Update app.py

app.py

@@ -4,13 +4,6 @@ import requests
 from streamlit.components.v1 import html
 import os
 
-# Import transformers and cache the help agent for performance
-@st.cache_resource
-def get_help_agent():
-    from transformers import pipeline
-    # Using BlenderBot 400M Distill as the public conversational model (used elsewhere)
-    return pipeline("conversational", model="facebook/blenderbot-400M-distill")
-
 # Custom CSS for professional look (fixed text color)
 def inject_custom_css():
     st.markdown("""
@@ -45,73 +38,71 @@ def inject_custom_css():
             color: black !important;
         }
 
-        .answer-btn {
-            border-radius: 12px !important;
-            padding: 0.5rem 1.5rem !important;
-            font-weight: 600 !important;
-            margin: 0.5rem !important;
-        }
-
-        .yes-btn {
+        .mic-button {
+            margin-left: 10px;
+            padding: 5px 10px !important;
+            border-radius: 5px;
             background: #6C63FF !important;
             color: white !important;
+            border: none;
+            cursor: pointer;
         }
 
-        .no-btn {
-            background: #FF6B6B !important;
-            color: white !important;
-        }
-
-        .final-reveal {
-            animation: fadeIn 2s;
-            font-size: 2.5rem;
-            color: #6C63FF;
-            text-align: center;
-            margin: 2rem 0;
-        }
-
-        @keyframes fadeIn {
-            from { opacity: 0; }
-            to { opacity: 1; }
-        }
-
-        .confetti {
-            position: fixed;
-            top: 0;
-            left: 0;
-            width: 100%;
-            height: 100%;
-            pointer-events: none;
+        .language-selector {
+            position: absolute;
+            top: 10px;
+            right: 10px;
             z-index: 1000;
         }
-        
-        .confidence-meter {
-            height: 10px;
-            background: linear-gradient(90deg, #FF6B6B 0%, #6C63FF 100%);
-            border-radius: 5px;
-            margin: 10px 0;
-        }
     </style>
     """, unsafe_allow_html=True)
 
-# Confetti animation
-def show_confetti():
-    html("""
-    <canvas id="confetti-canvas" class="confetti"></canvas>
-    <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/confetti.browser.min.js"></script>
+# Speech-to-text component
+def speech_to_text(language, key):
+    component = f"""
     <script>
-    const canvas = document.getElementById('confetti-canvas');
-    const confetti = confetti.create(canvas, { resize: true });
-    confetti({
-        particleCount: 150,
-        spread: 70,
-        origin: { y: 0.6 }
-    });
-    setTimeout(() => { canvas.remove(); }, 5000);
+    function startRecording{key}() {{
+        const recognition = new (window.SpeechRecognition || window.webkitSpeechRecognition)();
+        recognition.lang = '{language}';
+        recognition.interimResults = false;
+        recognition.maxAlternatives = 1;
+
+        recognition.start();
+
+        recognition.onresult = function(event) {{
+            const transcript = event.results[0][0].transcript;
+            window.location.href = window.location.origin + window.location.pathname + '?speech_input=' + 
+                                  encodeURIComponent(transcript) + '&speech_key={key}';
+        }};
+
+        recognition.onerror = function(event) {{
+            console.error('Speech recognition error:', event.error);
+        }};
+    }}
     </script>
-    """)
+    <button class="mic-button" onclick="startRecording{key}()">
+        🎤
+    </button>
+    """
+    html(component, height=50)
+
+# Handle speech input
+def handle_speech_input():
+    params = st.query_params
+    if 'speech_input' in params and 'speech_key' in params:
+        speech_input = params['speech_input']
+        speech_key = params['speech_key']
+        st.session_state[speech_key] = speech_input
+        st.query_params.clear()
+        st.experimental_rerun()
 
-# Enhanced AI question generation for guessing game using Llama model
+# Import transformers and cache the help agent for performance
+@st.cache_resource
+def get_help_agent():
+    from transformers import pipeline
+    return pipeline("conversational", model="facebook/blenderbot-400M-distill")
+
+# Enhanced AI question generation
 def ask_llama(conversation_history, category, is_final_guess=False):
     api_url = "https://api.groq.com/openai/v1/chat/completions"
     headers = {
@@ -128,16 +119,9 @@ def ask_llama(conversation_history, category, is_final_guess=False):
 6. For objects: ask about size, color, usage, material, or where it's found
 7. Never repeat questions and always make progress toward guessing"""
 
-    if is_final_guess:
-        prompt = f"""Based on these answers about a {category}, provide ONLY your final guess with no extra text:
-{conversation_history}"""
-    else:
-        prompt = "Ask your next strategic yes/no question that will best narrow down the possibilities."
-
     messages = [
         {"role": "system", "content": system_prompt},
-        *conversation_history,
-        {"role": "user", "content": prompt}
+        *conversation_history
     ]
 
     data = {
@@ -155,47 +139,18 @@ def ask_llama(conversation_history, category, is_final_guess=False):
         st.error(f"Error calling Llama API: {str(e)}")
         return "Could not generate question"
 
-# New function for the help AI assistant using the Hugging Face InferenceClient
-def ask_help_agent(query):
-    try:
-        from huggingface_hub import InferenceClient
-        # Initialize the client with the provided model
-        client = InferenceClient("HuggingFaceH4/zephyr-7b-beta", token=os.environ.get("HF_HUB_TOKEN"))
-        system_message = "You are a friendly Chatbot."
-        
-        # Build history from session state (if any)
-        history = []
-        if "help_conversation" in st.session_state:
-            for msg in st.session_state.help_conversation:
-                # Each history entry is a tuple: (user query, assistant response)
-                history.append((msg.get("query", ""), msg.get("response", "")))
-        
-        messages = [{"role": "system", "content": system_message}]
-        for user_msg, bot_msg in history:
-            if user_msg:
-                messages.append({"role": "user", "content": user_msg})
-            if bot_msg:
-                messages.append({"role": "assistant", "content": bot_msg})
-        messages.append({"role": "user", "content": query})
-        
-        response_text = ""
-        # Using streaming to collect the entire response from the model
-        for message in client.chat_completion(
-            messages,
-            max_tokens=150,
-            stream=True,
-            temperature=0.7,
-            top_p=0.95,
-        ):
-            token = message.choices[0].delta.content
-            response_text += token
-        return response_text
-    except Exception as e:
-        return f"Error in help agent: {str(e)}"
-
 # Main game logic
 def main():
     inject_custom_css()
+    handle_speech_input()
+
+    # Language selector
+    lang_options = {"English": "en-US", "Urdu": "ur-PK"}
+    with st.container():
+        st.markdown('<div class="language-selector">', unsafe_allow_html=True)
+        selected_lang = st.selectbox("Voice Language", options=list(lang_options.keys()), label_visibility="collapsed")
+        st.markdown('</div>', unsafe_allow_html=True)
+        selected_lang_code = lang_options[selected_lang]
 
     st.markdown('<div class="title">KASOTI</div>', unsafe_allow_html=True)
     st.markdown('<div class="subtitle">The Smart Guessing Game</div>', unsafe_allow_html=True)
@@ -208,7 +163,7 @@ def main():
         st.session_state.conversation_history = []
         st.session_state.category = None
         st.session_state.final_guess = None
-        st.session_state.help_conversation = []  # separate history for help agent
+        st.session_state.help_conversation = []
 
     # Start screen
     if st.session_state.game_state == "start":
@@ -222,22 +177,37 @@ def main():
                 <li><strong>Place</strong> - city, country, landmark, geographical location</li>
                 <li><strong>Object</strong> - everyday item, tool, vehicle, etc.</li>
             </ul>
-            <p>Type your category below to begin:</p>
+            <p>Type or speak your category below to begin:</p>
         </div>
         """, unsafe_allow_html=True)
 
         with st.form("start_form"):
-            category_input = st.text_input("Enter category (person/place/object):").strip().lower()
+            col1, col2 = st.columns([4, 1])
+            with col1:
+                category_input = st.text_input("Enter category (person/place/object):", 
+                                              key="category_input").strip().lower()
+            with col2:
+                st.write("")
+                st.write("")
+                speech_to_text(language=selected_lang_code, key="category_input")
+            
             if st.form_submit_button("Start Game"):
-                if not category_input:
+                urdu_to_english = {
+                    'شخص': 'person',
+                    'جگہ': 'place',
+                    'چیز': 'object'
+                }
+                category = urdu_to_english.get(category_input, category_input)
+                
+                if not category:
                     st.error("Please enter a category!")
-                elif category_input not in ["person", "place", "object"]:
+                elif category not in ["person", "place", "object"]:
                     st.error("Please enter either 'person', 'place', or 'object'!")
                 else:
-                    st.session_state.category = category_input
+                    st.session_state.category = category
                     first_question = ask_llama([
                         {"role": "user", "content": "Ask your first strategic yes/no question."}
-                    ], category_input)
+                    ], category)
                     st.session_state.questions = [first_question]
                     st.session_state.conversation_history = [
                         {"role": "assistant", "content": first_question}
@@ -249,7 +219,6 @@ def main():
     elif st.session_state.game_state == "gameplay":
         current_question = st.session_state.questions[st.session_state.current_q]
         
-        # Check if AI made a guess
         if "Final Guess:" in current_question:
             st.session_state.final_guess = current_question.split("Final Guess:")[1].strip()
             st.session_state.game_state = "confirm_guess"
@@ -260,15 +229,30 @@ def main():
                     unsafe_allow_html=True)
         
         with st.form("answer_form"):
-            answer_input = st.text_input("Your answer (yes/no/both):", 
-                                       key=f"answer_{st.session_state.current_q}").strip().lower()
+            col1, col2 = st.columns([4, 1])
+            with col1:
+                answer_input = st.text_input("Your answer (yes/no/both):", 
+                                           key=f"answer_{st.session_state.current_q}").strip().lower()
+            with col2:
+                st.write("")
+                st.write("")
+                speech_to_text(language=selected_lang_code, key=f"answer_{st.session_state.current_q}")
+            
             if st.form_submit_button("Submit"):
-                if answer_input not in ["yes", "no", "both"]:
+                # Map Urdu responses to English
+                urdu_to_english = {
+                    'ہاں': 'yes',
+                    'نہیں': 'no',
+                    'دونوں': 'both'
+                }
+                processed_answer = urdu_to_english.get(answer_input, answer_input)
+                
+                if processed_answer not in ["yes", "no", "both"]:
                     st.error("Please answer with 'yes', 'no', or 'both'!")
                 else:
-                    st.session_state.answers.append(answer_input)
+                    st.session_state.answers.append(processed_answer)
                     st.session_state.conversation_history.append(
-                        {"role": "user", "content": answer_input}
+                        {"role": "user", "content": processed_answer}
                     )
 
                     # Generate next response
@@ -277,7 +261,6 @@ def main():
                         st.session_state.category
                     )
                     
-                    # Check if AI made a guess
                     if "Final Guess:" in next_response:
                         st.session_state.final_guess = next_response.split("Final Guess:")[1].strip()
                         st.session_state.game_state = "confirm_guess"
@@ -288,44 +271,41 @@ def main():
                         )
                         st.session_state.current_q += 1
 
-                        # Stop after 20 questions max
                         if st.session_state.current_q >= 20:
                             st.session_state.game_state = "result"
                     
                     st.experimental_rerun()
 
-        # Side Help Option: independent chat with an AI help assistant using Hugging Face model
-        with st.expander("Need Help? Chat with AI Assistant"):
-            help_query = st.text_input("Enter your help query:", key="help_query")
-            if st.button("Send", key="send_help"):
-                if help_query:
-                    help_response = ask_help_agent(help_query)
-                    st.session_state.help_conversation.append({"query": help_query, "response": help_response})
-                else:
-                    st.error("Please enter a query!")
-            if st.session_state.help_conversation:
-                for msg in st.session_state.help_conversation:
-                    st.markdown(f"**You:** {msg['query']}")
-                    st.markdown(f"**Help Assistant:** {msg['response']}")
-
-    # Guess confirmation screen using text input response
+    # Guess confirmation screen
     elif st.session_state.game_state == "confirm_guess":
         st.markdown(f'<div class="question-box">🤖 My Final Guess:<br><br>'
                     f'<strong>Is it {st.session_state.final_guess}?</strong></div>',
                     unsafe_allow_html=True)
         
         with st.form("confirm_form"):
-            confirm_input = st.text_input("Type your answer (yes/no/both):", key="confirm_input").strip().lower()
+            col1, col2 = st.columns([4, 1])
+            with col1:
+                confirm_input = st.text_input("Type your answer (yes/no/both):", key="confirm_input").strip().lower()
+            with col2:
+                st.write("")
+                st.write("")
+                speech_to_text(language=selected_lang_code, key="confirm_input")
+            
             if st.form_submit_button("Submit"):
-                if confirm_input not in ["yes", "no", "both"]:
+                urdu_to_english = {
+                    'ہاں': 'yes',
+                    'نہیں': 'no',
+                    'دونوں': 'both'
+                }
+                processed_confirm = urdu_to_english.get(confirm_input, confirm_input)
+                
+                if processed_confirm not in ["yes", "no", "both"]:
                     st.error("Please answer with 'yes', 'no', or 'both'!")
                 else:
-                    if confirm_input == "yes":
+                    if processed_confirm == "yes":
                         st.session_state.game_state = "result"
                         st.experimental_rerun()
-                        st.stop()  # Immediately halt further execution
                     else:
-                        # Add negative response to history and continue gameplay
                         st.session_state.conversation_history.append(
                             {"role": "user", "content": "no"}
                         )
@@ -344,7 +324,6 @@ def main():
     # Result screen
     elif st.session_state.game_state == "result":
         if not st.session_state.final_guess:
-            # Generate final guess if not already made
             qa_history = "\n".join(
                 [f"Q{i+1}: {q}\nA: {a}" 
                  for i, (q, a) in enumerate(zip(st.session_state.questions, st.session_state.answers))]
@@ -357,7 +336,21 @@ def main():
             )
             st.session_state.final_guess = final_guess.split("Final Guess:")[-1].strip()
 
-        show_confetti()
+        html("""
+        <canvas id="confetti-canvas"></canvas>
+        <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/confetti.browser.min.js"></script>
+        <script>
+        const canvas = document.getElementById('confetti-canvas');
+        const confetti = confetti.create(canvas, { resize: true });
+        confetti({
+            particleCount: 150,
+            spread: 70,
+            origin: { y: 0.6 }
+        });
+        setTimeout(() => { canvas.remove(); }, 5000);
+        </script>
+        """)
+        
         st.markdown(f'<div class="final-reveal">🎉 It\'s...</div>', unsafe_allow_html=True)
         time.sleep(1)
         st.markdown(f'<div class="final-reveal" style="font-size:3.5rem;color:#6C63FF;">{st.session_state.final_guess}</div>',
@@ -370,4 +363,4 @@ def main():
             st.experimental_rerun()
 
 if __name__ == "__main__":
-    main()
+    main()