# Memory for user-provided factual information (in-session only)
+user_memory = []
+
+
+
+
+def is_question(text):
+    
+    QUESTION_STARTERS = [
+    "who", "what", "when", "where", "why", "which", "whose", "whom", "how",
+    "is", "are", "am", "was", "were",
+    "do", "does", "did",
+    "have", "has", "had",
+    "can", "could", "will", "would", "shall", "should",
+    "may", "might", "must",
+    "didn't", "isn't", "aren't", "wasn't", "weren't",
+    "won't", "shouldn't", "couldn't", "wouldn't"
+    ]
+
+    text_clean = text.strip().lower()
+    
+    if text_clean.endswith('?'):
+        return True
+    
+    # Rule 2: Starts with known question word
+    words = text_clean.split()
+    if words and words[0] in QUESTION_STARTERS:
+        return True
+    
+    
+    return False
+
+def is_fact(text):
+
+    if is_question(text):
+        return False
+    
+    candidate_labels = ["fact", "opinion", "speculation"]
+    result = classifier(text, candidate_labels)
+    
+    label = result['labels'][0]
+    score = result['scores'][0]
+
+    # Rule: High-confidence fact => keep as fact
+    if label == "fact" and score > 0.75:
+        final_label = "fact"
+
+    # Rule: If label is opinion or speculation, but sentence sounds assertive => custom fact
+    elif label in ["opinion", "speculation"] and score < 0.9:
+        final_label = "custom fact"
+
+    # Rule: Low-confidence fact => custom fact
+    elif label == "fact" and score <= 0.75:
+        final_label = "custom fact"
+
+    else:
+        final_label = "custom fact"
+
+    # check if there is "fact in final_label and return True
+    
+    is_fact = "fact" in final_label
+
+    return is_fact
+
+def clear_user_memory():
+    user_memory.clear()
+    return [], []  # Clear chat and state too (optional)
+
+
+def generate_answer(user_input):
+    print("📚 generate_answer:\n", user_input)
+
+    # Handle memory commands
+    if "what do you remember" in user_input.lower():
+        if not user_memory:
+            return "I don't remember anything yet."
+        return "Here's what I remember:\n" + "\n".join(f"- {fact}" for fact in user_memory)
+
+    if "forget everything" in user_input.lower():
+        user_memory.clear()
+        return "Okay, I’ve forgotten everything you told me."
+
+    if not user_input.strip():
+        return "Please enter a question."
+
+    # Store statements as memory
+    Isitquestion=True
+    if is_fact(user_input):
+        Isitquestion=False
+        if user_input not in user_memory:
+            user_memory.append(user_input)
+            print("✅ Added to memory:", user_input)
+
+
+
+    # Retrieve context from Chroma
+    docs = vectordb.similarity_search(user_input, k=3)
+    context = "\n\n".join([doc.page_content for doc in docs]) if docs else ""
+    memory_context = "\n".join(user_memory)
+
+    system_prompt = (
+        "You are a helpful AI assistant. Use ONLY the context and memory provided below. "
+        "If the answer is not in the context or memory, respond with: 'I don't know based on the context.'\n\n"
+        f"Context from documents:\n{context if context else 'None'}\n\n"
+        f"Memory from conversation:\n{memory_context if memory_context else 'None'}"
+    )
+
+    system_prompt = (
+        "You are a helpful AI assistant. "
+        "You MUST NOT use any knowledge from your pretraining. "
+        "Only use the information provided in the context or memory below. "
+        "If the information is not found in either, always reply with:\n"
+        "'I don't know based on the context.'\n\n"
+        f"Context from documents:\n{context if context else 'None'}\n\n"
+        f"Memory from conversation:\n{memory_context if memory_context else 'None'}"
+    )
+
+    system_prompt = (
+        "You are a helpful AI assistant.\n"
+        "You MUST NOT use any knowledge from your pretraining.\n"
+        "Use ONLY the information in the context or memory below.\n"
+        "ONLY include facts from memory or context **if they directly answer or support the user's input**.\n"
+        "If the information is not found, reply with:\n"
+        "'I don't know based on the context.'\n\n"
+        f"Context from documents:\n{context if context else 'None'}\n\n"
+        f"Memory from conversation:\n{memory_context if memory_context else 'None'}"
+    )
+
+    system_prompt = (
+        "You are a helpful AI assistant.\n"
+        "You must NEVER use your own knowledge or make any assumptions.\n"
+        "Only respond using the information provided in the CONTEXT and MEMORY sections below.\n"
+        "If the answer is not found there, you MUST reply with:\n"
+        "'I don't know based on the context.'\n"
+        "Do not guess or calculate anything that is not already mentioned.\n"
+        "Do not try to verify or correct any user-provided statements.\n\n"
+        f"CONTEXT:\n{context if context else 'None'}\n\n"
+        f"MEMORY:\n{memory_context if memory_context else 'None'}"
+    )
+
+
+    messages = [
+        {"role": "system", "content": system_prompt},
+        {"role": "user", "content": user_input}
+    ]
+
+    chat_completion = client.chat.completions.create(
+        model=GROQ_MODEL,
+        messages=messages
+    )
+
+    final_answer = chat_completion.choices[0].message.content.strip()
+    print("🤖 Model Response:", final_answer)
+
+
+    if not Isitquestion and final_answer.lower() == "i don't know based on the context." :
+        return "Got it"
+
+    return final_answer
+
+
+def search_google(query):
+    url = f"https://serpapi.com/search.json?q={query}&engine=google&api_key={SERPAPI_API_KEY}"
+    resp = requests.get(url)
+    data = resp.json()
+
+    if "answer_box" in data:
+        answer_box = data["answer_box"]
+        if "answer" in answer_box and answer_box["answer"]:
+            return answer_box["answer"]
+        if "snippet" in answer_box and answer_box["snippet"]:
+            return answer_box["snippet"]
+    
+    snippets = []
+    for i, result in enumerate(data.get("organic_results", [])[:3], 1):
+        snippet = result.get("snippet", "")
+        if snippet:
+            snippets.append(f"{i}. {snippet}")
+
+    if snippets:
+        return "\n\n".join(snippets)
+
+    return "No snippet found."
+
+
+
+def search_bing(query):
+    # Bing search via SerpAPI
+    url = f"https://serpapi.com/search.json?q={query}&engine=bing&api_key={SERPAPI_API_KEY}"
+    resp = requests.get(url)
+    data = resp.json()
+    try:
+        snippet = data['organic_results'][0].get('snippet', 'No snippet found.')
+    except Exception:
+        snippet = "No snippet found."
+    return snippet
+
+
+
+# Gradio UI
+def chat_interface(message, history):
+    if not message.strip():
+        return history, history, "" , gr.update(visible=False), gr.update(visible=False)  # hide buttons on empty input
+
+    reply = generate_answer(message)
+
+    messagegui = f"You said: {message}"
+    history.append({"role": "user", "content": messagegui})
+    history.append({"role": "assistant", "content": reply})
+
+    # Show buttons only if reply is exactly "I don't know based on the context."
+    show_buttons = reply.lower() == "i don't know based on the context."
+
+    return (
+        history,
+        history,
+        "",  # clear input box
+        gr.update(visible=show_buttons),
+        gr.update(visible=show_buttons),
+    )
+
+
+
+def google_search_button_click(history, state):
+    # Find last user message to search
+    last_user_message = ""
+    for msg in reversed(history):
+        if msg["role"] == "user":
+            # Strip "You said: " prefix to get original query
+            last_user_message = msg["content"].replace("You said: ", "", 1)
+            break
+
+    snippet = search_google(last_user_message)
+    history.append({"role": "assistant", "content": f"Google search snippet:\n{snippet}"})
+    # After search, keep buttons hidden until next unknown answer
+    return history, history, gr.update(visible=True), gr.update(visible=True)
+
+
+def bing_search_button_click(history, state):
+    last_user_message = ""
+    for msg in reversed(history):
+        if msg["role"] == "user":
+            last_user_message = msg["content"].replace("You said: ", "", 1)
+            break
+
+    snippet = search_bing(last_user_message)
+    history.append({"role": "assistant", "content": f"Bing search snippet:\n{snippet}"})
+    return history, history, gr.update(visible=True), gr.update(visible=True)
+
+
+# Launch UI
+with gr.Blocks() as demo:
+    gr.Markdown("## 🌍 Chatbot : countries and their capital city and population)")
+    chatbot = gr.Chatbot(label="Chat History", type="messages")
+
+    with gr.Row():
+        # Add buttons for search, initially hidden
+        google_btn = gr.Button("Search on Google", visible=False)
+        bing_btn = gr.Button("Search on Bing", visible=False)
+
+    msg = gr.Textbox(label="Your message", placeholder="Ask or tell me anything about countries and capital city", lines=2)
+    myAuthToken = gr.Textbox(label="Token", placeholder="Type here the AUTH Token", lines=1)
+
+    with gr.Row():
+        submit_btn = gr.Button("Submit")
+        clear = gr.Button("Clear chat")
+        clear_memory = gr.Button("Clear My Short Memory")  # 👈 New button
+
+
+
+
+    state = gr.State([])
+
+    submit_btn.click(chat_interface, [msg, state], [chatbot, state, msg, google_btn, bing_btn])
+    clear.click(lambda: ([], []), None, [chatbot, state])
+    clear_memory.click(clear_user_memory, None, [chatbot, state])
+
+    google_btn.click(google_search_button_click, [state, state], [chatbot, state, google_btn, bing_btn])
+    bing_btn.click(bing_search_button_click, [state, state], [chatbot, state, google_btn, bing_btn])
+
+
+demo.launch(server_name="0.0.0.0", server_port=7860, share=False)
+