app.py

import os
import base64
import io
import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
from dash import Dash, html, dcc, Input, Output, State, callback_context
import dash_bootstrap_components as dbc
from typing import Optional
from dotenv import load_dotenv
from pydantic import Field, SecretStr

# Fixed Langchain imports (using langchain-huggingface for v0.2+)
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.schema import Document
from langchain_openai import ChatOpenAI
from langchain_core.prompts import PromptTemplate
from langchain.chains import LLMChain

# Load environment variables
load_dotenv()

class ChatOpenRouter(ChatOpenAI):
    def __init__(self,
                 openai_api_key: Optional[str] = None,
                 **kwargs):
        openai_api_key = openai_api_key or os.environ.get("OPENROUTER_API_KEY")
        super().__init__(
            base_url="https://openrouter.ai/api/v1", 
            openai_api_key=openai_api_key, 
            **kwargs
        )

# Initialize OpenRouter model
openrouter_model = ChatOpenRouter(
    model="microsoft/phi-4-reasoning-plus",
    temperature=0.3,
    max_tokens=1500,
    model_kwargs={
        "top_p": 0.9,
        "frequency_penalty": 0.0,
        "presence_penalty": 0.0
    },
    streaming=False        
)

# Initialize Dash app
app = Dash(__name__, external_stylesheets=[dbc.themes.BOOTSTRAP])
server = app.server

# Initialize Langchain components (removed @st.cache_resource)
def init_langchain():
    """Initialize Langchain components"""
    try:
        # Use a lightweight model for embeddings
        embeddings = HuggingFaceEmbeddings(
            model_name="sentence-transformers/all-MiniLM-L6-v2",
            model_kwargs={'device': 'cpu'}
        )
        
        # Initialize text splitter
        text_splitter = RecursiveCharacterTextSplitter(
            chunk_size=1000,
            chunk_overlap=200
        )
        
        return embeddings, text_splitter
    except Exception as e:
        print(f"Error initializing Langchain: {e}")
        return None, None

# Global variables
embeddings, text_splitter = init_langchain()
vector_store = None

# App layout
app.layout = dbc.Container([
    dbc.Row([
        dbc.Col([
            html.H1("🤖 AI-Powered Data Analytics", className="text-center mb-4"),
            html.P("Upload data, ask questions, and get AI-powered insights!", 
                   className="text-center text-muted"),
            html.Hr(),
        ], width=12)
    ]),
    
    dbc.Row([
        dbc.Col([
            dbc.Card([
                dbc.CardBody([
                    html.H4("📁 Data Upload", className="card-title"),
                    dcc.Upload(
                        id='upload-data',
                        children=html.Div([
                            'Drag and Drop or ',
                            html.A('Select Files')
                        ]),
                        style={
                            'width': '100%',
                            'height': '60px',
                            'lineHeight': '60px',
                            'borderWidth': '1px',
                            'borderStyle': 'dashed',
                            'borderRadius': '5px',
                            'textAlign': 'center',
                            'margin': '10px'
                        },
                        multiple=False,
                        accept='.csv,.xlsx,.txt'
                    ),
                    
                    html.Div(id='upload-status', className="mt-2"),
                    html.Hr(),
                    
                    html.H4("🤖 AI Assistant", className="card-title"),
                    dbc.InputGroup([
                        dbc.Input(
                            id="ai-question",
                            placeholder="Ask questions about your data...",
                            type="text",
                            style={"fontSize": "14px"}
                        ),
                        dbc.Button(
                            "Ask AI", 
                            id="ask-button", 
                            color="primary",
                            n_clicks=0
                        )
                    ]),
                    
                    html.Div(id="ai-response", className="mt-3"),
                    html.Hr(),
                    
                    html.H4("📊 Quick Analytics", className="card-title"),
                    dbc.ButtonGroup([
                        dbc.Button("Summary Stats", id="stats-btn", size="sm"),
                        dbc.Button("Correlations", id="corr-btn", size="sm"),
                        dbc.Button("Missing Data", id="missing-btn", size="sm"),
                    ], className="w-100"),
                    
                    html.Div(id="quick-analytics", className="mt-3")
                ])
            ])
        ], width=4),
        
        dbc.Col([
            dbc.Card([
                dbc.CardBody([
                    html.H4("📈 Visualizations", className="card-title"),
                    dcc.Graph(id='main-graph', style={'height': '400px'}),
                ])
            ]),
            
            dbc.Card([
                dbc.CardBody([
                    html.H4("🔍 Data Explorer", className="card-title"),
                    html.Div(id='data-table')
                ])
            ], className="mt-3")
        ], width=8)
    ], className="mt-4"),
    
    # Store components
    dcc.Store(id='stored-data'),
    dcc.Store(id='data-context')
], fluid=True)

def create_vector_store(df):
    """Create vector store from dataframe"""
    global vector_store
    
    if embeddings is None:
        return False
    
    try:
        # Convert dataframe to documents
        documents = []
        
        # Add column information
        col_info = f"Dataset has {len(df)} rows and {len(df.columns)} columns.\n"
        col_info += f"Columns: {', '.join(df.columns)}\n"
        col_info += f"Data types: {df.dtypes.to_string()}\n"
        documents.append(Document(page_content=col_info, metadata={"type": "schema"}))
        
        # Add summary statistics
        summary = df.describe().to_string()
        documents.append(Document(page_content=f"Summary statistics:\n{summary}", 
                                metadata={"type": "statistics"}))
        
        # Add sample rows
        sample_data = df.head(10).to_string()
        documents.append(Document(page_content=f"Sample data:\n{sample_data}", 
                                metadata={"type": "sample"}))
        
        # Add correlation information for numeric columns
        numeric_cols = df.select_dtypes(include=['number']).columns
        if len(numeric_cols) > 1:
            corr = df[numeric_cols].corr().to_string()
            documents.append(Document(page_content=f"Correlations:\n{corr}", 
                                    metadata={"type": "correlation"}))
        
        # Create vector store
        vector_store = FAISS.from_documents(documents, embeddings)
        return True
        
    except Exception as e:
        print(f"Error creating vector store: {e}")
        return False

def get_ai_response(question, df):
    """Get AI response using OpenRouter LLM and RAG"""
    global vector_store
    
    if vector_store is None:
        return "Please upload data first to enable AI features."
    
    try:
        # Create data context for the LLM
        data_context = f"""
Dataset Information:
- Shape: {df.shape[0]} rows × {df.shape[1]} columns
- Columns: {', '.join(df.columns)}
- Data Types: {df.dtypes.to_dict()}
- Missing Values: {df.isnull().sum().to_dict()}

Sample Data (first 5 rows):
{df.head().to_string()}

Summary Statistics:
{df.describe().to_string()}
        """
        
        # Create a prompt template for data analysis
        prompt_template = PromptTemplate(
            input_variables=["question", "data_context"],
            template="""
You are a professional data analyst AI assistant. Based on the provided dataset information, answer the user's question with clear, actionable insights.

Dataset Context:
{data_context}

User Question: {question}

Please provide a helpful, accurate response with:
1. Direct answer to the question
2. Key insights or patterns you notice
3. Recommendations or next steps if applicable

Use emojis and markdown formatting to make your response engaging and easy to read.
            """
        )
        
        # Create LLM chain
        llm_chain = LLMChain(
            llm=openrouter_model,
            prompt=prompt_template
        )
        
        # Get response from OpenRouter
        response = llm_chain.run(
            question=question,
            data_context=data_context
        )
        
        return response
        
    except Exception as e:
        # Fallback to basic responses if OpenRouter fails
        print(f"OpenRouter error: {e}")
        return f"""🤖 **AI Assistant** (Limited Mode): 
        I encountered an issue with the AI service. Here's basic info about your data:
        
        📊 **Quick Summary**:
        - Shape: {df.shape[0]} rows × {df.shape[1]} columns
        - Columns: {', '.join(df.columns)}
        - Missing values: {df.isnull().sum().sum()} total
        
        Please check your OPENROUTER_API_KEY configuration.
        """

def parse_contents(contents, filename):
    """Parse uploaded file contents"""
    content_type, content_string = contents.split(',')
    decoded = base64.b64decode(content_string)
    
    try:
        if 'csv' in filename:
            df = pd.read_csv(io.StringIO(decoded.decode('utf-8')))
        elif 'xls' in filename:
            df = pd.read_excel(io.BytesIO(decoded))
        else:
            return None, "Unsupported file type"
        
        return df, None
    except Exception as e:
        return None, f"Error processing file: {str(e)}"

@app.callback(
    [Output('stored-data', 'data'),
     Output('upload-status', 'children'),
     Output('data-table', 'children')],
    [Input('upload-data', 'contents')],
    [State('upload-data', 'filename')]
)
def update_data(contents, filename):
    """Update data when file is uploaded"""
    if contents is None:
        return None, "", ""
    
    df, error = parse_contents(contents, filename)
    
    if error:
        return None, dbc.Alert(error, color="danger"), ""
    
    # Create vector store for AI
    vector_success = create_vector_store(df)
    
    # Create data table preview
    table = dbc.Table.from_dataframe(
        df.head(10), 
        striped=True, 
        bordered=True, 
        hover=True, 
        size='sm'
    )
    
    ai_status = "🤖 AI Ready" if vector_success else "⚠️ AI Limited"
    
    success_msg = dbc.Alert([
        html.H6(f"✅ File uploaded successfully! {ai_status}"),
        html.P(f"Shape: {df.shape[0]} rows × {df.shape[1]} columns"),
        html.P(f"Columns: {', '.join(df.columns.tolist())}")
    ], color="success")
    
    return df.to_dict('records'), success_msg, table

@app.callback(
    Output('ai-response', 'children'),
    [Input('ask-button', 'n_clicks')],
    [State('ai-question', 'value'),
     State('stored-data', 'data')]
)
def handle_ai_question(n_clicks, question, data):
    """Handle AI question"""
    if not n_clicks or not question or not data:
        return ""
    
    df = pd.DataFrame(data)
    response = get_ai_response(question, df)
    
    return dbc.Alert(
        dcc.Markdown(response),
        color="info"
    )

@app.callback(
    Output('quick-analytics', 'children'),
    [Input('stats-btn', 'n_clicks'),
     Input('corr-btn', 'n_clicks'),
     Input('missing-btn', 'n_clicks')],
    [State('stored-data', 'data')]
)
def quick_analytics(stats_clicks, corr_clicks, missing_clicks, data):
    """Handle quick analytics buttons"""
    if not data:
        return ""
    
    df = pd.DataFrame(data)
    ctx = callback_context
    
    if not ctx.triggered:
        return ""
    
    button_id = ctx.triggered[0]['prop_id'].split('.')[0]
    
    if button_id == 'stats-btn':
        stats = df.describe()
        return dbc.Alert([
            html.H6("📊 Summary Statistics"),
            dbc.Table.from_dataframe(stats.reset_index(), size='sm')
        ], color="light")
    
    elif button_id == 'corr-btn':
        numeric_df = df.select_dtypes(include=['number'])
        if len(numeric_df.columns) > 1:
            corr = numeric_df.corr()
            fig = px.imshow(corr, text_auto=True, aspect="auto", 
                          title="Correlation Matrix")
            return dcc.Graph(figure=fig, style={'height': '300px'})
        return dbc.Alert("No numeric columns for correlation analysis", color="warning")
    
    elif button_id == 'missing-btn':
        missing = df.isnull().sum()
        missing = missing[missing > 0]
        if missing.empty:
            return dbc.Alert("✅ No missing values!", color="success")
        return dbc.Alert([
            html.H6("⚠️ Missing Values"),
            html.Pre(missing.to_string())
        ], color="warning")
    
    return ""

@app.callback(
    Output('main-graph', 'figure'),
    [Input('stored-data', 'data')]
)
def update_main_graph(data):
    """Update main visualization"""
    if not data:
        return {}
    
    df = pd.DataFrame(data)
    
    # Create a smart default visualization
    numeric_cols = df.select_dtypes(include=['number']).columns
    categorical_cols = df.select_dtypes(include=['object']).columns
    
    if len(numeric_cols) >= 2:
        # Scatter plot for numeric data
        fig = px.scatter(df, x=numeric_cols[0], y=numeric_cols[1],
                        title=f"Relationship: {numeric_cols[1]} vs {numeric_cols[0]}")
    elif len(numeric_cols) >= 1 and len(categorical_cols) >= 1:
        # Bar chart for mixed data
        fig = px.bar(df, x=categorical_cols[0], y=numeric_cols[0],
                    title=f"Distribution: {numeric_cols[0]} by {categorical_cols[0]}")
    elif len(numeric_cols) >= 1:
        # Histogram for single numeric
        fig = px.histogram(df, x=numeric_cols[0],
                         title=f"Distribution of {numeric_cols[0]}")
    else:
        # Default message
        fig = go.Figure()
        fig.add_annotation(text="Upload data to see visualizations", 
                         x=0.5, y=0.5, showarrow=False)
    
    fig.update_layout(template="plotly_white")
    return fig

if __name__ == '__main__':
    app.run_server(host='0.0.0.0', port=7860, debug=False)