Added new datasets and fixed gemini models

.gitattributes

@@ -34,5 +34,3 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 Data.csv filter=lfs diff=lfs merge=lfs -text
-CPCB_data.csv filter=lfs diff=lfs merge=lfs -text
-AQ_met_data.csv filter=lfs diff=lfs merge=lfs -text

AQ_met_data.csv

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d060bcf1c8cc8bc7c7fa8016fa573202218e3b434885e7022481fd488c5f8198
-size 143177747

app.py

@@ -19,14 +19,6 @@ from datasets import load_dataset, get_dataset_config_info, Dataset
 from PIL import Image
 import time
 import uuid
-import asyncio
-
-# Gemini API requires async
-try:
-    asyncio.get_running_loop()
-except RuntimeError:
-    loop = asyncio.new_event_loop()
-    asyncio.set_event_loop(loop)
 
 # Page config with beautiful theme
 st.set_page_config(
@@ -36,1117 +28,913 @@ st.set_page_config(
     initial_sidebar_state="expanded"
 )
 
-# CRITICAL: CSS must be loaded FIRST for immediate blue message styling
+# Custom CSS for beautiful styling
 st.markdown("""
 <style>
-/* User message styling - MUST be defined early */
-.user-message {
-    background: #3b82f6 !important;
-    color: white !important;
-    padding: 0.75rem 1rem !important;
-    border-radius: 7px !important;
-    max-width: 95% !important;
+/* Clean app background */
+.stApp {
+    background-color: #ffffff;
+    color: #212529;
+    font-family: 'Segoe UI', sans-serif;
 }
 
-/* Assistant message styling */
-.assistant-message {
-    background: #f1f5f9 !important;
-    color: #334155 !important;
-    padding: 0.75rem 1rem !important;
-    border-radius: 12px !important;
-    max-width: 85% !important;
+/* Reduce main container padding */
+.main .block-container {
+    padding-top: 0.5rem;
+    padding-bottom: 3rem;
+    max-width: 100%;
 }
 
-.assistant-info {
-    font-size: 0.875rem !important;
-    color: #6b7280 !important;
-    margin-bottom: 5px !important;
+/* Remove excessive spacing */
+.element-container {
+    margin-bottom: 0.5rem !important;
 }
-</style>
-""", unsafe_allow_html=True)
-
-# JavaScript for interactions
-# st.markdown("""
-# <script>
-# function scrollToBottom() {
-#     setTimeout(function() {
-#         const mainContainer = document.querySelector('.main-container');
-#         if (mainContainer) {
-#             mainContainer.scrollTop = mainContainer.scrollHeight;
-#         }
-#         window.scrollTo(0, document.body.scrollHeight);
-#     }, 100);
-# }
-
-# function toggleCode(header) {
-#     const codeBlock = header.nextElementSibling;
-#     const toggleText = header.querySelector('.toggle-text');
-    
-#     if (codeBlock.style.display === 'none') {
-#         codeBlock.style.display = 'block';
-#         toggleText.textContent = 'Click to collapse';
-#     } else {
-#         codeBlock.style.display = 'none';
-#         toggleText.textContent = 'Click to expand';
-#     }
-# }
-# </script>
-# """, unsafe_allow_html=True)
-
-# FORCE reload environment variables
-load_dotenv(override=True)
-
-# Get API keys
-Groq_Token = os.getenv("GROQ_API_KEY")
-hf_token = os.getenv("HF_TOKEN")
-gemini_token = os.getenv("GEMINI_TOKEN")
 
-# Model order is decided by this
-models = {
-    "gpt-oss-120b": "openai/gpt-oss-120b",
-    "qwen3-32b": "qwen/qwen3-32b",
-    "gpt-oss-20b": "openai/gpt-oss-20b",
-    "llama4 maverik":"meta-llama/llama-4-maverick-17b-128e-instruct",
-    "llama3.3": "llama-3.3-70b-versatile",
-    "deepseek-R1": "deepseek-r1-distill-llama-70b",
-    "gemini-2.5-flash": "gemini-2.5-flash",
-    "gemini-2.5-pro": "gemini-2.5-pro",
-    "gemini-2.5-flash-lite": "gemini-2.5-flash-lite",
-    "gemini-2.0-flash": "gemini-2.0-flash",
-    "gemini-2.0-flash-lite": "gemini-2.0-flash-lite",
-    # "llama4 scout":"meta-llama/llama-4-scout-17b-16e-instruct"
-    # "llama3.1": "llama-3.1-8b-instant"
+/* Fix sidebar spacing */
+[data-testid="stSidebar"] .element-container {
+    margin-bottom: 0.25rem !important;
 }
 
-self_path = os.path.dirname(os.path.abspath(__file__))
-
-# Initialize session ID for this session
-if "session_id" not in st.session_state:
-    st.session_state.session_id = str(uuid.uuid4())
-
-def upload_feedback(feedback, error, output, last_prompt, code, status):
-    """Enhanced feedback upload function with better logging and error handling"""
-    try:
-        if not hf_token or hf_token.strip() == "":
-            st.warning("Cannot upload feedback - HF_TOKEN not available")
-            return False
-
-        # Create comprehensive feedback data
-        feedback_data = {
-            "timestamp": datetime.now().isoformat(),
-            "session_id": st.session_state.session_id,
-            "feedback_score": feedback.get("score", ""),
-            "feedback_comment": feedback.get("text", ""),
-            "user_prompt": last_prompt,
-            "ai_output": str(output),
-            "generated_code": code or "",
-            "error_message": error or "",
-            "is_image_output": status.get("is_image", False),
-            "success": not bool(error)
-        }
+/* Sidebar */
+[data-testid="stSidebar"] {
+    background-color: #f8f9fa;
+    border-right: 1px solid #dee2e6;
+    padding: 1rem;
+}
 
-        # Create unique folder name with timestamp
-        timestamp_str = datetime.now().strftime("%Y%m%d_%H%M%S")
-        random_id = str(uuid.uuid4())[:8]
-        folder_name = f"feedback_{timestamp_str}_{random_id}"
-        
-        # Create markdown feedback file
-        markdown_content = f"""# VayuChat Feedback Report
+/* Optimize sidebar scrolling */
+[data-testid="stSidebar"] > div:first-child {
+    height: 100vh;
+    overflow-y: auto;
+    padding-bottom: 2rem;
+}
 
-## Session Information
-- **Timestamp**: {feedback_data['timestamp']}
-- **Session ID**: {feedback_data['session_id']}
+[data-testid="stSidebar"]::-webkit-scrollbar {
+    width: 6px;
+}
 
-## User Interaction
-**Prompt**: {feedback_data['user_prompt']}
+[data-testid="stSidebar"]::-webkit-scrollbar-track {
+    background: #f1f1f1;
+    border-radius: 3px;
+}
 
-## AI Response
-**Output**: {feedback_data['ai_output']}
+[data-testid="stSidebar"]::-webkit-scrollbar-thumb {
+    background: #c1c1c1;
+    border-radius: 3px;
+}
 
-## Generated Code
-```python
-{feedback_data['generated_code']}
-```
+[data-testid="stSidebar"]::-webkit-scrollbar-thumb:hover {
+    background: #a1a1a1;
+}
 
-## Technical Details
-- **Error Message**: {feedback_data['error_message']}
-- **Is Image Output**: {feedback_data['is_image_output']}
-- **Success**: {feedback_data['success']}
+/* Main title */
+.main-title {
+    text-align: center;
+    color: #343a40;
+    font-size: 2.5rem;
+    font-weight: 700;
+    margin-bottom: 0.5rem;
+}
 
-## User Feedback
-- **Score**: {feedback_data['feedback_score']}
-- **Comments**: {feedback_data['feedback_comment']}
-"""
+/* Subtitle */
+.subtitle {
+    text-align: center;
+    color: #6c757d;
+    font-size: 1.1rem;
+    margin-bottom: 1.5rem;
+}
 
-        # Save markdown file locally
-        markdown_filename = f"{folder_name}.md"
-        markdown_local_path = f"/tmp/{markdown_filename}"
-        
-        with open(markdown_local_path, "w", encoding="utf-8") as f:
-            f.write(markdown_content)
+/* Instructions */
+.instructions {
+    background-color: #f1f3f5;
+    border-left: 4px solid #0d6efd;
+    padding: 1rem;
+    margin-bottom: 1.5rem;
+    border-radius: 6px;
+    color: #495057;
+    text-align: left;
+}
 
-        # Upload to Hugging Face
-        api = HfApi(token=hf_token)
-        
-        # Upload markdown feedback
-        api.upload_file(
-            path_or_fileobj=markdown_local_path,
-            path_in_repo=f"data/{markdown_filename}",
-            repo_id="SustainabilityLabIITGN/VayuChat_Feedback",
-            repo_type="dataset",
-        )
-        
-        # Upload image if it exists and is an image output
-        if status.get("is_image", False) and isinstance(output, str) and os.path.exists(output):
-            try:
-                image_filename = f"{folder_name}_plot.png"
-                api.upload_file(
-                    path_or_fileobj=output,
-                    path_in_repo=f"data/{image_filename}",
-                    repo_id="SustainabilityLabIITGN/VayuChat_Feedback",
-                    repo_type="dataset",
-                )
-            except Exception as img_error:
-                print(f"Error uploading image: {img_error}")
-        
-        # Clean up local files
-        if os.path.exists(markdown_local_path):
-            os.remove(markdown_local_path)
-        
-        st.success("Feedback uploaded successfully!")
-        return True
-        
-    except Exception as e:
-        st.error(f"Error uploading feedback: {e}")
-        print(f"Feedback upload error: {e}")
-        return False
+/* Quick prompt buttons */
+.quick-prompt-container {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 8px;
+    margin-bottom: 1.5rem;
+    padding: 1rem;
+    background-color: #f8f9fa;
+    border-radius: 10px;
+    border: 1px solid #dee2e6;
+}
 
-# Filter available models
-available_models = []
-model_names = list(models.keys())
-groq_models = []
-gemini_models = []
-for model_name in model_names:
-    if "gemini" not in model_name:
-        groq_models.append(model_name)
-    else:
-        gemini_models.append(model_name)
-if Groq_Token and Groq_Token.strip():
-    available_models.extend(groq_models)
-if gemini_token and gemini_token.strip():
-    available_models.extend(gemini_models)
+.quick-prompt-btn {
+    background-color: #0d6efd;
+    color: white;
+    border: none;
+    padding: 8px 16px;
+    border-radius: 20px;
+    font-size: 0.9rem;
+    cursor: pointer;
+    transition: all 0.2s ease;
+    white-space: nowrap;
+}
 
-if not available_models:
-    st.error("No API keys available! Please set up your API keys in the .env file")
-    st.stop()
+.quick-prompt-btn:hover {
+    background-color: #0b5ed7;
+    transform: translateY(-2px);
+}
 
-# Set GPT-OSS-120B as default if available
-default_index = 0
-if "gpt-oss-120b" in available_models:
-    default_index = available_models.index("gpt-oss-120b")
-elif "deepseek-R1" in available_models:
-    default_index = available_models.index("deepseek-R1")
+/* User message styling */
+.user-message {
+    background: #3b82f6;
+    color: white;
+    padding: 0.75rem 1rem;
+    border-radius: 12px;
+    max-width: 70%;
+}
 
-# Compact header - everything perfectly aligned at same height
-st.markdown("""
-<style>
-.header-container {
-    display: flex; 
-    align-items: center; 
-    justify-content: center;
-    gap: 12px;
-    border-bottom: 1px solid #e5e7eb;
+.user-info {
+    font-size: 0.875rem;
+    opacity: 0.9;
+    margin-bottom: 3px;
 }
 
-.header-container img {
-    height: 80px;
+/* Assistant message styling */
+.assistant-message {
+    background: #f1f5f9;
+    color: #334155;
+    padding: 0.75rem 1rem;
+    border-radius: 12px;
+    max-width: 70%;
 }
 
-.header-container h1 {
-    padding: 0.25rem 0;
-    margin: 0; 
-    font-size: 1.5rem; 
-    font-weight: 700; 
-    color: #2563eb;
+.assistant-info {
+    font-size: 0.875rem;
+    color: #6b7280;
+    margin-bottom: 5px;
 }
 
-/* 🔹 Responsive: On small screens stack vertically */
-@media (max-width: 768px) {
-    .header-container {
-        flex-direction: column;
-        text-align: center;
-        gap: 0;
-        padding: 0 0 0.40rem;
-    }
-    .header-container img {
-        height: 60px;
-    }
-    .header-container h1 {
-        padding: 0 0;
-        font-size: 1.25rem;
-    }
+/* Processing indicator */
+.processing-indicator {
+    background: linear-gradient(135deg, #a8edea 0%, #fed6e3 100%);
+    color: #333;
+    padding: 1rem 1.5rem;
+    border-radius: 12px;
+    margin: 1rem 0;
+    margin-left: 0;
+    margin-right: auto;
+    max-width: 70%;
+    box-shadow: 0 2px 10px rgba(0,0,0,0.1);
+    animation: pulse 2s infinite;
 }
-</style>
-<div class="header-container">
-    <img src="https://sustainability-lab.github.io/images/logo_light.svg" />
-    <div style="display: flex; flex-direction: column; line-height: 1.2;">
-        <h1>VayuChat</h1>
-        <span>AI Air Quality Analysis • Sustainability Lab, IIT Gandhinagar</span>
-    </div>
-</div>
-""", unsafe_allow_html=True)
-
-# Load data with caching for better performance
-@st.cache_data
-def load_data():
-    return preprocess_and_load_df(join(self_path, "Data.csv"))
-
-try:
-    df = load_data()
-    # Data loaded silently - no success message needed
-except Exception as e:
-    st.error(f"Error loading data: {e}")
-    st.stop()
-
-inference_server = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
-image_path = "IITGN_Logo.png"
-
-# Clean sidebar  
-with st.sidebar:
-    # Model selector at top of sidebar for easy access
-    model_name = st.selectbox(
-        "🤖 AI Model:",
-        available_models,
-        index=default_index,
-        help="Choose your AI model - easily accessible without scrolling!"
-    )
-    
-    st.markdown("---")
-    
-    # Quick Queries Section
-    st.markdown("### 💭 Quick Queries")
-    
-    # Load quick prompts with caching
-    @st.cache_data
-    def load_questions():
-        questions = []
-        questions_file = join(self_path, "questions.txt")
-        if os.path.exists(questions_file):
-            try:
-                with open(questions_file, 'r', encoding='utf-8') as f:
-                    content = f.read()
-                    questions = [q.strip() for q in content.split("\n") if q.strip()]
-            except Exception as e:
-                questions = []
-        return questions
-    
-    questions = load_questions()
-    
-    # Add default prompts if file doesn't exist or is empty
-    if not questions:
-        questions = [
-            "Which month had highest pollution?",
-            "Which city has worst air quality?", 
-            "Show annual PM2.5 average",
-            "Plot monthly average PM2.5 for 2023",
-            "List all cities by pollution level",
-            "Compare winter vs summer pollution",
-            "Show seasonal pollution patterns",
-            "Which areas exceed WHO guidelines?",
-            "What are peak pollution hours?",
-            "Show PM10 vs PM2.5 comparison",
-            "Which station records highest variability in PM2.5?",
-            "Calculate pollution improvement rate year-over-year by city",
-            "Identify cities with PM2.5 levels consistently above 50 μg/m³ for >6 months",
-            "Find correlation between PM2.5 and PM10 across different seasons and cities",
-            "Compare weekday vs weekend levels",
-            "Plot yearly trend analysis",
-            "Show pollution distribution by city",
-            "Create correlation plot between pollutants"
-        ]
-    
-    # Quick query buttons in sidebar
-    selected_prompt = None
-    
-    
-    # Show all questions but in a scrollable format
-    if len(questions) > 0:
-        st.markdown("**Select a question to analyze:**")
-        
-        # Getting Started section with simple questions
-        getting_started_questions = questions[:10]  # First 10 simple questions
-        with st.expander("🚀 Getting Started - Simple Questions", expanded=True):
-            for i, q in enumerate(getting_started_questions):
-                if st.button(q, key=f"start_q_{i}", use_container_width=True, help=f"Analyze: {q}"):
-                    selected_prompt = q
-                    st.session_state.last_selected_prompt = q
-        
-        # Create expandable sections for better organization
-        with st.expander("📊 NCAP Funding & Policy Analysis", expanded=False):
-            for i, q in enumerate([q for q in questions if any(word in q.lower() for word in ['ncap', 'funding', 'investment', 'rupee'])]):
-                if st.button(q, key=f"ncap_q_{i}", use_container_width=True, help=f"Analyze: {q}"):
-                    selected_prompt = q
-                    st.session_state.last_selected_prompt = q
-        
-        with st.expander("🌬️ Meteorology & Environmental Factors", expanded=False):
-            for i, q in enumerate([q for q in questions if any(word in q.lower() for word in ['wind', 'temperature', 'humidity', 'rainfall', 'meteorological', 'monsoon', 'barometric'])]):
-                if st.button(q, key=f"met_q_{i}", use_container_width=True, help=f"Analyze: {q}"):
-                    selected_prompt = q
-                    st.session_state.last_selected_prompt = q
-        
-        with st.expander("👥 Population & Demographics", expanded=False):
-            for i, q in enumerate([q for q in questions if any(word in q.lower() for word in ['population', 'capita', 'density', 'exposure'])]):
-                if st.button(q, key=f"pop_q_{i}", use_container_width=True, help=f"Analyze: {q}"):
-                    selected_prompt = q
-                    st.session_state.last_selected_prompt = q
-        
-        with st.expander("🏭 Multi-Pollutant Analysis", expanded=False):
-            for i, q in enumerate([q for q in questions if any(word in q.lower() for word in ['ozone', 'no2', 'correlation', 'multi-pollutant', 'interaction'])]):
-                if st.button(q, key=f"multi_q_{i}", use_container_width=True, help=f"Analyze: {q}"):
-                    selected_prompt = q
-                    st.session_state.last_selected_prompt = q
-        
-        with st.expander("📈 Other Analysis Questions", expanded=False):
-            remaining_questions = [q for q in questions if not any(any(word in q.lower() for word in category) for category in [
-                ['ncap', 'funding', 'investment', 'rupee'],
-                ['wind', 'temperature', 'humidity', 'rainfall', 'meteorological', 'monsoon', 'barometric'],
-                ['population', 'capita', 'density', 'exposure'],
-                ['ozone', 'no2', 'correlation', 'multi-pollutant', 'interaction']
-            ])]
-            for i, q in enumerate(remaining_questions):
-                if st.button(q, key=f"other_q_{i}", use_container_width=True, help=f"Analyze: {q}"):
-                    selected_prompt = q
-                    st.session_state.last_selected_prompt = q
-    
-    st.markdown("---")
-    
-    
-    # Clear Chat Button
-    if st.button("Clear Chat", use_container_width=True):
-        st.session_state.responses = []
-        st.session_state.processing = False
-        st.session_state.session_id = str(uuid.uuid4())
-        try:
-            st.rerun()
-        except AttributeError:
-            st.experimental_rerun()
-
-# Initialize session state first
-if "responses" not in st.session_state:
-    st.session_state.responses = []
-if "processing" not in st.session_state:
-    st.session_state.processing = False
-if "session_id" not in st.session_state:
-    st.session_state.session_id = str(uuid.uuid4())
-
-
-
-
-def show_custom_response(response):
-    """Custom response display function with improved styling"""
-    role = response.get("role", "assistant")
-    content = response.get("content", "")
-    
-    if role == "user":
-        # User message with right alignment - CSS now loaded at top of file
-        st.markdown(f"""
-        <div style='display: flex; justify-content: flex-end; margin: 1rem 0;'>
-            <div class='user-message'>
-                {content}
-            </div>
-        </div>
-        """, unsafe_allow_html=True)
-    elif role == "assistant":
-        # Check if content is an image filename - don't display the filename text
-        is_image_path = isinstance(content, str) and any(ext in content for ext in ['.png', '.jpg', '.jpeg'])
-        
-        # Check if content is a pandas DataFrame
-        import pandas as pd
-        is_dataframe = isinstance(content, pd.DataFrame)
-        
-        # Check for errors first and display them with special styling
-        error = response.get("error")
-        timestamp = response.get("timestamp", "")
-        timestamp_display = f" • {timestamp}" if timestamp else ""
-        
-        if error:
-            st.markdown(f"""
-            <div style='display: flex; justify-content: flex-start; margin: 1rem 0;'>
-                <div class='assistant-message'>
-                    <div class='assistant-info'>VayuChat{timestamp_display}</div>
-                    <div class='error-message'>
-                        ⚠️ <strong>Error:</strong> {error}
-                        <br><br>
-                        <em>💡 Try rephrasing your question or being more specific about what you'd like to analyze.</em>
-                    </div>
-                </div>
-            </div>
-            """, unsafe_allow_html=True)
-        # Assistant message with left alignment - reduced margins
-        elif not is_image_path and not is_dataframe:
-            st.markdown(f"""
-            <div style='display: flex; justify-content: flex-start; margin: 1rem 0;'>
-                <div class='assistant-message'>
-                    <div class='assistant-info'>VayuChat{timestamp_display}</div>
-                    {content if isinstance(content, str) else str(content)}
-                </div>
-            </div>
-            """, unsafe_allow_html=True)
-        elif is_dataframe:
-            # Display DataFrame with nice formatting
-            st.markdown(f"""
-            <div style='display: flex; justify-content: flex-start; margin: 1rem 0;'>
-                <div class='assistant-message'>
-                    <div class='assistant-info'>VayuChat{timestamp_display}</div>
-                    Here are the results:
-                </div>
-            </div>
-            """, unsafe_allow_html=True)
-            
-            # Add context info for dataframes
-            st.markdown("""
-            <div class='context-info'>
-                💡 This table is interactive - click column headers to sort, or scroll to view all data.
-            </div>
-            """, unsafe_allow_html=True)
-            
-            # Display dataframe with built-in download functionality
-            st.dataframe(
-                content, 
-                use_container_width=True,
-                hide_index=True,
-                column_config=None
-            )
-        
-        # Show generated code with Streamlit expander
-        if response.get("gen_code"):
-            with st.expander("📋 View Generated Code", expanded=False):
-                st.code(response["gen_code"], language="python")
-        
-        # Check if this is a plot response (plots are now displayed directly via st.pyplot)
-        is_plot_response = isinstance(content, str) and "Plot displayed successfully" in content
-        
-        # Try to display image if content is a file path (for backward compatibility)
-        try:
-            if isinstance(content, str) and (content.endswith('.png') or content.endswith('.jpg')):
-                if os.path.exists(content):
-                    # Display image with better styling and reasonable width
-                    st.markdown("""
-                    <div style='margin: 1rem 0; display: flex; justify-content: center;'>
-                    </div>
-                    """, unsafe_allow_html=True)
-                    st.image(content, width=1080, caption="Generated Visualization")
-                    return {"is_image": True}
-            # Also handle case where content shows filename but we want to show image
-            elif isinstance(content, str) and any(ext in content for ext in ['.png', '.jpg']):
-                # Extract potential filename from content
-                import re
-                filename_match = re.search(r'([^/\\]+\.(?:png|jpg|jpeg))', content)
-                if filename_match:
-                    filename = filename_match.group(1)
-                    if os.path.exists(filename):
-                        st.markdown("""
-                        <div style='margin: 1rem 0; display: flex; justify-content: center;'>
-                        </div>
-                        """, unsafe_allow_html=True)
-                        st.image(filename, width=1080, caption="Generated Visualization")
-                        return {"is_image": True}
-        except:
-            pass
-            
-        return {"is_image": False}
 
+@keyframes pulse {
+    0% { opacity: 1; }
+    50% { opacity: 0.7; }
+    100% { opacity: 1; }
+}
 
-# Chat history
-# Display chat history
-for response_id, response in enumerate(st.session_state.responses):
-    status = show_custom_response(response)
-    
-    # Show feedback section for assistant responses
-    if response["role"] == "assistant":
-        feedback_key = f"feedback_{int(response_id/2)}"
-        error = response.get("error", "")
-        output = response.get("content", "")
-        last_prompt = response.get("last_prompt", "")
-        code = response.get("gen_code", "")
-
+/* Feedback box */
+.feedback-section {
+    background-color: #f8f9fa;
+    border: 1px solid #dee2e6;
+    padding: 1rem;
+    border-radius: 8px;
+    margin: 1rem 0;
+}
 
-        # Beautiful action bar with feedback and retry
-        st.markdown('<div style="margin: 1.5rem 0 0.5rem 0;"></div>', unsafe_allow_html=True)  # Spacer
-        
-        if "feedback" in st.session_state.responses[response_id]:
-            # Show submitted feedback nicely
-            feedback_data = st.session_state.responses[response_id]["feedback"]
-            col1, col2 = st.columns([3, 1])
-            with col1:
-                st.markdown(f"""
-                <div style='
-                    background: linear-gradient(135deg, #ecfdf5 0%, #d1fae5 100%);
-                    border: 1px solid #a7f3d0;
-                    border-radius: 8px;
-                    padding: 0.75rem 1rem;
-                    display: flex;
-                    align-items: center;
-                    gap: 8px;
-                '>
-                    <span style='font-size: 1.1rem;'>{feedback_data.get('score', '')}</span>
-                    <span style='color: #059669; font-weight: 500; font-size: 0.9rem;'>
-                        Thanks for your feedback!
-                    </span>
-                </div>
-                """, unsafe_allow_html=True)
-            with col2:
-                if st.button("🔄 Retry", key=f"retry_{response_id}", use_container_width=True):
-                    user_prompt = ""
-                    if response_id > 0:
-                        user_prompt = st.session_state.responses[response_id-1].get("content", "")
-                    if user_prompt:
-                        if response_id > 0:
-                            retry_prompt = st.session_state.responses[response_id-1].get("content", "")
-                            del st.session_state.responses[response_id]
-                            del st.session_state.responses[response_id-1]
-                            st.session_state.follow_up_prompt = retry_prompt
-                            st.rerun()
-        else:
-            # Clean feedback and retry layout
-            col1, col2, col3, col4 = st.columns([2, 2, 1, 1])
-            
-            with col1:
-                if st.button("✨ Excellent", key=f"{feedback_key}_excellent", use_container_width=True):
-                    feedback = {"score": "✨ Excellent", "text": ""}
-                    st.session_state.responses[response_id]["feedback"] = feedback
-                    st.rerun()
-            
-            with col2:
-                if st.button("🔧 Needs work", key=f"{feedback_key}_poor", use_container_width=True):
-                    feedback = {"score": "🔧 Needs work", "text": ""}
-                    st.session_state.responses[response_id]["feedback"] = feedback
-                    st.rerun()
-            
-            with col4:
-                if st.button("🔄 Retry", key=f"retry_{response_id}", use_container_width=True):
-                    user_prompt = ""
-                    if response_id > 0:
-                        user_prompt = st.session_state.responses[response_id-1].get("content", "")
-                    if user_prompt:
-                        if response_id > 0:
-                            retry_prompt = st.session_state.responses[response_id-1].get("content", "")
-                            del st.session_state.responses[response_id]
-                            del st.session_state.responses[response_id-1]
-                            st.session_state.follow_up_prompt = retry_prompt
-                            st.rerun()
+/* Success and error messages */
+.success-message {
+    background-color: #d1e7dd;
+    color: #0f5132;
+    padding: 1rem;
+    border-radius: 6px;
+    border: 1px solid #badbcc;
+}
 
-# Chat input with better guidance
-prompt = st.chat_input("💬 Ask about air quality trends, pollution analysis, or city comparisons...", key="main_chat")
+.error-message {
+    background-color: #f8d7da;
+    color: #842029;
+    padding: 1rem;
+    border-radius: 6px;
+    border: 1px solid #f5c2c7;
+}
 
-# Handle selected prompt from quick prompts
-if selected_prompt:
-    prompt = selected_prompt
+/* Chat input styling like mockup */
+.stChatInput {
+    border-radius: 8px;
+    border: 1px solid #d1d5db;
+    background: #ffffff;
+    padding: 0.75rem 1rem;
+    font-size: 1rem;
+}
 
-# Handle follow-up prompts from quick action buttons
-if st.session_state.get("follow_up_prompt") and not st.session_state.get("processing"):
-    prompt = st.session_state.follow_up_prompt
-    st.session_state.follow_up_prompt = None  # Clear the follow-up prompt
+.stChatInput:focus {
+    border-color: #3b82f6;
+    box-shadow: 0 0 0 3px rgba(59, 130, 246, 0.1);
+}
 
-# Handle new queries
-if prompt and not st.session_state.get("processing"):
-    # Prevent duplicate processing
-    if "last_prompt" in st.session_state:
-        last_prompt = st.session_state["last_prompt"]
-        last_model_name = st.session_state.get("last_model_name", "")
-        if (prompt == last_prompt) and (model_name == last_model_name):
-            prompt = None
+/* Button */
+.stButton > button {
+    background-color: #0d6efd;
+    color: white;
+    border-radius: 6px;
+    padding: 0.5rem 1.25rem;
+    border: none;
+    font-weight: 600;
+    transition: background-color 0.2s ease;
+}
 
-    if prompt:
-        # Add user input to chat history
-        user_response = get_from_user(prompt)
-        st.session_state.responses.append(user_response)
-        
-        # Set processing state
-        st.session_state.processing = True
-        st.session_state.current_model = model_name
-        st.session_state.current_question = prompt
-        
-        # Rerun to show processing indicator
-        st.rerun()
+.stButton > button:hover {
+    background-color: #0b5ed7;
+}
 
-# Process the question if we're in processing state
-if st.session_state.get("processing"):
-    # Enhanced processing indicator like Claude Code
-    st.markdown("""
-    <div style='padding: 1rem; text-align: center; background: #f8fafc; border-radius: 8px; margin: 1rem 0;'>
-        <div style='display: flex; align-items: center; justify-content: center; gap: 0.5rem; color: #475569;'>
-            <div style='font-weight: 500;'>🤖 Processing with """ + str(st.session_state.get('current_model', 'Unknown')) + """</div>
-            <div class='dots' style='display: inline-flex; gap: 2px;'>
-                <div class='dot' style='width: 4px; height: 4px; background: #3b82f6; border-radius: 50%; animation: bounce 1.4s infinite ease-in-out;'></div>
-                <div class='dot' style='width: 4px; height: 4px; background: #3b82f6; border-radius: 50%; animation: bounce 1.4s infinite ease-in-out; animation-delay: 0.16s;'></div>
-                <div class='dot' style='width: 4px; height: 4px; background: #3b82f6; border-radius: 50%; animation: bounce 1.4s infinite ease-in-out; animation-delay: 0.32s;'></div>
-            </div>
-        </div>
-        <div style='font-size: 0.75rem; color: #6b7280; margin-top: 0.25rem;'>Analyzing data and generating response...</div>
-    </div>
-    <style>
-    @keyframes bounce {
-        0%, 80%, 100% { transform: scale(0.8); opacity: 0.5; }
-        40% { transform: scale(1.2); opacity: 1; }
-    }
-    </style>
-    """, unsafe_allow_html=True)
-    
-    prompt = st.session_state.get("current_question")
-    model_name = st.session_state.get("current_model")
-    
-    try:
-        response = ask_question(model_name=model_name, question=prompt)
-        
-        if not isinstance(response, dict):
-            response = {
-                "role": "assistant",
-                "content": "Error: Invalid response format",
-                "gen_code": "",
-                "ex_code": "",
-                "last_prompt": prompt,
-                "error": "Invalid response format",
-                "timestamp": datetime.now().strftime("%H:%M")
-            }
-        
-        response.setdefault("role", "assistant")
-        response.setdefault("content", "No content generated")
-        response.setdefault("gen_code", "")
-        response.setdefault("ex_code", "")
-        response.setdefault("last_prompt", prompt)
-        response.setdefault("error", None)
-        response.setdefault("timestamp", datetime.now().strftime("%H:%M"))
-        
-    except Exception as e:
-        response = {
-            "role": "assistant",
-            "content": f"Sorry, I encountered an error: {str(e)}",
-            "gen_code": "",
-            "ex_code": "",
-            "last_prompt": prompt,
-            "error": str(e),
-            "timestamp": datetime.now().strftime("%H:%M")
-        }
+/* Sidebar button styling - smaller, left-aligned */
+[data-testid="stSidebar"] .stButton > button {
+    background-color: #f8fafc;
+    color: #475569;
+    border: 1px solid #e2e8f0;
+    padding: 0.375rem 0.75rem;
+    font-size: 0.65rem;
+    font-weight: normal;
+    text-align: left;
+    white-space: normal;
+    height: auto;
+    line-height: 1.2;
+    transition: all 0.2s ease;
+    cursor: pointer;
+    margin-bottom: 0.25rem;
+    width: 100%;
+    display: flex;
+    justify-content: flex-start;
+}
 
-    st.session_state.responses.append(response)
-    st.session_state["last_prompt"] = prompt
-    st.session_state["last_model_name"] = model_name
-    st.session_state.processing = False
-    
-    # Clear processing state
-    if "current_model" in st.session_state:
-        del st.session_state.current_model
-    if "current_question" in st.session_state:
-        del st.session_state.current_question
-    
-    st.rerun()
+[data-testid="stSidebar"] .stButton > button:hover {
+    background-color: #e0f2fe;
+    border-color: #0ea5e9;
+    color: #0c4a6e;
+}
 
-# Close chat container
-st.markdown("</div>", unsafe_allow_html=True)
+[data-testid="stSidebar"] .stButton > button:active {
+    transform: translateY(0);
+    box-shadow: none;
+}
 
-# Minimal auto-scroll - only scroll when processing
-if st.session_state.get("processing"):
-    st.markdown("<script>scrollToBottom();</script>", unsafe_allow_html=True)
+/* Code container styling */
+.code-container {
+    margin: 1rem 0;
+    border: 1px solid #d1d5db;
+    border-radius: 12px;
+    background: white;
+    box-shadow: 0 1px 3px rgba(0, 0, 0, 0.1);
+}
 
-    # Dataset Info Section (matching mockup)
-    st.markdown("### Dataset Info")
-    st.markdown("""
-    <div style='background: #f1f5f9; border-radius: 8px; padding: 1rem; margin-bottom: 1rem;'>
-        <h4 style='margin: 0 0 0.5rem 0; color: #1e293b; font-size: 0.9rem;'>PM2.5 Air Quality Data</h4>
-        <p style='margin: 0; font-size: 0.75rem; color: #475569;'><strong>Time Range:</strong> 2022 - 2023</p>
-        <p style='margin: 0; font-size: 0.75rem; color: #475569;'><strong>Locations:</strong> 300+ cities across India</p>
-        <p style='margin: 0; font-size: 0.75rem; color: #475569;'><strong>Records:</strong> 100,000+ measurements</p>
-    </div>
-    """, unsafe_allow_html=True)
-    
+.code-header {
+    display: flex;
+    justify-content: space-between;
+    align-items: center;
+    padding: 0.875rem 1.25rem;
+    background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%);
+    border-bottom: 1px solid #e2e8f0;
+    cursor: pointer;
+    transition: all 0.2s ease;
+    border-radius: 12px 12px 0 0;
+}
 
-# streamlit adds each markdown's div, so its better to keep this in the last
-# Custom CSS for beautiful styling
-st.markdown("""
-<style>
-/* Clean app background */
-.stApp {
-    background-color: #ffffff;
-    color: #212529;
-    font-family: 'Segoe UI', sans-serif;
+.code-header:hover {
+    background: linear-gradient(135deg, #e2e8f0 0%, #cbd5e1 100%);
 }
 
-/* Reduce main container padding */
-.main .block-container {
-    padding-top: 0px;
-    padding-bottom: 3rem;
-    max-width: 100%;
+.code-title {
+    font-size: 0.9rem;
+    font-weight: 600;
+    color: #1e293b;
+    display: flex;
+    align-items: center;
+    gap: 0.5rem;
 }
 
-/* Remove excessive spacing */
-.element-container {
-    margin-bottom: 0.5rem !important;
+.code-title:before {
+    content: "⚡";
+    font-size: 0.8rem;
 }
 
-/* Fix sidebar spacing */
-[data-testid="stSidebar"] .element-container {
-    margin-bottom: 0.25rem !important;
+.toggle-text {
+    font-size: 0.75rem;
+    color: #64748b;
+    font-weight: 500;
 }
 
-/* Sidebar */
-[data-testid="stSidebar"] {
-    background-color: #f8f9fa;
-    border-right: 1px solid #dee2e6;
-    padding: 1rem;
+.code-block {
+    background: linear-gradient(135deg, #0f172a 0%, #1e293b 100%);
+    color: #e2e8f0;
+    padding: 1.5rem;
+    font-family: 'SF Mono', 'Monaco', 'Menlo', 'Consolas', monospace;
+    font-size: 0.875rem;
+    overflow-x: auto;
+    line-height: 1.6;
+    border-radius: 0 0 12px 12px;
 }
 
-/* Optimize sidebar scrolling */
-[data-testid="stSidebar"] > div:first-child {
-    height: 100vh;
-    overflow-y: auto;
-    padding-bottom: 2rem;
+.answer-container {
+    background: #f8fafc;
+    border: 1px solid #e2e8f0;
+    border-radius: 8px;
+    padding: 1.5rem;
+    margin: 1rem 0;
 }
 
-[data-testid="stSidebar"]::-webkit-scrollbar {
-    width: 6px;
+.answer-text {
+    font-size: 1.125rem;
+    color: #1e293b;
+    line-height: 1.6;
+    margin-bottom: 1rem;
 }
 
-[data-testid="stSidebar"]::-webkit-scrollbar-track {
-    background: #f1f1f1;
-    border-radius: 3px;
+.answer-highlight {
+    background: #fef3c7;
+    padding: 0.125rem 0.375rem;
+    border-radius: 4px;
+    font-weight: 600;
+    color: #92400e;
 }
 
-[data-testid="stSidebar"]::-webkit-scrollbar-thumb {
-    background: #c1c1c1;
-    border-radius: 3px;
+.context-info {
+    background: #f1f5f9;
+    border-left: 4px solid #3b82f6;
+    padding: 0.75rem 1rem;
+    margin: 1rem 0;
+    font-size: 0.875rem;
+    color: #475569;
 }
 
-[data-testid="stSidebar"]::-webkit-scrollbar-thumb:hover {
-    background: #a1a1a1;
+/* Hide default menu and footer */
+#MainMenu {visibility: hidden;}
+footer {visibility: hidden;}
+header {visibility: hidden;}
+
+/* Auto scroll */
+.main-container {
+    height: 70vh;
+    overflow-y: auto;
 }
+</style>
+""", unsafe_allow_html=True)
 
-/* Main title */
-.main-title {
-    text-align: center;
-    color: #343a40;
-    font-size: 2.5rem;
-    font-weight: 700;
-    margin-bottom: 0.5rem;
+# JavaScript for interactions
+st.markdown("""
+<script>
+function scrollToBottom() {
+    setTimeout(function() {
+        const mainContainer = document.querySelector('.main-container');
+        if (mainContainer) {
+            mainContainer.scrollTop = mainContainer.scrollHeight;
+        }
+        window.scrollTo(0, document.body.scrollHeight);
+    }, 100);
 }
 
-/* Subtitle */
-.subtitle {
-    text-align: center;
-    color: #6c757d;
-    font-size: 1.1rem;
-    margin-bottom: 1.5rem;
+function toggleCode(header) {
+    const codeBlock = header.nextElementSibling;
+    const toggleText = header.querySelector('.toggle-text');
+    
+    if (codeBlock.style.display === 'none') {
+        codeBlock.style.display = 'block';
+        toggleText.textContent = 'Click to collapse';
+    } else {
+        codeBlock.style.display = 'none';
+        toggleText.textContent = 'Click to expand';
+    }
+}
+</script>
+""", unsafe_allow_html=True)
+
+# FORCE reload environment variables
+load_dotenv(override=True)
+
+# Get API keys
+Groq_Token = os.getenv("GROQ_API_KEY")
+hf_token = os.getenv("HF_TOKEN")
+gemini_token = os.getenv("GEMINI_TOKEN")
+
+models = {
+    "gpt-oss-20b": "openai/gpt-oss-20b",
+    "gpt-oss-120b": "openai/gpt-oss-120b",
+    "llama3.1": "llama-3.1-8b-instant",
+    "llama3.3": "llama-3.3-70b-versatile",
+    "deepseek-R1": "deepseek-r1-distill-llama-70b",
+    "llama4 maverik":"meta-llama/llama-4-maverick-17b-128e-instruct",
+    "llama4 scout":"meta-llama/llama-4-scout-17b-16e-instruct",
+    "gemini-pro": "gemini-1.5-pro"
 }
 
-/* Instructions */
-.instructions {
-    background-color: #f1f3f5;
-    border-left: 4px solid #0d6efd;
-    padding: 1rem;
-    margin-bottom: 1.5rem;
-    border-radius: 6px;
-    color: #495057;
-    text-align: left;
-}
+self_path = os.path.dirname(os.path.abspath(__file__))
+
+# Initialize session ID for this session
+if "session_id" not in st.session_state:
+    st.session_state.session_id = str(uuid.uuid4())
+
+def upload_feedback(feedback, error, output, last_prompt, code, status):
+    """Enhanced feedback upload function with better logging and error handling"""
+    try:
+        if not hf_token or hf_token.strip() == "":
+            st.warning("Cannot upload feedback - HF_TOKEN not available")
+            return False
+
+        # Create comprehensive feedback data
+        feedback_data = {
+            "timestamp": datetime.now().isoformat(),
+            "session_id": st.session_state.session_id,
+            "feedback_score": feedback.get("score", ""),
+            "feedback_comment": feedback.get("text", ""),
+            "user_prompt": last_prompt,
+            "ai_output": str(output),
+            "generated_code": code or "",
+            "error_message": error or "",
+            "is_image_output": status.get("is_image", False),
+            "success": not bool(error)
+        }
+
+        # Create unique folder name with timestamp
+        timestamp_str = datetime.now().strftime("%Y%m%d_%H%M%S")
+        random_id = str(uuid.uuid4())[:8]
+        folder_name = f"feedback_{timestamp_str}_{random_id}"
+        
+        # Create markdown feedback file
+        markdown_content = f"""# VayuChat Feedback Report
+
+## Session Information
+- **Timestamp**: {feedback_data['timestamp']}
+- **Session ID**: {feedback_data['session_id']}
+
+## User Interaction
+**Prompt**: {feedback_data['user_prompt']}
+
+## AI Response
+**Output**: {feedback_data['ai_output']}
+
+## Generated Code
+```python
+{feedback_data['generated_code']}
+```
+
+## Technical Details
+- **Error Message**: {feedback_data['error_message']}
+- **Is Image Output**: {feedback_data['is_image_output']}
+- **Success**: {feedback_data['success']}
+
+## User Feedback
+- **Score**: {feedback_data['feedback_score']}
+- **Comments**: {feedback_data['feedback_comment']}
+"""
+
+        # Save markdown file locally
+        markdown_filename = f"{folder_name}.md"
+        markdown_local_path = f"/tmp/{markdown_filename}"
+        
+        with open(markdown_local_path, "w", encoding="utf-8") as f:
+            f.write(markdown_content)
+
+        # Upload to Hugging Face
+        api = HfApi(token=hf_token)
+        
+        # Upload markdown feedback
+        api.upload_file(
+            path_or_fileobj=markdown_local_path,
+            path_in_repo=f"data/{markdown_filename}",
+            repo_id="SustainabilityLabIITGN/VayuChat_Feedback",
+            repo_type="dataset",
+        )
+        
+        # Upload image if it exists and is an image output
+        if status.get("is_image", False) and isinstance(output, str) and os.path.exists(output):
+            try:
+                image_filename = f"{folder_name}_plot.png"
+                api.upload_file(
+                    path_or_fileobj=output,
+                    path_in_repo=f"data/{image_filename}",
+                    repo_id="SustainabilityLabIITGN/VayuChat_Feedback",
+                    repo_type="dataset",
+                )
+            except Exception as img_error:
+                print(f"Error uploading image: {img_error}")
+        
+        # Clean up local files
+        if os.path.exists(markdown_local_path):
+            os.remove(markdown_local_path)
+        
+        st.success("Feedback uploaded successfully!")
+        return True
+        
+    except Exception as e:
+        st.error(f"Error uploading feedback: {e}")
+        print(f"Feedback upload error: {e}")
+        return False
+
+# Filter available models
+available_models = []
+model_names = list(models.keys())
+groq_models = []
+gemini_models = []
+for model_name in model_names:
+    if "gemini" not in model_name:
+        groq_models.append(model_name)
+    else:
+        gemini_models.append(model_name)
+if Groq_Token and Groq_Token.strip():
+    available_models.extend(groq_models)
+if gemini_token and gemini_token.strip():
+    available_models.extend(gemini_models)
+
+if not available_models:
+    st.error("No API keys available! Please set up your API keys in the .env file")
+    st.stop()
+
+# Set DeepSeek-R1 as default if available
+default_index = 0
+if "deepseek-R1" in available_models:
+    default_index = available_models.index("deepseek-R1")
 
-/* Quick prompt buttons */
-.quick-prompt-container {
-    display: flex;
-    flex-wrap: wrap;
-    gap: 8px;
-    margin-bottom: 1.5rem;
-    padding: 1rem;
-    background-color: #f8f9fa;
-    border-radius: 10px;
-    border: 1px solid #dee2e6;
-}
+# Simple header - just title and model selector
+col1, col2 = st.columns([3, 1])
+with col1:
+    st.title("VayuChat")
+with col2:
+    model_name = st.selectbox(
+        "Model:",
+        available_models,
+        index=default_index,
+        help="Choose your AI model"
+    )
 
-.quick-prompt-btn {
-    background-color: #0d6efd;
-    color: white;
-    border: none;
-    padding: 8px 16px;
-    border-radius: 20px;
-    font-size: 0.9rem;
-    cursor: pointer;
-    transition: all 0.2s ease;
-    white-space: nowrap;
-}
 
-.quick-prompt-btn:hover {
-    background-color: #0b5ed7;
-    transform: translateY(-2px);
-}
+# Load data with caching for better performance
+@st.cache_data
+def load_data():
+    return preprocess_and_load_df(join(self_path, "Data.csv"))
 
-/* User message styling */
-.user-message {
-    background: #3b82f6;
-    color: white;
-    padding: 0.75rem 1rem;
-    border-radius: 7px;
-    max-width: 95%;
-}
+try:
+    df = load_data()
+    # Data loaded silently - no success message needed
+except Exception as e:
+    st.error(f"Error loading data: {e}")
+    st.stop()
 
-.user-info {
-    font-size: 0.875rem;
-    opacity: 0.9;
-    margin-bottom: 3px;
-}
+inference_server = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
+image_path = "IITGN_Logo.png"
 
-/* Assistant message styling */
-.assistant-message {
-    background: #f1f5f9;
-    color: #334155;
-    padding: 0.75rem 1rem;
-    border-radius: 12px;
-    max-width: 85%;
-}
+# Clean sidebar  
+with st.sidebar:
+    # Quick Queries Section - moved to top
+    st.markdown("### Quick Queries")
+    
+    # Load quick prompts with caching
+    @st.cache_data
+    def load_questions():
+        questions = []
+        questions_file = join(self_path, "questions.txt")
+        if os.path.exists(questions_file):
+            try:
+                with open(questions_file, 'r', encoding='utf-8') as f:
+                    content = f.read()
+                    questions = [q.strip() for q in content.split("\n") if q.strip()]
+            except Exception as e:
+                questions = []
+        return questions
+    
+    questions = load_questions()
+    
+    # Add default prompts if file doesn't exist or is empty
+    if not questions:
+        questions = [
+            "Which month had highest pollution?",
+            "Which city has worst air quality?", 
+            "Show annual PM2.5 average",
+            "Plot monthly average PM2.5 for 2023",
+            "List all cities by pollution level",
+            "Compare winter vs summer pollution",
+            "Show seasonal pollution patterns",
+            "Which areas exceed WHO guidelines?",
+            "What are peak pollution hours?",
+            "Show PM10 vs PM2.5 comparison",
+            "Which station records highest variability in PM2.5?",
+            "Calculate pollution improvement rate year-over-year by city",
+            "Identify cities with PM2.5 levels consistently above 50 μg/m³ for >6 months",
+            "Find correlation between PM2.5 and PM10 across different seasons and cities",
+            "Compare weekday vs weekend levels",
+            "Plot yearly trend analysis",
+            "Show pollution distribution by city",
+            "Create correlation plot between pollutants"
+        ]
+    
+    # Quick query buttons in sidebar
+    selected_prompt = None
+    
+    
+    for i, question in enumerate(questions[:20]):  # Show more questions including policy-focused ones
+        # Simple left-aligned buttons without icons for cleaner look
+        if st.button(question, key=f"sidebar_prompt_{i}", use_container_width=True, help=f"Click to analyze: {question}"):
+            if question != st.session_state.get("last_selected_prompt"):
+                selected_prompt = question
+                st.session_state.last_selected_prompt = question
+    
+    st.markdown("---")
+    
+    
+    # Clear Chat Button
+    if st.button("Clear Chat", use_container_width=True):
+        st.session_state.responses = []
+        st.session_state.processing = False
+        st.session_state.session_id = str(uuid.uuid4())
+        try:
+            st.rerun()
+        except AttributeError:
+            st.experimental_rerun()
 
-.assistant-info {
-    font-size: 0.875rem;
-    color: #6b7280;
-    margin-bottom: 5px;
-}
+# Initialize session state first
+if "responses" not in st.session_state:
+    st.session_state.responses = []
+if "processing" not in st.session_state:
+    st.session_state.processing = False
+if "session_id" not in st.session_state:
+    st.session_state.session_id = str(uuid.uuid4())
 
-/* Processing indicator */
-.processing-indicator {
-    background: linear-gradient(135deg, #a8edea 0%, #fed6e3 100%);
-    color: #333;
-    padding: 1rem 1.5rem;
-    border-radius: 12px;
-    margin: 1rem 0;
-    margin-left: 0;
-    margin-right: auto;
-    max-width: 70%;
-    box-shadow: 0 2px 10px rgba(0,0,0,0.1);
-    animation: pulse 2s infinite;
-}
 
-@keyframes pulse {
-    0% { opacity: 1; }
-    50% { opacity: 0.7; }
-    100% { opacity: 1; }
-}
 
-/* Feedback box */
-.feedback-section {
-    background-color: #f8f9fa;
-    border: 1px solid #dee2e6;
-    padding: 1rem;
-    border-radius: 8px;
-    margin: 1rem 0;
-}
 
-/* Success and error messages */
-.success-message {
-    background-color: #d1e7dd;
-    color: #0f5132;
-    padding: 1rem;
-    border-radius: 6px;
-    border: 1px solid #badbcc;
-}
+def show_custom_response(response):
+    """Custom response display function with improved styling"""
+    role = response.get("role", "assistant")
+    content = response.get("content", "")
+    
+    if role == "user":
+        # User message with right alignment - reduced margins
+        st.markdown(f"""
+        <div style='display: flex; justify-content: flex-end; margin: 1rem 0;'>
+            <div class='user-message'>
+                {content}
+            </div>
+        </div>
+        """, unsafe_allow_html=True)
+    elif role == "assistant":
+        # Check if content is an image filename - don't display the filename text
+        is_image_path = isinstance(content, str) and any(ext in content for ext in ['.png', '.jpg', '.jpeg'])
+        
+        # Check if content is a pandas DataFrame
+        import pandas as pd
+        is_dataframe = isinstance(content, pd.DataFrame)
+        
+        # Check for errors first and display them with special styling
+        error = response.get("error")
+        timestamp = response.get("timestamp", "")
+        timestamp_display = f" • {timestamp}" if timestamp else ""
+        
+        if error:
+            st.markdown(f"""
+            <div style='display: flex; justify-content: flex-start; margin: 1rem 0;'>
+                <div class='assistant-message'>
+                    <div class='assistant-info'>VayuChat{timestamp_display}</div>
+                    <div class='error-message'>
+                        ⚠️ <strong>Error:</strong> {error}
+                        <br><br>
+                        <em>💡 Try rephrasing your question or being more specific about what you'd like to analyze.</em>
+                    </div>
+                </div>
+            </div>
+            """, unsafe_allow_html=True)
+        # Assistant message with left alignment - reduced margins
+        elif not is_image_path and not is_dataframe:
+            st.markdown(f"""
+            <div style='display: flex; justify-content: flex-start; margin: 1rem 0;'>
+                <div class='assistant-message'>
+                    <div class='assistant-info'>VayuChat{timestamp_display}</div>
+                    {content if isinstance(content, str) else str(content)}
+                </div>
+            </div>
+            """, unsafe_allow_html=True)
+        elif is_dataframe:
+            # Display DataFrame with nice formatting
+            st.markdown(f"""
+            <div style='display: flex; justify-content: flex-start; margin: 1rem 0;'>
+                <div class='assistant-message'>
+                    <div class='assistant-info'>VayuChat{timestamp_display}</div>
+                    Here are the results:
+                </div>
+            </div>
+            """, unsafe_allow_html=True)
+            
+            # Add context info for dataframes
+            st.markdown("""
+            <div class='context-info'>
+                💡 This table is interactive - click column headers to sort, or scroll to view all data.
+            </div>
+            """, unsafe_allow_html=True)
+            
+            st.dataframe(content, use_container_width=True)
+        
+        # Show generated code with Streamlit expander
+        if response.get("gen_code"):
+            with st.expander("📋 View Generated Code", expanded=False):
+                st.code(response["gen_code"], language="python")
+        
+        # Try to display image if content is a file path
+        try:
+            if isinstance(content, str) and (content.endswith('.png') or content.endswith('.jpg')):
+                if os.path.exists(content):
+                    # Display image without showing filename
+                    st.image(content, use_column_width=True)
+                    return {"is_image": True}
+            # Also handle case where content shows filename but we want to show image
+            elif isinstance(content, str) and any(ext in content for ext in ['.png', '.jpg']):
+                # Extract potential filename from content
+                import re
+                filename_match = re.search(r'([^/\\]+\.(?:png|jpg|jpeg))', content)
+                if filename_match:
+                    filename = filename_match.group(1)
+                    if os.path.exists(filename):
+                        st.image(filename, use_column_width=True)
+                        return {"is_image": True}
+        except:
+            pass
+            
+        return {"is_image": False}
 
-.error-message {
-    background-color: #f8d7da;
-    color: #842029;
-    padding: 1rem;
-    border-radius: 6px;
-    border: 1px solid #f5c2c7;
-}
 
-/* Chat input styling - Fixed alignment */
-# .stChatInput {
-#     border-radius: 12px !important;
-#     border: 2px solid #e5e7eb !important;
-#     background: #ffffff !important;
-#     padding: 0.75rem 1rem !important;
-#     font-size: 1rem !important;
-#     width: 100% !important;
-#     max-width: 70% !important;
-#     margin: 0 !important;
-#     box-shadow: 0 1px 3px rgba(0, 0, 0, 0.1) !important;
-#     transition: all 0.2s ease !important;
-# }
-
-# .stChatInput:focus {
-#     border-color: #3b82f6 !important;
-#     box-shadow: 0 0 0 3px rgba(59, 130, 246, 0.1) !important;
-#     outline: none !important;
-# }
-
-/* Chat input container */
-.stChatInput > div {
-    padding: 0 !important;
-    margin: 0 !important;
-}
+# Chat history
+# Display chat history
+for response_id, response in enumerate(st.session_state.responses):
+    status = show_custom_response(response)
+    
+    # Show feedback section for assistant responses
+    if response["role"] == "assistant":
+        feedback_key = f"feedback_{int(response_id/2)}"
+        error = response.get("error", "")
+        output = response.get("content", "")
+        last_prompt = response.get("last_prompt", "")
+        code = response.get("gen_code", "")
 
-/* Chat input text area */
-# .stChatInput textarea {
-#     border: none !important;
-#     background: transparent !important;
-#     padding: 0 !important;
-#     margin: 0 !important;
-#     font-size: 1rem !important;
-#     line-height: 1.5 !important;
-#     resize: none !important;
-#     outline: none !important;
-# }
-
-/* Chat input placeholder */
-# .stChatInput textarea::placeholder {
-#     color: #9ca3af !important;
-#     font-style: normal !important;
-# }
-
-.st-emotion-cache-f4ro0r {
-    align-items = center;
-}
 
-/* Fix the main chat input container alignment */
-[data-testid="stChatInput"] {
-    position: fixed !important;
-    bottom: 0.5rem !important;
-    left: 6rem !important;
-    right: 0 !important;
-    background: #ffffff !important;
-    width: 65% !important;  
-    box-shadow: 0 -2px 10px rgba(0, 0, 0, 0.1) !important;
-}
+        if "feedback" in st.session_state.responses[response_id]:
+            feedback_data = st.session_state.responses[response_id]["feedback"]
+            st.markdown(f"""
+            <div class='feedback-section'>
+                <strong>Your Feedback:</strong> {feedback_data.get('score', '')} 
+                {f"- {feedback_data.get('text', '')}" if feedback_data.get('text') else ""}
+            </div>
+            """, unsafe_allow_html=True)
+        else:
+            # Simple feedback
+            st.markdown("**Rate this response:**")
+            col1, col2 = st.columns(2)
+            with col1:
+                good = st.button("👍 Good", key=f"{feedback_key}_good")
+            with col2:
+                poor = st.button("👎 Needs work", key=f"{feedback_key}_poor")
+                
+            if good or poor:
+                if good:
+                    thumbs = "👍 Good"
+                else:
+                    thumbs = "👎 Needs work"
+                comments = st.text_input("Optional comment:", key=f"{feedback_key}_comments")
+                
+                feedback = {"score": thumbs, "text": comments}
+                st.session_state.responses[response_id]["feedback"] = feedback
+                st.success("Thanks for your feedback!")
+                st.rerun()
 
-/* Adjust main content to account for fixed chat input */
-.main .block-container {
-    padding-bottom: 100px !important;
-}
+# Chat input with better guidance
+prompt = st.chat_input("💬 Ask about air quality trends, compare cities, or request visualizations...", key="main_chat")
 
-/* Chat input button styling */
-[data-testid="stChatInput"] button {
-    background: #3b82f6 !important;
-    color: white !important;
-    border: none !important;
-    border-radius: 12px !important;
-    font-weight: 600 !important;
-    transition: background-color 0.2s ease !important;
-}
+# Handle selected prompt from quick prompts
+if selected_prompt:
+    prompt = selected_prompt
 
-[data-testid="stChatInput"] button:hover {
-    background: #2563eb !important;
-}
+# Handle follow-up prompts from quick action buttons
+if st.session_state.get("follow_up_prompt") and not st.session_state.get("processing"):
+    prompt = st.session_state.follow_up_prompt
+    st.session_state.follow_up_prompt = None  # Clear the follow-up prompt
 
-/* Textarea inside chat input */
-[data-testid="stChatInput"] [data-baseweb="textarea"] {
-    border: 2px solid #3b82f6 !important;
-    border-radius: 12px !important;
-    font-size: 16px !important;
-    color: #111 !important;
+# Handle new queries
+if prompt and not st.session_state.get("processing"):
+    # Prevent duplicate processing
+    if "last_prompt" in st.session_state:
+        last_prompt = st.session_state["last_prompt"]
+        last_model_name = st.session_state.get("last_model_name", "")
+        if (prompt == last_prompt) and (model_name == last_model_name):
+            prompt = None
 
-    width: 100% !important;   /* fill the parent container */
-    box-sizing: border-box !important;
-}
+    if prompt:
+        # Add user input to chat history
+        user_response = get_from_user(prompt)
+        st.session_state.responses.append(user_response)
+        
+        # Set processing state
+        st.session_state.processing = True
+        st.session_state.current_model = model_name
+        st.session_state.current_question = prompt
+        
+        # Rerun to show processing indicator
+        st.rerun()
 
-/* Ensure proper spacing from sidebar */
-@media (min-width: 768px) {
-    [data-testid="stChatInput"] {
-        margin-left: 21rem !important; /* Account for sidebar width */
+# Process the question if we're in processing state
+if st.session_state.get("processing"):
+    # Enhanced processing indicator like Claude Code
+    st.markdown("""
+    <div style='padding: 1rem; text-align: center; background: #f8fafc; border-radius: 8px; margin: 1rem 0;'>
+        <div style='display: flex; align-items: center; justify-content: center; gap: 0.5rem; color: #475569;'>
+            <div style='font-weight: 500;'>🤖 Processing with """ + str(st.session_state.get('current_model', 'Unknown')) + """</div>
+            <div class='dots' style='display: inline-flex; gap: 2px;'>
+                <div class='dot' style='width: 4px; height: 4px; background: #3b82f6; border-radius: 50%; animation: bounce 1.4s infinite ease-in-out;'></div>
+                <div class='dot' style='width: 4px; height: 4px; background: #3b82f6; border-radius: 50%; animation: bounce 1.4s infinite ease-in-out; animation-delay: 0.16s;'></div>
+                <div class='dot' style='width: 4px; height: 4px; background: #3b82f6; border-radius: 50%; animation: bounce 1.4s infinite ease-in-out; animation-delay: 0.32s;'></div>
+            </div>
+        </div>
+        <div style='font-size: 0.75rem; color: #6b7280; margin-top: 0.25rem;'>Analyzing data and generating response...</div>
+    </div>
+    <style>
+    @keyframes bounce {
+        0%, 80%, 100% { transform: scale(0.8); opacity: 0.5; }
+        40% { transform: scale(1.2); opacity: 1; }
     }
-}
-
-/* Code container styling */
-.code-container {
-    margin: 1rem 0;
-    border: 1px solid #d1d5db;
-    border-radius: 12px;
-    background: white;
-    box-shadow: 0 1px 3px rgba(0, 0, 0, 0.1);
-}
-
-.code-header {
-    display: flex;
-    justify-content: space-between;
-    align-items: center;
-    padding: 0.875rem 1.25rem;
-    background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%);
-    border-bottom: 1px solid #e2e8f0;
-    cursor: pointer;
-    transition: all 0.2s ease;
-    border-radius: 12px 12px 0 0;
-}
-
-.code-header:hover {
-    background: linear-gradient(135deg, #e2e8f0 0%, #cbd5e1 100%);
-}
-
-.code-title {
-    font-size: 0.9rem;
-    font-weight: 600;
-    color: #1e293b;
-    display: flex;
-    align-items: center;
-    gap: 0.5rem;
-}
-
-.code-title:before {
-    content: "⚡";
-    font-size: 0.8rem;
-}
-
-.toggle-text {
-    font-size: 0.75rem;
-    color: #64748b;
-    font-weight: 500;
-}
-
-.code-block {
-    background: linear-gradient(135deg, #0f172a 0%, #1e293b 100%);
-    color: #e2e8f0;
-    padding: 1.5rem;
-    font-family: 'SF Mono', 'Monaco', 'Menlo', 'Consolas', monospace;
-    font-size: 0.875rem;
-    overflow-x: auto;
-    line-height: 1.6;
-    border-radius: 0 0 12px 12px;
-}
+    </style>
+    """, unsafe_allow_html=True)
+    
+    prompt = st.session_state.get("current_question")
+    model_name = st.session_state.get("current_model")
+    
+    try:
+        response = ask_question(model_name=model_name, question=prompt)
+        
+        if not isinstance(response, dict):
+            response = {
+                "role": "assistant",
+                "content": "Error: Invalid response format",
+                "gen_code": "",
+                "ex_code": "",
+                "last_prompt": prompt,
+                "error": "Invalid response format",
+                "timestamp": datetime.now().strftime("%H:%M")
+            }
+        
+        response.setdefault("role", "assistant")
+        response.setdefault("content", "No content generated")
+        response.setdefault("gen_code", "")
+        response.setdefault("ex_code", "")
+        response.setdefault("last_prompt", prompt)
+        response.setdefault("error", None)
+        response.setdefault("timestamp", datetime.now().strftime("%H:%M"))
+        
+    except Exception as e:
+        response = {
+            "role": "assistant",
+            "content": f"Sorry, I encountered an error: {str(e)}",
+            "gen_code": "",
+            "ex_code": "",
+            "last_prompt": prompt,
+            "error": str(e),
+            "timestamp": datetime.now().strftime("%H:%M")
+        }
 
-.answer-container {
-    background: #f8fafc;
-    border: 1px solid #e2e8f0;
-    border-radius: 8px;
-    padding: 1.5rem;
-    margin: 1rem 0;
-}
+    st.session_state.responses.append(response)
+    st.session_state["last_prompt"] = prompt
+    st.session_state["last_model_name"] = model_name
+    st.session_state.processing = False
+    
+    # Clear processing state
+    if "current_model" in st.session_state:
+        del st.session_state.current_model
+    if "current_question" in st.session_state:
+        del st.session_state.current_question
+    
+    st.rerun()
 
-.answer-text {
-    font-size: 1.125rem;
-    color: #1e293b;
-    line-height: 1.6;
-    margin-bottom: 1rem;
-}
+# Close chat container
+st.markdown("</div>", unsafe_allow_html=True)
 
-.answer-highlight {
-    background: #fef3c7;
-    padding: 0.125rem 0.375rem;
-    border-radius: 4px;
-    font-weight: 600;
-    color: #92400e;
-}
+# Minimal auto-scroll - only scroll when processing
+if st.session_state.get("processing"):
+    st.markdown("<script>scrollToBottom();</script>", unsafe_allow_html=True)
 
-.context-info {
-    background: #f1f5f9;
-    border-left: 4px solid #3b82f6;
-    padding: 0.75rem 1rem;
-    margin: 1rem 0;
-    font-size: 0.875rem;
-    color: #475569;
-}
+# Beautiful sidebar footer
+# with st.sidebar:
+#     st.markdown("---")
+#     st.markdown("""
+#     <div class='contact-section'>
+#         <h4>📄 Paper on VayuChat</h4>
+#         <p>Learn more about VayuChat in our <a href='https://arxiv.org/abs/2411.12760' target='_blank'>Research Paper</a>.</p>
+#     </div>
+#     """, unsafe_allow_html=True)
+    
+    # Dataset Info Section (matching mockup)
+    st.markdown("### Dataset Info")
+    st.markdown("""
+    <div style='background: #f1f5f9; border-radius: 8px; padding: 1rem; margin-bottom: 1rem;'>
+        <h4 style='margin: 0 0 0.5rem 0; color: #1e293b; font-size: 0.9rem;'>PM2.5 Air Quality Data</h4>
+        <p style='margin: 0; font-size: 0.75rem; color: #475569;'><strong>Time Range:</strong> 2022 - 2023</p>
+        <p style='margin: 0; font-size: 0.75rem; color: #475569;'><strong>Locations:</strong> 300+ cities across India</p>
+        <p style='margin: 0; font-size: 0.75rem; color: #475569;'><strong>Records:</strong> 100,000+ measurements</p>
+    </div>
+    """, unsafe_allow_html=True)
 
-/* Hide default menu and footer */
-#MainMenu {visibility: hidden;}
-footer {visibility: hidden;}
-header {visibility: hidden;}
 
-/* Auto scroll */
-.main-container {
-    height: 70vh;
-    overflow-y: auto;
-}
-</style>
-""", unsafe_allow_html=True)

ncap_funding_data.csv

@@ -1,118 +0,0 @@
-S. No.,state,city,Amount released during FY 2019-20,Amount released during FY 2020-21,Amount released during FY 2021-22,Total fund released,Utilisation as on June 2022
-1,Andhra Pradesh,Vijaywada,6.0,,,6.0,22.91
-2,Andhra Pradesh,Guntur,0.12,0.76,1.96,2.84,22.91
-3,Andhra Pradesh,Kurnool,0.06,0.76,1.36,2.18,22.91
-4,Andhra Pradesh,Nellore,0.06,0.76,1.92,2.74,22.91
-5,Andhra Pradesh,Visakhapatnam,0.12,,,0.12,22.91
-6,Andhra Pradesh,Srikakulam,,2.0,0.49,2.49,22.91
-7,Andhra Pradesh,Chitoor,,2.0,0.46,2.46,22.91
-8,Andhra Pradesh,Ongole,,2.0,0.64,2.64,22.91
-9,Andhra Pradesh,vizianagaram,,2.0,0.83,2.83,22.91
-10,Andhra Pradesh,Eluru,,2.0,0.82,2.82,22.91
-11,Andhra Pradesh,Rajahmundry,,2.0,1.13,3.13,22.91
-12,Andhra Pradesh,Anantapur,,2.0,1.04,3.04,22.91
-13,Andhra Pradesh,Kadapa,,1.0,0.83,1.83,22.91
-14,Assam,Guwahati,0.12,5.0,,5.12,1.45
-15,Assam,Nagaon,0.06,2.0,,2.06,1.45
-16,Assam,Nalbari,0.06,1.0,,1.06,1.45
-17,Assam,Sibsagar,0.06,2.0,,2.06,1.45
-18,Assam,Silchar,0.06,2.0,,2.06,1.45
-19,Bihar,Patna,10.0,,,10.0,15.2
-20,Bihar,Gaya,0.1,2.0,1.9,4.0,15.2
-21,Bihar,Muzaffarpur,0.1,5.0,2.5,7.6,15.2
-22,Chandigarh,Chandigarh,8.28,5.0,4.61,17.89,10.83
-23,Chhattisgarh,Raipur,6.0,,,6.0,2.76
-24,Chhattisgarh,Durg Bhilainagar,6.0,,,6.0,2.76
-25,Chhattisgarh,Korba,0.06,1.0,,1.06,2.76
-26,Delhi,Delhi,,,11.25,11.25,
-27,Gujarat,Surat,6.0,,,6.0,12.0
-28,Gujarat,Ahmedabad,6.0,,,6.0,12.0
-29,Himachal Pradesh,Baddi (Baddi&nalagarh considered twin during FY 20-21),0.06,3.0,0.2,3.26,6.35
-30,Himachal Pradesh,Nalagarh,0.06,,0.06,0.12,6.35
-31,Himachal Pradesh,Paonta Sahib,0.06,1.0,0.1,1.16,6.35
-32,Himachal Pradesh,Sunder Nagar,0.06,1.0,0.08,1.14,6.35
-33,Himachal Pradesh,Kala Amb,,3.0,0.0,3.0,6.35
-34,Himachal Pradesh,Damtal,,1.0,0.01,1.01,6.35
-35,Himachal Pradesh,Parwanoo,,1.0,0.03,1.03,6.35
-36,Jammu & Kashmir,Jammu,0.12,3.0,4.89,8.01,0.12
-37,Jammu & Kashmir,Srinagar,,5.0,7.95,12.95,0.12
-38,Jharkhand,Dhanbad,6.0,,,6.0,3.0
-39,Karnataka,Bangalore,6.0,,,6.0,7.39
-40,Karnataka,Gulburga,0.12,0.38,2.24,2.74,7.39
-41,Karnataka,Hubli-Dharwad,0.12,0.38,3.68,4.18,7.39
-42,Karnataka,Devangere,0.06,0.76,1.4,2.22,7.39
-43,Madhya Pradesh,Bhopal,10.0,,,10.0,20.96
-44,Madhya Pradesh,Gwalior,10.0,,,10.0,20.96
-45,Madhya Pradesh,Indore,0.2,,,0.2,20.96
-46,Madhya Pradesh,Ujjain,0.2,0.38,2.33,2.91,20.96
-47,Madhya Pradesh,Sagar,0.1,0.76,1.36,2.22,20.96
-48,Madhya Pradesh,Dewas,0.1,0.38,1.33,1.81,20.96
-49,Maharashtra,Mumbai,9.5,,,9.5,29.92
-50,Maharashtra,Nagpur,9.45,,,9.45,29.92
-51,Maharashtra,Navi Mumbai,9.45,,,9.45,29.92
-52,Maharashtra,Pune,9.45,,,9.45,29.92
-53,Maharashtra,Amravati,0.2,1.14,2.91,4.25,29.92
-54,Maharashtra,Aurangabad,0.2,,,0.2,29.92
-55,Maharashtra,Nashik,0.2,,,0.2,29.92
-56,Maharashtra,Kolhapur,0.2,0.76,,0.96,29.92
-57,Maharashtra,Sangli,0.2,0.76,1.72,2.68,29.92
-58,Maharashtra,Solapur,0.2,0.38,4.2,4.78,29.92
-59,Maharashtra,Ulhasnagar,0.2,1.9,,2.1,29.92
-60,Maharashtra,Akola,0.1,1.14,1.47,2.71,29.92
-61,Maharashtra,Badlapur,0.1,1.9,,2.0,29.92
-62,Maharashtra,Chandrapur,0.1,1.14,,1.24,29.92
-63,Maharashtra,Jalgaon,0.1,0.76,,0.86,29.92
-64,Maharashtra,Jalna,0.1,1.14,,1.24,29.92
-65,Maharashtra,Latur,0.1,0.38,1.6,2.08,29.92
-66,Meghalaya,Byrnihat,,3.0,0.0,3.0,1.97
-67,Nagaland,Dimapur,0.06,3.0,0.53,3.59,6.12
-68,Nagaland,Kohima,0.06,3.0,0.4,3.46,6.12
-69,Odisha,Twin City Bhubaneshwar & Cuttack,6.0,,,6.0,3.62
-70,Odisha,Balasore,0.06,0.76,,0.82,3.62
-71,Odisha,Rourkela,0.06,1.14,,1.2,3.62
-72,Odisha,Angul,0.06,1.14,,1.2,3.62
-73,Odisha,Kalinga Nagar,,3.0,,3.0,3.62
-74,Odisha,Talcher,,,0.22,0.22,3.62
-75,Odisha,Cuttack,,,3.42,3.42,3.62
-76,Punjab,Ludhiana,6.0,,,6.0,3.02
-77,Punjab,Amritsar,6.0,,,6.0,3.02
-78,Punjab,Jalandhar,0.12,4.0,,4.12,3.02
-79,Punjab,Khanna,0.06,1.9,,1.96,3.02
-80,Punjab,Gobindgarh,0.06,3.0,,3.06,3.02
-81,Punjab,NayaNangal,0.06,1.0,,1.06,3.02
-82,Punjab,Dera Baba Nanak,0.06,0.76,,0.82,3.02
-83,Punjab,Patiala,0.06,4.0,,4.06,3.02
-84,Punjab,DeraBassi,0.06,0.38,,0.44,3.02
-85,Rajasthan,Jaipur,6.0,,,6.0,12.55
-86,Rajasthan,Jodhpur,6.0,,,6.0,12.55
-87,Rajasthan,Kota,6.0,,,6.0,12.55
-88,Rajasthan,Alwar,0.06,1.9,,1.96,12.55
-89,Rajasthan,Udaipur,0.06,1.9,,1.96,12.55
-90,Tamil Nadu,Tuticorin,0.06,3.0,,3.06,
-91,Telangana,Hyderabad,10.8,,,10.8,9.72
-92,Telangana,Nalgonda,0.1,0.38,0.47,0.95,9.72
-93,Telangana,Patencheru,0.1,0.38,,0.48,9.72
-94,Telangana,Sangareddy,,2.0,0.32,2.32,9.72
-95,Uttar Pradesh,Agra,9.45,,,9.45,30.57
-96,Uttar Pradesh,Allahabad,9.45,,,9.45,30.57
-97,Uttar Pradesh,Kanpur,9.45,,,9.45,30.57
-98,Uttar Pradesh,Lucknow,9.45,,,9.45,30.57
-99,Uttar Pradesh,Varanasi,9.47,,,9.47,30.57
-100,Uttar Pradesh,Moradabad,0.2,1.9,,2.1,30.57
-101,Uttar Pradesh,Bareily,0.2,1.9,,2.1,30.57
-102,Uttar Pradesh,Firozabad,0.2,1.9,,2.1,30.57
-103,Uttar Pradesh,Jhansi,0.2,1.14,,1.34,30.57
-104,Uttar Pradesh,Khurja,0.1,1.9,,2.0,30.57
-105,Uttar Pradesh,Anpara,0.1,1.14,,1.24,30.57
-106,Uttar Pradesh,Gajraula,0.1,1.14,,1.24,30.57
-107,Uttar Pradesh,Raebareli,0.1,1.14,,1.24,30.57
-108,Uttar Pradesh,Gorakhpur,,,9.64,9.64,30.57
-109,Uttar Pradesh,Noida,,,6.67,6.67,30.57
-110,Uttarakhand,Kashipur,0.06,3.0,0.79,3.85,8.15
-111,Uttarakhand,Rishikesh,0.06,5.0,,5.06,8.15
-112,Uttarakhand,Dehradun,,3.0,4.88,7.88,8.15
-113,West Bengal,Kolkata,6.0,,,6.0,19.0
-114,West Bengal,Howrah,,5.0,,5.0,19.0
-115,West Bengal,Haldia,,3.0,,3.0,19.0
-116,West Bengal,Durgapur,,3.0,,3.0,19.0
-117,West Bengal,Barrackpore,,2.0,,2.0,19.0

new_system_prompt.txt

@@ -1,65 +0,0 @@
-Generate Python code to answer the user's question about air quality data.
-
-SCOPE VALIDATION (MANDATORY FIRST STEP):
-- ONLY answer questions about: air quality, pollution (PM2.5, PM10, NO2, ozone, etc.), meteorology (wind, temperature, humidity), NCAP funding, Indian cities/states environmental data
-- If question is NOT about air quality/pollution/environmental data, generate ONLY this code:
-  answer = "I can only help with air quality and pollution data analysis. Please ask about PM2.5, pollution trends, city comparisons, meteorological factors, or NCAP funding."
-- Examples of REJECTED topics: general Python coding, politics, personal questions, unrelated data analysis
-- For rejected questions: write only the answer assignment - no other code needed
-
-CRITICAL: Only generate Python code - no explanations, no thinking, just clean executable code.
-
-OUTPUT TYPES (store result in 'answer' variable):
-1. PLOTS: For visualization questions → save plot and store filename: answer = filename
-2. TEXT: For simple questions → store direct string: answer = "The highest PM2.5 city is Delhi"  
-3. DATAFRAMES: For rankings/lists → store DataFrame: answer = result_df
-
-AVAILABLE LIBRARIES:
-- pandas, numpy (data manipulation)
-- matplotlib, seaborn, plotly (visualization) 
-- statsmodels, scikit-learn (analysis)
-- geopandas (geospatial analysis)
-
-IMPORT REQUIREMENTS:
-- Always import what you use: import seaborn as sns, import numpy as np
-- Standard imports are already available: pandas as pd, matplotlib.pyplot as plt
-
-ESSENTIAL RULES:
-
-DATA SAFETY:
-- Always check if data exists: if df.empty: answer = "No data available"
-- For city-specific questions: filter first: df_city = df[df['City'].str.contains('CityName', case=False)]
-- Check sufficient data: if len(df_filtered) < 10: answer = "Insufficient data"
-- Use .dropna() to remove missing values before analysis
-
-PLOTTING REQUIREMENTS:
-- Create plots for visualization requests: fig, ax = plt.subplots(figsize=(9, 6))
-- Save plots with ULTRA high resolution: filename = f"plot_{uuid.uuid4().hex[:8]}.png"; plt.savefig(filename, dpi=1200, bbox_inches='tight', facecolor='white', edgecolor='none')
-- Close plots: plt.close()
-- Store filename: answer = filename
-- For non-plots: answer = "text result"
-
-BASIC ERROR PREVENTION:
-- Use try/except for complex operations
-- Validate results: if pd.isna(result): answer = "Analysis inconclusive"
-- For correlations: check len(data) > 20 before calculating
-- Use simple matplotlib plotting - avoid complex visualizations
-
-PLOTTING BEST PRACTICES:
-- Check data exists in each category before plotting
-- For comparisons (>, <): ensure both categories have data
-- Example: high_wind = df[df['WS'] > 3]; low_wind = df[df['WS'] <= 3]
-- If category is empty: create simple bar chart instead of box plots
-- Add data count labels: plt.text() to show sample sizes
-
-TECHNICAL REQUIREMENTS:
-- Save final result in variable called 'answer'
-- Use exact column names: 'PM2.5 (µg/m³)', 'WS (m/s)', etc.
-- Handle dates with pd.to_datetime() if needed
-- Round numerical results: round(value, 2)
-
-MANDATORY: ALWAYS END CODE WITH ANSWER ASSIGNMENT
-- Every code block MUST end with: answer = [result]
-- If analysis fails: answer = "Unable to complete analysis with available data"
-- If plotting fails: answer = "Unable to generate visualization"
-- NEVER leave answer variable unset - this will cause system failure

questions.txt

@@ -1,30 +1,28 @@
-Which city has the highest average PM2.5 levels in 2023?
-Show monthly PM2.5 trends for Delhi in 2023
-Compare PM2.5 levels between winter and summer months
-Which month had the highest pollution levels in Mumbai?
-Calculate the average PM2.5 for all cities in November 2023
-Rank top 10 cities by the highest PM2.5 pollution levels
-Show seasonal pollution patterns across all cities
-Compare all pollutant levels between weekdays and weekends
-Which cities exceed the WHO PM2.5 guidelines of 15 µg/m³?
-Plot yearly PM2.5 trends from 2020 to 2023 for major cities
-How much NCAP funding did Delhi receive vs Mumbai?
-Which NCAP cities achieved the best PM2.5 reduction?
-Does wind speed above 3 m/s reduce PM2.5 levels in Delhi?
-Show the correlation between temperature and PM2.5 in the summer months
-Which states in India with populations above 1 million have average PM2.5 concentrations exceeding 60 µg/m³?
-Compare PM2.5 levels in high-funded vs low-funded NCAP cities
-Does rainfall help reduce pollution levels during the monsoon?
-Which meteorological factor correlates most with PM2.5 reduction?
-Show monthly PM2.5 trends for the top 5 Indian states by population
-Does humidity above 80% help reduce pollution in coastal cities?
-Compare NO2 vs PM2.5 levels in traffic-heavy areas
-Which NCAP-funded cities still exceed WHO guidelines?
-Show relationship between city population and average PM2.5
-Compare PM2.5 improvement rates: Delhi vs Mumbai vs Kolkata
-Create simple scatter plot of PM2.5 vs PM10 correlation
-Show state-wise average PM2.5 levels for policy planning
-Which cities need immediate intervention with PM2.5 above 60 µg/m³?
-Compare pollution trends between North vs South Indian cities
-Show seasonal variation in PM2.5 across different climate zones
-Identify cities with consistent pollution improvement over time
+Which month in 2023 had the highest average PM2.5 pollution levels?
+Which city has the worst air quality based on average PM2.5 levels?
+Calculate the overall annual average PM2.5 concentration across all cities in 2023
+Plot monthly average PM2.5 trends for 2023 across all cities
+Rank all cities by average PM2.5 levels from highest to lowest pollution
+Compare average PM2.5 levels: winter months (Dec-Feb) vs summer months (Apr-Jun)
+Show seasonal PM2.5 patterns: which season has highest pollution levels?
+Which cities have annual average PM2.5 exceeding WHO guideline of 15 μg/m³?
+Identify the top 10 most polluted cities based on PM2.5 in 2023
+Compare PM10 vs PM2.5 correlation strength across different cities
+Which station in Ahmedabad shows highest PM2.5 variability in winter 2023?
+Calculate PM2.5 improvement rate from 2022 to 2023 for Mumbai vs Delhi
+Identify Gujarat cities with PM2.5 >50 μg/m³ for 6+ consecutive months in 2023
+Compare PM2.5 vs PM10 correlation: winter vs summer across top 5 polluted cities
+Compare average pollution levels: weekdays (Mon-Fri) vs weekends (Sat-Sun)
+Plot year-over-year PM2.5 trend from 2022 to 2023 for major cities
+Show PM2.5 distribution histogram across all cities and time periods
+Create scatter plot showing PM2.5 vs PM10 correlation with trend line
+Which cities need immediate emergency intervention with PM2.5 >100 μg/m³?
+Identify cities showing consistent pollution improvement for policy replication
+Which months require targeted interventions based on highest pollution spikes?
+Compare pollution levels between industrial vs non-industrial cities for policy focus
+Rank states by average air quality to prioritize national-level resource allocation
+Which cities have PM2.5 levels in 'hazardous' category (>250 μg/m³) requiring urgent action?
+Calculate population-weighted pollution exposure to identify areas affecting most people
+Show cities with worsening pollution trends that need immediate policy intervention
+Compare pollution reduction success: which cities improved most and how?
+Identify seasonal emergency periods when public health advisories should be issued

src.py

@@ -20,24 +20,19 @@ hf_token = os.getenv("HF_TOKEN")
 gemini_token = os.getenv("GEMINI_TOKEN")
 
 # Debug print (remove in production)
-# print(f"Debug - Groq Token: {'Present' if Groq_Token else 'Missing'}")
-# print(f"Debug - Groq Token Value: {Groq_Token[:10] + '...' if Groq_Token else 'None'}")
-# print(f"Debug - Gemini Token: {'Present' if gemini_token else 'Missing'}")
+print(f"Debug - Groq Token: {'Present' if Groq_Token else 'Missing'}")
+print(f"Debug - Groq Token Value: {Groq_Token[:10] + '...' if Groq_Token else 'None'}")
+print(f"Debug - Gemini Token: {'Present' if gemini_token else 'Missing'}")
 
 models = {
-    "gpt-oss-120b": "openai/gpt-oss-120b",
-     "qwen3-32b": "qwen/qwen3-32b",
     "gpt-oss-20b": "openai/gpt-oss-20b",
-    "llama4 maverik":"meta-llama/llama-4-maverick-17b-128e-instruct",
+    "gpt-oss-120b": "openai/gpt-oss-120b",
+    "llama3.1": "llama-3.1-8b-instant",
     "llama3.3": "llama-3.3-70b-versatile",
     "deepseek-R1": "deepseek-r1-distill-llama-70b",
-    "gemini-2.5-flash": "gemini-2.5-flash",
-    "gemini-2.5-pro": "gemini-2.5-pro",
-    "gemini-2.5-flash-lite": "gemini-2.5-flash-lite",
-    "gemini-2.0-flash": "gemini-2.0-flash",
-    "gemini-2.0-flash-lite": "gemini-2.0-flash-lite",
-    # "llama4 scout":"meta-llama/llama-4-scout-17b-16e-instruct"
-    # "llama3.1": "llama-3.1-8b-instant"
+    "llama4 maverik":"meta-llama/llama-4-maverick-17b-128e-instruct",
+    "llama4 scout":"meta-llama/llama-4-scout-17b-16e-instruct",
+    "gemini-pro": "gemini-1.5-pro"
 }
 
 def log_interaction(user_query, model_name, response_content, generated_code, execution_time, error_message=None, is_image=False):
@@ -101,284 +96,461 @@ def preprocess_and_load_df(path: str) -> pd.DataFrame:
         raise Exception(f"Error loading dataframe: {e}")
 
 
+
 def get_from_user(prompt):
     """Format user prompt"""
     return {"role": "user", "content": prompt}
 
 
+
+
 def ask_question(model_name, question):
     """Ask question with comprehensive error handling and logging"""
     start_time = datetime.now()
-    # ------------------------
-    # Helper functions
-    # ------------------------
-    def make_error_response(msg, log_msg, content=None):
-        """Build error response + log it"""
-        execution_time = (datetime.now() - start_time).total_seconds()
-        log_interaction(
-            user_query=question,
-            model_name=model_name,
-            response_content=content or msg,
-            generated_code="",
-            execution_time=execution_time,
-            error_message=log_msg,
-            is_image=False
-        )
-        return {
-            "role": "assistant",
-            "content": content or msg,
-            "gen_code": "",
-            "ex_code": "",
-            "last_prompt": question,
-            "error": log_msg
-        }
-    def validate_api_token(token, token_name, msg_if_missing):
-        """Check for missing/empty API tokens"""
-        if not token or token.strip() == "":
-            return make_error_response(
-                msg="Missing or empty API token",
-                log_msg="Missing or empty API token",
-                content=msg_if_missing
-            )
-        return None  # OK
-    def run_safe_exec(full_code, df=None, extra_globals=None):
-        """Safely execute generated code and handle errors"""
-        local_vars = {}
+    try:
+        # Reload environment variables to get fresh values
+        load_dotenv(override=True)
+        fresh_groq_token = os.getenv("GROQ_API_KEY")
+        fresh_gemini_token = os.getenv("GEMINI_TOKEN")
         
-        # Force matplotlib to use ULTRA high resolution settings in exec environment
-        # Skip style file and set everything manually to ensure it works
-        plt.rcParams['figure.dpi'] = 1200
-        plt.rcParams['savefig.dpi'] = 1200 
-        plt.rcParams['figure.figsize'] = [9, 6]
-        plt.rcParams['figure.facecolor'] = 'white'
-        plt.rcParams['savefig.facecolor'] = 'white'
-        plt.rcParams['savefig.bbox'] = 'tight'
-        plt.rcParams['font.size'] = 11
-        plt.rcParams['axes.titlesize'] = 14
-        plt.rcParams['axes.labelsize'] = 12
-        plt.rcParams['xtick.labelsize'] = 10
-        plt.rcParams['ytick.labelsize'] = 10
-        plt.rcParams['legend.fontsize'] = 10
+        print(f"ask_question - Fresh Groq Token: {'Present' if fresh_groq_token else 'Missing'}")
         
-        global_vars = {
-            'pd': pd, 'plt': plt, 'os': os,
-            'sns': __import__('seaborn'),
-            'uuid': __import__('uuid'),
-            'calendar': __import__('calendar'),
-            'np': __import__('numpy'),
-            'df': df,  # <-- pass your DataFrame here
-            'st': __import__('streamlit')  # Add streamlit for st.pyplot()
-        }
-
-        # allow user to inject more globals (optional)
-        if extra_globals:
-            global_vars.update(extra_globals)
-
-        try:
-            exec(full_code, global_vars, local_vars)
-            return (
-                local_vars.get('answer', "Code executed but no result was saved in 'answer' variable"),
-                None
-            )
-        except Exception as code_error:
-            return None, str(code_error)
-
-    # ------------------------
-    # Step 1: Reload env vars
-    # ------------------------
-    load_dotenv(override=True)
-    fresh_groq_token = os.getenv("GROQ_API_KEY")
-    fresh_gemini_token = os.getenv("GEMINI_TOKEN")
-    # ------------------------
-    # Step 2: Init LLM
-    # ------------------------
-    try:
-        if "gemini" in model_name:
-            token_error = validate_api_token(
-                fresh_gemini_token,
-                "GEMINI_TOKEN",
-                "Gemini API token not available or empty. Please set GEMINI_TOKEN in your environment variable."
+        # Check API availability with fresh tokens
+        if model_name == "gemini-pro":
+            if not fresh_gemini_token or fresh_gemini_token.strip() == "":
+                execution_time = (datetime.now() - start_time).total_seconds()
+                error_msg = "Missing or empty API token"
+                
+                # Log the failed interaction
+                log_interaction(
+                    user_query=question,
+                    model_name=model_name,
+                    response_content="Gemini API token not available or empty",
+                    generated_code="",
+                    execution_time=execution_time,
+                    error_message=error_msg,
+                    is_image=False
+                )
+                
+                return {
+                    "role": "assistant", 
+                    "content": "Gemini API token not available or empty. Please set GEMINI_TOKEN in your environment variables.", 
+                    "gen_code": "", 
+                    "ex_code": "", 
+                    "last_prompt": question,
+                    "error": error_msg
+                }
+            llm = ChatGoogleGenerativeAI(
+                model=models[model_name], 
+                google_api_key=fresh_gemini_token,
+                temperature=0
             )
-            if token_error:
-                return token_error
+        else:
+            if not fresh_groq_token or fresh_groq_token.strip() == "":
+                execution_time = (datetime.now() - start_time).total_seconds()
+                error_msg = "Missing or empty API token"
+                
+                # Log the failed interaction
+                log_interaction(
+                    user_query=question,
+                    model_name=model_name,
+                    response_content="Groq API token not available or empty",
+                    generated_code="",
+                    execution_time=execution_time,
+                    error_message=error_msg,
+                    is_image=False
+                )
+                
+                return {
+                    "role": "assistant", 
+                    "content": "Groq API token not available or empty. Please set GROQ_API_KEY in your environment variables and restart the application.", 
+                    "gen_code": "", 
+                    "ex_code": "", 
+                    "last_prompt": question,
+                    "error": error_msg
+                }
             
+            # Test the API key by trying to create the client
             try:
-                llm = ChatGoogleGenerativeAI(
-                    model=models[model_name],
-                    google_api_key=fresh_gemini_token,
-                    temperature=0
+                llm = ChatGroq(
+                    model=models[model_name], 
+                    api_key=fresh_groq_token,
+                    temperature=0.1
                 )
-                # Gemini requires async call
-                llm.invoke("Test")
-                # print("Gemini API key test successful")
+                # Test with a simple call to verify the API key works
+                test_response = llm.invoke("Test")
+                print("API key test successful")
             except Exception as api_error:
-                return make_error_response(
-                    msg="API Connection Error",
-                    log_msg=str(api_error),
-                    content="API Key Error: Your Gemini API key appears to be invalid, expired, or restricted. Please check your GEMINI_TOKEN in the .env file."
-                    if "organization_restricted"in str(api_error).lower() or "unauthorized" in str(api_error).lower()
-                    else f"API Connection Error: {api_error}"
+                execution_time = (datetime.now() - start_time).total_seconds()
+                error_msg = str(api_error)
+                
+                if "organization_restricted" in error_msg.lower() or "unauthorized" in error_msg.lower():
+                    response_content = "API Key Error: Your Groq API key appears to be invalid, expired, or restricted. Please check your API key in the .env file."
+                    log_error_msg = f"API key validation failed: {error_msg}"
+                else:
+                    response_content = f"API Connection Error: {error_msg}"
+                    log_error_msg = error_msg
+                
+                # Log the failed interaction
+                log_interaction(
+                    user_query=question,
+                    model_name=model_name,
+                    response_content=response_content,
+                    generated_code="",
+                    execution_time=execution_time,
+                    error_message=log_error_msg,
+                    is_image=False
                 )
+                
+                return {
+                    "role": "assistant", 
+                    "content": response_content, 
+                    "gen_code": "", 
+                    "ex_code": "", 
+                    "last_prompt": question,
+                    "error": log_error_msg
+                }
 
-        else:
-            token_error = validate_api_token(
-                fresh_groq_token,
-                "GROQ_API_KEY",
-                "Groq API token not available or empty. Please set GROQ_API_KEY in your environment variables and restart the application."
+        # Check if data file exists
+        if not os.path.exists("Data.csv"):
+            execution_time = (datetime.now() - start_time).total_seconds()
+            error_msg = "Data file not found"
+            
+            # Log the failed interaction
+            log_interaction(
+                user_query=question,
+                model_name=model_name,
+                response_content="Data.csv file not found",
+                generated_code="",
+                execution_time=execution_time,
+                error_message=error_msg,
+                is_image=False
             )
-            if token_error:
-                return token_error
+            
+            return {
+                "role": "assistant", 
+                "content": "Data.csv file not found. Please ensure the data file is in the correct location.", 
+                "gen_code": "", 
+                "ex_code": "", 
+                "last_prompt": question,
+                "error": error_msg
+            }
 
-            try:
-                llm = ChatGroq(
-                    model=models[model_name],
-                    api_key=fresh_groq_token,
-                    temperature=0
-                )
-                llm.invoke("Test")  # test API key
-                # print("Groq API key test successful")
-            except Exception as api_error:
-                return make_error_response(
-                    msg="API Connection Error",
-                    log_msg=str(api_error),
-                    content="API Key Error: Your Groq API key appears to be invalid, expired, or restricted. Please check your GROQ_API_KEY in the .env file."
-                    if "organization_restricted"in str(api_error).lower() or "unauthorized" in str(api_error).lower()
-                    else f"API Connection Error: {api_error}"
-                )
-    except Exception as e:
-        return make_error_response(str(e), str(e))
-    # ------------------------
-    # Step 3: Check AQ_met_data.csv
-    # ------------------------
-    if not os.path.exists("AQ_met_data.csv"):
-        return make_error_response(
-            msg="Data file not found",
-            log_msg="Data file not found",
-            content="AQ_met_data.csv file not found. Please ensure the data file is in the correct location."
-        )
+        df_check = pd.read_csv("Data.csv")
+        df_check["Timestamp"] = pd.to_datetime(df_check["Timestamp"])
+        df_check = df_check.head(5)
 
-    df = pd.read_csv("AQ_met_data.csv")
-    df["Timestamp"] = pd.to_datetime(df["Timestamp"])
-    new_line = "\n"
-    states_df = pd.read_csv("states_data.csv")
-    ncap_df = pd.read_csv("ncap_funding_data.csv")
-    
-    # Template for user query
-    template = f"""```python
+        new_line = "\n"
+
+        template = f"""```python
 import pandas as pd
 import matplotlib.pyplot as plt
-import seaborn as sns
-import streamlit as st
 import uuid
 import calendar
 import numpy as np
-# Set professional matplotlib styling with high resolution
-plt.style.use('vayuchat.mplstyle')
-df = pd.read_csv("AQ_met_data.csv")
+
+# Set professional matplotlib styling
+plt.rcParams.update({{
+    'font.size': 12,
+    'figure.dpi': 400,
+    'figure.facecolor': 'white',
+    'axes.facecolor': 'white',
+    'axes.edgecolor': '#e2e8f0',
+    'axes.linewidth': 1.2,
+    'axes.labelcolor': '#374151',
+    'axes.spines.top': False,
+    'axes.spines.right': False,
+    'axes.spines.left': True,
+    'axes.spines.bottom': True,
+    'axes.grid': True,
+    'grid.color': '#f1f5f9',
+    'grid.linewidth': 0.8,
+    'grid.alpha': 0.7,
+    'xtick.color': '#6b7280',
+    'ytick.color': '#6b7280',
+    'text.color': '#374151',
+    'figure.figsize': [12, 6],
+    'axes.prop_cycle': plt.cycler('color', ['#3b82f6', '#ef4444', '#10b981', '#f59e0b', '#8b5cf6', '#06b6d4'])
+}})
+
+df = pd.read_csv("Data.csv")
 df["Timestamp"] = pd.to_datetime(df["Timestamp"])
-states_df = pd.read_csv("states_data.csv")
-ncap_df = pd.read_csv("ncap_funding_data.csv")
-# df is pandas DataFrame with air quality data from India. Data frequency is daily from 2017 to 2024. The data has the following columns and data types:
-{new_line.join(map(lambda x: '# '+x, str(df.dtypes).split(new_line)))}
-# states_df is a pandas DataFrame of state-wise population, area and whether state is union territory or not of India.
-{new_line.join(map(lambda x: '# '+x, str(states_df.dtypes).split(new_line)))}
-# ncap_df is a pandas DataFrame of funding given to the cities of India from 2019-2022, under The National Clean Air Program (NCAP).
-{new_line.join(map(lambda x: '# '+x, str(ncap_df.dtypes).split(new_line)))}
+
+# Available columns and data types:
+{new_line.join(map(lambda x: '# '+x, str(df_check.dtypes).split(new_line)))}
+
 # Question: {question.strip()}
 # Generate code to answer the question and save result in 'answer' variable
 # If creating a plot, save it with a unique filename and store the filename in 'answer'
 # If returning text/numbers, store the result directly in 'answer'
 ```"""
 
-    # Read system prompt from txt file
-    with open("new_system_prompt.txt", "r", encoding="utf-8") as f:
-        system_prompt = f.read().strip()
-
-    messages = [
-        {
-            "role": "system",
-            "content": system_prompt
-        },
-        {
-            "role": "user",
-            "content": f"""Complete the following code to answer the user's question: 
-            {template}"""
-        }
-    ]
-    
-    # ------------------------
-    # Step 4: Call model
-    # ------------------------
-    try:
-        response = llm.invoke(messages)
-        answer = response.content
-    except Exception as e:
-        return make_error_response(f"Error: {e}", str(e))
+        system_prompt = """Generate Python code to answer the user's question about air quality data.
+
+CRITICAL: Only generate Python code - no explanations, no thinking, just clean executable code.
+
+AVAILABLE LIBRARIES:
+You can use these pre-installed libraries:
+- pandas, numpy (data manipulation)
+- matplotlib, seaborn, plotly (visualization) 
+- statsmodels (statistical modeling, trend analysis)
+- scikit-learn (machine learning, regression)
+- geopandas (geospatial analysis)
+
+LIBRARY USAGE RULES:
+- For trend analysis: Use numpy.polyfit(x, y, 1) for simple linear trends
+- For regression: Use sklearn.linear_model.LinearRegression() for robust regression
+- For statistical modeling: Use statsmodels only if needed, otherwise use numpy/sklearn
+- Always import libraries at the top: import numpy as np, from sklearn.linear_model import LinearRegression
+- Handle missing libraries gracefully with try-except around imports
+
+OUTPUT TYPE REQUIREMENTS:
+1. PLOT GENERATION (for "plot", "chart", "visualize", "show trend", "graph"):
+   - MUST create matplotlib figure with proper labels, title, legend
+   - MUST save plot: filename = f"plot_{uuid.uuid4().hex[:8]}.png" 
+   - MUST call plt.savefig(filename, dpi=300, bbox_inches='tight')
+   - MUST call plt.close() to prevent memory leaks
+   - MUST store filename in 'answer' variable: answer = filename
+   - Handle empty data gracefully before plotting
+
+2. TEXT ANSWERS (for simple "Which", "What", single values):
+   - Store direct string answer in 'answer' variable
+   - Example: answer = "December had the highest pollution"
+
+3. DATAFRAMES (for lists, rankings, comparisons, multiple results):
+   - Create clean DataFrame with descriptive column names
+   - Sort appropriately for readability
+   - Store DataFrame in 'answer' variable: answer = result_df
+
+MANDATORY SAFETY & ROBUSTNESS RULES:
+
+DATA VALIDATION (ALWAYS CHECK):
+- Check if DataFrame exists and not empty: if df.empty: answer = "No data available"
+- Validate required columns exist: if 'PM2.5' not in df.columns: answer = "Required data not available"
+- Check for sufficient data: if len(df) < 10: answer = "Insufficient data for analysis"  
+- Remove invalid/missing values: df = df.dropna(subset=['PM2.5', 'city', 'Timestamp'])
+- Use early exit pattern: if condition: answer = "error message"; else: continue with analysis
+
+OPERATION SAFETY (PREVENT CRASHES):
+- Wrap risky operations in try-except blocks
+- Check denominators before division: if denominator == 0: continue
+- Validate indexing bounds: if idx >= len(array): continue
+- Check for empty results after filtering: if result_df.empty: answer = "No data found"
+- Convert data types explicitly: pd.to_numeric(), .astype(int), .astype(str)
+- Handle timezone issues with datetime operations
+- NO return statements - this is script context, use if/else logic flow
+
+PLOT GENERATION (MANDATORY FOR PLOTS):
+- Check data exists before plotting: if plot_data.empty: answer = "No data to plot"
+- Always create new figure: plt.figure(figsize=(12, 8))
+- Add comprehensive labels: plt.title(), plt.xlabel(), plt.ylabel()
+- Handle long city names: plt.xticks(rotation=45, ha='right') 
+- Use tight layout: plt.tight_layout()
+- CRITICAL PLOT SAVING SEQUENCE (no return statements):
+  1. filename = f"plot_{uuid.uuid4().hex[:8]}.png"
+  2. plt.savefig(filename, dpi=300, bbox_inches='tight')
+  3. plt.close()
+  4. answer = filename
+- Use if/else logic: if data_valid: create_plot(); answer = filename else: answer = "error"
+
+CRITICAL CODING PRACTICES:
+
+DATA VALIDATION & SAFETY:
+- Always check if DataFrames/Series are empty before operations: if df.empty: return
+- Use .dropna() to handle missing values or .fillna() with appropriate defaults
+- Validate column names exist before accessing: if 'column' in df.columns
+- Check data types before operations: df['col'].dtype, isinstance() checks
+- Handle edge cases: empty results, single row/column DataFrames, all NaN columns
+- Use .copy() when modifying DataFrames to avoid SettingWithCopyWarning
+
+VARIABLE & TYPE HANDLING:
+- Use descriptive variable names (avoid single letters in complex operations)
+- Ensure all variables are defined before use - initialize with defaults
+- Convert pandas/numpy objects to proper Python types before operations
+- Convert datetime/period objects appropriately: .astype(str), .dt.strftime(), int()
+- Always cast to appropriate types for indexing: int(), str(), list()
+- CRITICAL: Convert pandas/numpy values to int before list indexing: int(value) for calendar.month_name[int(month_value)]
+- Use explicit type conversions rather than relying on implicit casting
+
+PANDAS OPERATIONS:
+- Reference DataFrame properly: df['column'] not 'column' in operations
+- Use .loc/.iloc correctly for indexing - avoid chained indexing
+- Use .reset_index() after groupby operations when needed for clean DataFrames
+- Sort results for consistent output: .sort_values(), .sort_index()
+- Use .round() for numerical results to avoid excessive decimals
+- Chain operations carefully - split complex chains for readability
+
+MATPLOTLIB & PLOTTING:
+- Always call plt.close() after saving plots to prevent memory leaks
+- Use descriptive titles, axis labels, and legends
+- Handle cases where no data exists for plotting
+- Use proper figure sizing: plt.figure(figsize=(width, height))
+- Convert datetime indices to strings for plotting if needed
+- Use color palettes consistently
+
+ERROR PREVENTION:
+- Use try-except blocks for operations that might fail
+- Check denominators before division operations
+- Validate array/list lengths before indexing
+- Use .get() method for dictionary access with defaults
+- Handle timezone-aware vs naive datetime objects consistently
+- Use proper string formatting and encoding for text output
+
+TECHNICAL REQUIREMENTS:
+- Save final result in variable called 'answer'
+- For TEXT: Store the direct answer as a string in 'answer'
+- For PLOTS: Save with unique filename f"plot_{{uuid.uuid4().hex[:8]}}.png" and store filename in 'answer'
+- For DATAFRAMES: Store the pandas DataFrame directly in 'answer' (e.g., answer = result_df)
+- Always use .iloc or .loc properly for pandas indexing
+- Close matplotlib figures with plt.close() to prevent memory leaks
+- Use proper column name checks before accessing columns
+- For dataframes, ensure proper column names and sorting for readability
+"""
+
+        query = f"""{system_prompt}
+
+Complete the following code to answer the user's question:
+
+{template}
+"""
+        
+        # Make API call
+        if model_name == "gemini-pro":
+            response = llm.invoke(query)
+            answer = response.content
+        else:
+            response = llm.invoke(query)
+            answer = response.content
+        
+        # Extract and execute code with enhanced error handling
+        try:
+            if "```python" in answer:
+                code_part = answer.split("```python")[1].split("```")[0]
+            else:
+                code_part = answer
             
-    # ------------------------
-    # Step 5: Extract code
-    # ------------------------
-    code_part = answer.split("```python")[1].split("```")[0] if "```python" in answer else answer
-    full_code = f"""
+            full_code = f"""
 {template.split("```python")[1].split("```")[0]}
 {code_part}
 """
-    answer_result, code_error = run_safe_exec(full_code, df, extra_globals={'states_df': states_df, 'ncap_df': ncap_df})
-
-    execution_time = (datetime.now() - start_time).total_seconds()
-    if code_error:
-        # Friendly error messages
-        msg = "I encountered an error while analyzing your data. "
-        if "syntax" in code_error.lower():
-            msg += "There was a syntax error in the generated code. Please try rephrasing your question."
-        elif "not defined" in code_error.lower():
-            msg += "Variable naming error occurred. Please try asking the question again."
-        elif "division by zero" in code_error.lower():
-            msg += "Calculation involved division by zero, possibly due to missing data."
-        elif "no data" in code_error.lower() or "empty" in code_error.lower():
-            msg += "No relevant data was found for your query."
+            
+            # Execute code in a controlled environment with better error handling
+            local_vars = {}
+            global_vars = {
+                'pd': pd,
+                'plt': plt,
+                'os': os,
+                'uuid': __import__('uuid'),
+                'calendar': __import__('calendar'),
+                'np': __import__('numpy')
+            }
+            
+            exec(full_code, global_vars, local_vars)
+            
+            # Get the answer
+            if 'answer' in local_vars:
+                answer_result = local_vars['answer']
+            else:
+                answer_result = "Code executed but no result was saved in 'answer' variable"
+            
+            execution_time = (datetime.now() - start_time).total_seconds()
+            
+            # Determine if output is an image
+            is_image = isinstance(answer_result, str) and any(answer_result.endswith(ext) for ext in ['.png', '.jpg', '.jpeg'])
+            
+            # Log successful interaction
+            log_interaction(
+                user_query=question,
+                model_name=model_name,
+                response_content=str(answer_result),
+                generated_code=full_code,
+                execution_time=execution_time,
+                error_message=None,
+                is_image=is_image
+            )
+                
+            return {
+                "role": "assistant", 
+                "content": answer_result, 
+                "gen_code": full_code, 
+                "ex_code": full_code, 
+                "last_prompt": question,
+                "error": None
+            }
+            
+        except Exception as code_error:
+            execution_time = (datetime.now() - start_time).total_seconds()
+            error_msg = str(code_error)
+            
+            # Classify and provide user-friendly error messages
+            user_friendly_msg = "I encountered an error while analyzing your data. "
+            
+            if "unmatched" in error_msg.lower() or "invalid syntax" in error_msg.lower():
+                user_friendly_msg += "There was a syntax error in the generated code (missing brackets or quotes). Please try rephrasing your question or try again."
+            elif "not defined" in error_msg.lower():
+                user_friendly_msg += "There was a variable naming error in the generated code. Please try asking the question again."
+            elif "has no attribute" in error_msg.lower():
+                user_friendly_msg += "There was an issue accessing data properties. Please try a simpler version of your question."
+            elif "division by zero" in error_msg.lower():
+                user_friendly_msg += "The calculation involved division by zero, possibly due to missing data. Please try a different time period or location."
+            elif "empty" in error_msg.lower() or "no data" in error_msg.lower():
+                user_friendly_msg += "No relevant data was found for your query. Please try adjusting the time period, location, or criteria."
+            else:
+                user_friendly_msg += f"Technical error: {error_msg}"
+            
+            user_friendly_msg += "\n\n💡 **Suggestions:**\n- Try rephrasing your question\n- Use simpler terms\n- Check if the data exists for your specified criteria"
+            
+            # Log the failed code execution
+            log_interaction(
+                user_query=question,
+                model_name=model_name,
+                response_content=user_friendly_msg,
+                generated_code=full_code if 'full_code' in locals() else "",
+                execution_time=execution_time,
+                error_message=error_msg,
+                is_image=False
+            )
+            
+            return {
+                "role": "assistant", 
+                "content": user_friendly_msg, 
+                "gen_code": full_code if 'full_code' in locals() else "", 
+                "ex_code": full_code if 'full_code' in locals() else "", 
+                "last_prompt": question,
+                "error": error_msg
+            }
+            
+    except Exception as e:
+        execution_time = (datetime.now() - start_time).total_seconds()
+        error_msg = str(e)
+        
+        # Handle specific API errors
+        if "organization_restricted" in error_msg:
+            response_content = "API Organization Restricted: Your API key access has been restricted. Please check your Groq API key or try generating a new one."
+            log_error_msg = "API access restricted"
+        elif "rate_limit" in error_msg.lower():
+            response_content = "Rate limit exceeded. Please wait a moment and try again."
+            log_error_msg = "Rate limit exceeded"
         else:
-            msg += f"Technical error: {code_error}"
-
-        msg += "\n\n💡 **Suggestions:**\n- Try rephrasing your question\n- Use simpler terms\n- Check if the data exists for your specified criteria"
-
+            response_content = f"Error: {error_msg}"
+            log_error_msg = error_msg
+        
+        # Log the failed interaction
         log_interaction(
             user_query=question,
             model_name=model_name,
-            response_content=msg,
-            generated_code=full_code,
+            response_content=response_content,
+            generated_code="",
             execution_time=execution_time,
-            error_message=code_error,
+            error_message=log_error_msg,
             is_image=False
         )
+        
         return {
-            "role": "assistant",
-            "content": msg,
-            "gen_code": full_code,
-            "ex_code": full_code,
+            "role": "assistant", 
+            "content": response_content, 
+            "gen_code": "", 
+            "ex_code": "", 
             "last_prompt": question,
-            "error": code_error
-        }
-        
-    # ------------------------
-    # Step 7: Success logging
-    # ------------------------
-    is_image = isinstance(answer_result, str) and answer_result.endswith(('.png', '.jpg', '.jpeg'))
-    log_interaction(
-        user_query=question,
-        model_name=model_name,
-        response_content=str(answer_result),
-        generated_code=full_code,
-        execution_time=execution_time,
-        error_message=None,
-        is_image=is_image
-    )
-
-    return {
-        "role": "assistant",
-        "content": answer_result,
-        "gen_code": full_code,
-        "ex_code": full_code,
-        "last_prompt": question,
-        "error": None
-    }
+            "error": log_error_msg
+        }

states_data.csv

@@ -1,32 +0,0 @@
-state,population,area (km2),isUnionTerritory
-Uttar Pradesh,199812341,240928,False
-Maharashtra,112374333,307713,False
-Bihar,104099452,94163,False
-West Bengal,91276115,88752,False
-Madhya Pradesh,72626809,308252,False
-Tamil Nadu,72147030,130058,False
-Rajasthan,68548437,342239,False
-Karnataka,61095297,191791,False
-Gujarat,60439692,196024,False
-Andhra Pradesh,49577103,162975,False
-Odisha,41974219,155707,False
-Telangana,35003674,112077,False
-Kerala,33406061,38863,False
-Jharkhand,32988134,79716,False
-Assam,31205576,78438,False
-Punjab,27743338,50362,False
-Chhattisgarh,25545198,135192,False
-Delhi,16787941,1484,True
-Haryana,25351462,44212,False
-Jammu and Kashmir,12267032,42241,True
-Uttarakhand,10086292,53483,False
-Himachal Pradesh,6864602,55673,False
-Tripura,3673917,10491,False
-Manipur,2570390,22327,False
-Meghalaya,2966889,22429,False
-Nagaland,1978502,16579,False
-Arunachal Pradesh,1383727,83743,False
-Puducherry,1247953,479,True
-Mizoram,1097206,21081,False
-Chandigarh,1055450,114,True
-Sikkim,610577,7096,False

system_prompt.txt

@@ -3,7 +3,6 @@ I have a pandas dataframe data of PM2.5 and PM10.
 * Frequency of data is daily.
 * `pollution` generally means `PM2.5`.
 * You already have df, so don't read the csv file 
-* Available libraries: pandas, matplotlib, numpy, seaborn, plotly, geopandas, statsmodels, scikit-learn
 * Don't print anything, but save result in a variable `answer` and make it global.
 * Unless explicitly mentioned, don't consider the result as a plot.
 * PM2.5 guidelines: India: 60, WHO: 15.

test_image.py

@@ -1,129 +0,0 @@
-import pandas as pd
-import matplotlib.pyplot as plt
-import seaborn as sns
-import uuid
-import calendar
-import numpy as np
-# Set professional matplotlib styling with high resolution
-#plt.style.use('vayuchat.mplstyle')
-df = pd.read_csv("AQ_met_data.csv")
-df["Timestamp"] = pd.to_datetime(df["Timestamp"])
-states_df = pd.read_csv("states_data.csv")
-ncap_df = pd.read_csv("ncap_funding_data.csv")
-# df is pandas DataFrame with air quality data from India. Data frequency is daily from 2017 to 2024. The data has the following columns and data types:
-# Unnamed: 0                int64
-# Timestamp        datetime64[ns]
-# State                    object
-# City                     object
-# Station                  object
-# site_id                  object
-# Year                      int64
-# PM2.5 (µg/m³)           float64
-# PM10 (µg/m³)            float64
-# NO (µg/m³)              float64
-# NO2 (µg/m³)             float64
-# NOx (ppb)               float64
-# NH3 (µg/m³)             float64
-# SO2 (µg/m³)             float64
-# CO (mg/m³)              float64
-# Ozone (µg/m³)           float64
-# AT (°C)                 float64
-# RH (%)                  float64
-# WS (m/s)                float64
-# WD (deg)                float64
-# RF (mm)                 float64
-# TOT-RF (mm)             float64
-# SR (W/mt2)              float64
-# BP (mmHg)               float64
-# VWS (m/s)               float64
-# dtype: object
-# states_df is a pandas DataFrame of state-wise population, area and whether state is union territory or not of India.
-# state               object
-# population           int64
-# area (km2)           int64
-# isUnionTerritory      bool
-# dtype: object
-# ncap_df is a pandas DataFrame of funding given to the cities of India from 2019-2022, under The National Clean Air Program (NCAP).
-# S. No.                                 int64
-# state                                 object
-# city                                  object
-# Amount released during FY 2019-20    float64
-# Amount released during FY 2020-21    float64
-# Amount released during FY 2021-22    float64
-# Total fund released                  float64
-# Utilisation as on June 2022          float64
-# dtype: object
-# Question: Compare the wind speed and PM2.5 levels during Delhi’s most polluted week (highest PM2.5) in December 2024 with the previous 15 days and the following 15 days on a time series plot.
-# Generate code to answer the question and save result in 'answer' variable
-# If creating a plot, save it with a unique filename and store the filename in 'answer'
-# If returning text/numbers, store the result directly in 'answer'
-
-
-import pandas as pd
-import matplotlib.pyplot as plt
-import seaborn as sns
-import uuid
-import numpy as np
-
-# Ensure data is loaded
-if df.empty:
-    answer = "No data available"
-else:
-    try:
-        # Filter for Delhi in December 2024
-        df_delhi = df[
-            (df['City'].str.contains('Delhi', case=False, na=False)) &
-            (df['Timestamp'].dt.year == 2024) &
-            (df['Timestamp'].dt.month == 12)
-        ].copy()
-        df_delhi = df_delhi.dropna(subset=['PM2.5 (µg/m³)', 'WS (m/s)'])
-        df_delhi = df_delhi.sort_values('Timestamp')
-
-        # Need at least a full week to compute rolling mean
-        if len(df_delhi) < 7:
-            answer = "Insufficient data"
-        else:
-            # Compute 7‑day rolling mean of PM2.5
-            df_delhi['PM2.5_roll7'] = df_delhi['PM2.5 (µg/m³)'].rolling(window=7, min_periods=7).mean()
-            # Identify the window with the highest mean PM2.5
-            max_idx = df_delhi['PM2.5_roll7'].idxmax()
-            max_end_date = df_delhi.loc[max_idx, 'Timestamp']
-            max_start_date = max_end_date - pd.Timedelta(days=6)
-
-            # Define extended window: 15 days before start and 15 days after end
-            ext_start = max_start_date - pd.Timedelta(days=15)
-            ext_end = max_end_date + pd.Timedelta(days=15)
-
-            # Filter data for the extended period
-            mask = (df_delhi['Timestamp'] >= ext_start) & (df_delhi['Timestamp'] <= ext_end)
-            df_plot = df_delhi.loc[mask].copy()
-
-            if df_plot.empty or len(df_plot) < 30:
-                answer = "Insufficient data"
-            else:
-                # Plot time series
-                plt.figure(figsize=(9, 6))
-                ax1 = plt.gca()
-                sns.lineplot(data=df_plot, x='Timestamp', y='PM2.5 (µg/m³)', ax=ax1,
-                             label='PM2.5 (µg/m³)', color='tab:red')
-                ax1.set_ylabel('PM2.5 (µg/m³)', color='tab:red')
-                ax1.tick_params(axis='y', labelcolor='tab:red')
-
-                ax2 = ax1.twinx()
-                sns.lineplot(data=df_plot, x='Timestamp', y='WS (m/s)', ax=ax2,
-                             label='Wind Speed (m/s)', color='tab:blue')
-                ax2.set_ylabel('Wind Speed (m/s)', color='tab:blue')
-                ax2.tick_params(axis='y', labelcolor='tab:blue')
-
-                plt.title('Delhi – PM2.5 and Wind Speed around Most Polluted Week (Dec 2024)')
-                plt.xlabel('Date')
-                plt.tight_layout()
-
-                # Save plot
-                filename = f"plot.png"
-                plt.savefig(filename, dpi=1200, bbox_inches='tight', facecolor='white')
-                plt.close()
-
-                answer = filename
-    except Exception as e:
-        answer = "Unable to complete analysis with available data"

vayuchat.mplstyle

@@ -1,93 +0,0 @@
-# VayuChat - Modern Professional Style
-# Inspired by modern data visualization best practices
-
-# Typography & Layout
-font.size: 11
-font.family: sans-serif
-font.sans-serif: Inter, SF Pro Display, Segoe UI, system-ui, Arial
-figure.titlesize: 14
-axes.titlesize: 12
-axes.labelsize: 10
-xtick.labelsize: 9
-ytick.labelsize: 9
-legend.fontsize: 9
-
-# Figure & DPI - Ultra High Resolution
-figure.dpi: 1200
-figure.facecolor: white
-figure.edgecolor: none
-figure.figsize: 9, 6
-figure.autolayout: True
-
-# Modern Color Palette (inspired by Tailwind/GitHub)
-axes.prop_cycle: cycler('color', ['2563eb', 'dc2626', '059669', 'ea580c', '7c3aed', '0891b2', 'be123c', '16a34a', 'c2410c', '9333ea'])
-
-# Axes Styling
-axes.facecolor: white
-axes.edgecolor: e5e7eb
-axes.linewidth: 1
-axes.labelcolor: 374151
-axes.axisbelow: True
-axes.spines.left: True
-axes.spines.bottom: True
-axes.spines.top: False
-axes.spines.right: False
-
-# Grid (subtle and clean)
-axes.grid: True
-grid.color: f3f4f6
-grid.linewidth: 0.8
-grid.alpha: 0.7
-axes.grid.axis: both
-
-# Ticks
-xtick.direction: out
-ytick.direction: out
-xtick.major.size: 4
-ytick.major.size: 4
-xtick.minor.size: 2
-ytick.minor.size: 2
-xtick.color: 6b7280
-ytick.color: 6b7280
-xtick.major.pad: 7
-ytick.major.pad: 7
-
-# Legend
-legend.frameon: True
-legend.fancybox: True
-legend.shadow: False
-legend.framealpha: 0.95
-legend.facecolor: white
-legend.edgecolor: e5e7eb
-legend.borderpad: 0.8
-legend.columnspacing: 2
-legend.handlelength: 1.5
-legend.handletextpad: 0.8
-
-# Lines & Markers
-lines.linewidth: 2.5
-lines.markersize: 7
-lines.solid_capstyle: round
-patch.linewidth: 0.5
-patch.facecolor: 3b82f6
-patch.edgecolor: none
-patch.antialiased: True
-
-# Scatter plots
-scatter.marker: o
-scatter.edgecolors: white
-
-# Bars
-patch.force_edgecolor: False
-
-# Text & Annotations
-text.color: 1f2937
-text.antialiased: True
-
-# Savefig - Ultra High Resolution
-savefig.dpi: 1200
-savefig.facecolor: white
-savefig.edgecolor: none
-savefig.bbox: tight
-savefig.pad_inches: 0.2
-savefig.format: png