Upload src/finee/ui.py with huggingface_hub

src/finee/ui.py

@@ -0,0 +1,452 @@
+"""
+FinEE UI - Gradio Interface
+============================
+
+Beautiful web UI for financial entity extraction with:
+- Message extraction demo
+- Batch processing
+- File upload (PDF/Image)
+- Analytics dashboard
+- Chat interface
+
+Author: Ranjit Behera
+"""
+
+import json
+from typing import Optional
+
+try:
+    import gradio as gr
+except ImportError:
+    raise ImportError("Please install gradio: pip install gradio")
+
+
+# ============================================================================
+# EXTRACTION FUNCTIONS
+# ============================================================================
+
+def extract_entities(
+    message: str,
+    use_rag: bool = True,
+    use_llm: bool = False
+) -> tuple:
+    """Extract entities from a message."""
+    if not message.strip():
+        return "{}", "", "Please enter a message"
+    
+    try:
+        # Try to import FinEE
+        try:
+            from finee import FinancialExtractor
+            from finee.rag import RAGEngine
+            
+            extractor = FinancialExtractor(use_llm=use_llm)
+            result = extractor.extract(message)
+            
+            # RAG enhancement
+            rag_info = ""
+            if use_rag:
+                rag = RAGEngine()
+                context = rag.retrieve(message)
+                
+                if context.merchant_info:
+                    if not result.get("merchant"):
+                        result["merchant"] = context.merchant_info["name"]
+                    if not result.get("category"):
+                        result["category"] = context.merchant_info["category"]
+                    
+                    rag_info = f"""
+### RAG Context
+- **Merchant**: {context.merchant_info['name']}
+- **Category**: {context.merchant_info['category']}
+- **Confidence Boost**: +{context.confidence_boost:.0%}
+"""
+            
+            # Format output
+            json_output = json.dumps(result, indent=2, ensure_ascii=False)
+            
+            # Create summary
+            summary_parts = []
+            if result.get("amount"):
+                summary_parts.append(f"💰 **Amount**: ₹{result['amount']:,.2f}")
+            if result.get("type"):
+                emoji = "📤" if result["type"] == "debit" else "📥"
+                summary_parts.append(f"{emoji} **Type**: {result['type'].upper()}")
+            if result.get("merchant"):
+                summary_parts.append(f"🏪 **Merchant**: {result['merchant']}")
+            if result.get("beneficiary"):
+                summary_parts.append(f"👤 **Beneficiary**: {result['beneficiary']}")
+            if result.get("category"):
+                summary_parts.append(f"📂 **Category**: {result['category']}")
+            if result.get("bank"):
+                summary_parts.append(f"🏦 **Bank**: {result['bank']}")
+            if result.get("reference"):
+                summary_parts.append(f"🔗 **Reference**: {result['reference']}")
+            
+            summary = "\n".join(summary_parts) if summary_parts else "No entities extracted"
+            summary += rag_info
+            
+            return json_output, summary, "✅ Extraction successful!"
+            
+        except ImportError:
+            # Fallback mock extraction
+            import re
+            result = {}
+            
+            # Amount
+            amount_match = re.search(r'Rs\.?\s*([\d,]+(?:\.\d{2})?)', message)
+            if amount_match:
+                result['amount'] = float(amount_match.group(1).replace(',', ''))
+            
+            # Type
+            if any(w in message.lower() for w in ['debit', 'debited', 'paid']):
+                result['type'] = 'debit'
+            elif any(w in message.lower() for w in ['credit', 'credited']):
+                result['type'] = 'credit'
+            
+            # Bank
+            banks = ['HDFC', 'ICICI', 'SBI', 'Axis', 'Kotak']
+            for bank in banks:
+                if bank.lower() in message.lower():
+                    result['bank'] = bank
+                    break
+            
+            json_output = json.dumps(result, indent=2)
+            summary = f"💰 Amount: ₹{result.get('amount', 'N/A')}\n📤 Type: {result.get('type', 'N/A')}"
+            
+            return json_output, summary, "⚠️ Using mock extractor (install finee for full functionality)"
+    
+    except Exception as e:
+        return "{}", "", f"❌ Error: {str(e)}"
+
+
+def batch_extract(messages_text: str, use_rag: bool = True) -> str:
+    """Extract from multiple messages."""
+    if not messages_text.strip():
+        return "Please enter messages (one per line)"
+    
+    messages = [m.strip() for m in messages_text.split('\n') if m.strip()]
+    results = []
+    
+    for i, msg in enumerate(messages, 1):
+        json_out, summary, status = extract_entities(msg, use_rag, False)
+        try:
+            data = json.loads(json_out)
+            data['_message'] = msg[:50] + '...' if len(msg) > 50 else msg
+            results.append(data)
+        except:
+            results.append({'error': status, '_message': msg[:50]})
+    
+    return json.dumps(results, indent=2, ensure_ascii=False)
+
+
+def analyze_transactions(transactions_json: str) -> str:
+    """Analyze transactions and generate insights."""
+    try:
+        transactions = json.loads(transactions_json)
+        
+        if not isinstance(transactions, list):
+            return "Please provide a list of transactions"
+        
+        # Calculate stats
+        total_debit = sum(t.get('amount', 0) for t in transactions if t.get('type') == 'debit')
+        total_credit = sum(t.get('amount', 0) for t in transactions if t.get('type') == 'credit')
+        
+        # Category breakdown
+        categories = {}
+        for t in transactions:
+            cat = t.get('category', 'other')
+            if cat not in categories:
+                categories[cat] = {'total': 0, 'count': 0}
+            categories[cat]['total'] += t.get('amount', 0)
+            categories[cat]['count'] += 1
+        
+        # Format report
+        report = f"""
+## 📊 Transaction Analysis
+
+### Summary
+- **Total Transactions**: {len(transactions)}
+- **Total Debits**: ₹{total_debit:,.2f}
+- **Total Credits**: ₹{total_credit:,.2f}
+- **Net Flow**: ₹{total_credit - total_debit:,.2f}
+
+### Category Breakdown
+"""
+        
+        sorted_cats = sorted(categories.items(), key=lambda x: x[1]['total'], reverse=True)
+        for cat, data in sorted_cats:
+            pct = (data['total'] / total_debit * 100) if total_debit > 0 else 0
+            report += f"- **{cat.title()}**: ₹{data['total']:,.2f} ({pct:.1f}%) - {data['count']} txns\n"
+        
+        return report
+    
+    except json.JSONDecodeError:
+        return "❌ Invalid JSON. Please paste valid transaction data."
+    except Exception as e:
+        return f"❌ Analysis error: {str(e)}"
+
+
+def chat_response(message: str, history: list) -> str:
+    """Handle chat messages."""
+    if not message.strip():
+        return ""
+    
+    message_lower = message.lower()
+    
+    # Intent detection
+    if any(w in message_lower for w in ['extract', 'parse', 'analyze this']):
+        # Try to extract from the message
+        _, summary, _ = extract_entities(message)
+        return f"I extracted these entities:\n\n{summary}"
+    
+    elif any(w in message_lower for w in ['how much', 'spent', 'spending']):
+        return "To analyze your spending, please share your transaction messages or paste extracted JSON data."
+    
+    elif any(w in message_lower for w in ['help', 'what can']):
+        return """I can help you with:
+
+1. **Extract Entities** - Paste a bank SMS/email and I'll extract the details
+2. **Batch Processing** - Process multiple messages at once
+3. **Analyze Spending** - Get insights from your transactions
+4. **Categorize** - Understand your spending categories
+
+Try pasting a bank message like:
+`HDFC Bank: Rs.2,500 debited from A/c XX1234 on 12-Jan-26. UPI:swiggy@ybl`"""
+    
+    elif 'hello' in message_lower or 'hi' in message_lower:
+        return "Hello! 👋 I'm FinEE, your financial entity extraction assistant. Paste a bank message and I'll extract the details!"
+    
+    else:
+        # Try extraction as default
+        json_out, summary, status = extract_entities(message)
+        if summary and "No entities" not in summary:
+            return f"I found these in your message:\n\n{summary}"
+        else:
+            return "I'm not sure what you mean. Try pasting a bank SMS or email, or type 'help' for more options."
+
+
+# ============================================================================
+# GRADIO UI
+# ============================================================================
+
+def create_ui():
+    """Create the Gradio interface."""
+    
+    # Custom CSS
+    custom_css = """
+    .gradio-container {
+        font-family: 'Inter', sans-serif !important;
+    }
+    .main-header {
+        text-align: center;
+        margin-bottom: 20px;
+    }
+    .output-json {
+        font-family: 'Monaco', 'Menlo', monospace !important;
+    }
+    """
+    
+    with gr.Blocks(
+        title="FinEE - Financial Entity Extractor",
+        theme=gr.themes.Soft(
+            primary_hue="blue",
+            secondary_hue="gray",
+        ),
+        css=custom_css
+    ) as demo:
+        
+        # Header
+        gr.Markdown("""
+        # 🏦 FinEE - Financial Entity Extractor
+        
+        Extract structured data from Indian banking SMS, emails, and statements.
+        
+        [![GitHub](https://img.shields.io/badge/GitHub-Repo-blue)](https://github.com/Ranjitbehera0034/Finance-Entity-Extractor)
+        [![HuggingFace](https://img.shields.io/badge/🤗-Dataset-yellow)](https://huggingface.co/datasets/Ranjit0034/finee-dataset)
+        [![PyPI](https://img.shields.io/badge/PyPI-finee-orange)](https://pypi.org/project/finee/)
+        """)
+        
+        with gr.Tabs():
+            # Tab 1: Single Extraction
+            with gr.Tab("🔍 Extract", id="extract"):
+                with gr.Row():
+                    with gr.Column(scale=1):
+                        input_message = gr.Textbox(
+                            label="Bank Message",
+                            placeholder="Paste your bank SMS, email, or notification here...",
+                            lines=4,
+                        )
+                        
+                        with gr.Row():
+                            use_rag = gr.Checkbox(label="Use RAG", value=True, info="Context-aware extraction")
+                            use_llm = gr.Checkbox(label="Use LLM", value=False, info="For complex cases")
+                        
+                        extract_btn = gr.Button("Extract Entities", variant="primary")
+                        
+                        # Examples
+                        gr.Examples(
+                            examples=[
+                                "HDFC Bank: Rs.2,500 debited from A/c XX1234 on 12-Jan-26. UPI:swiggy@ybl. Ref:123456789012",
+                                "SBI: Rs.15,000 credited to A/c XX4567 from rahul.sharma@oksbi. NEFT Ref: N987654321",
+                                "ICICI: Your EMI of Rs.12,500 for Loan A/c XX8901 debited on 01-Jan-26",
+                                "Axis Bank: Rs.999 debited for Netflix subscription. Card XX5678",
+                                "Kotak: Rs.50,000 transferred to Zerodha Broking. Ref: 456789012345",
+                            ],
+                            inputs=input_message,
+                        )
+                    
+                    with gr.Column(scale=1):
+                        status_output = gr.Textbox(label="Status", interactive=False)
+                        summary_output = gr.Markdown(label="Summary")
+                        json_output = gr.Code(label="JSON Output", language="json")
+                
+                extract_btn.click(
+                    extract_entities,
+                    inputs=[input_message, use_rag, use_llm],
+                    outputs=[json_output, summary_output, status_output]
+                )
+            
+            # Tab 2: Batch Processing
+            with gr.Tab("📋 Batch", id="batch"):
+                with gr.Row():
+                    with gr.Column():
+                        batch_input = gr.Textbox(
+                            label="Messages (one per line)",
+                            placeholder="Paste multiple messages, one per line...",
+                            lines=10,
+                        )
+                        batch_rag = gr.Checkbox(label="Use RAG", value=True)
+                        batch_btn = gr.Button("Process All", variant="primary")
+                    
+                    with gr.Column():
+                        batch_output = gr.Code(label="Results", language="json", lines=20)
+                
+                batch_btn.click(
+                    batch_extract,
+                    inputs=[batch_input, batch_rag],
+                    outputs=batch_output
+                )
+            
+            # Tab 3: Analytics
+            with gr.Tab("📊 Analytics", id="analytics"):
+                with gr.Row():
+                    with gr.Column():
+                        analytics_input = gr.Code(
+                            label="Transaction Data (JSON)",
+                            language="json",
+                            lines=15,
+                            value="""[
+  {"amount": 2500, "type": "debit", "category": "food", "merchant": "Swiggy"},
+  {"amount": 15000, "type": "credit", "category": "transfer"},
+  {"amount": 999, "type": "debit", "category": "entertainment", "merchant": "Netflix"},
+  {"amount": 5000, "type": "debit", "category": "shopping", "merchant": "Amazon"}
+]"""
+                        )
+                        analyze_btn = gr.Button("Analyze", variant="primary")
+                    
+                    with gr.Column():
+                        analytics_output = gr.Markdown(label="Analysis Report")
+                
+                analyze_btn.click(
+                    analyze_transactions,
+                    inputs=analytics_input,
+                    outputs=analytics_output
+                )
+            
+            # Tab 4: Chat
+            with gr.Tab("💬 Chat", id="chat"):
+                chatbot = gr.Chatbot(
+                    label="FinEE Assistant",
+                    height=400,
+                    placeholder="Ask me to extract entities or analyze your transactions..."
+                )
+                
+                with gr.Row():
+                    chat_input = gr.Textbox(
+                        label="Message",
+                        placeholder="Type a message or paste a bank SMS...",
+                        scale=4
+                    )
+                    send_btn = gr.Button("Send", variant="primary", scale=1)
+                
+                def respond(message, history):
+                    if not message.strip():
+                        return "", history
+                    response = chat_response(message, history)
+                    history.append((message, response))
+                    return "", history
+                
+                send_btn.click(respond, [chat_input, chatbot], [chat_input, chatbot])
+                chat_input.submit(respond, [chat_input, chatbot], [chat_input, chatbot])
+            
+            # Tab 5: About
+            with gr.Tab("ℹ️ About", id="about"):
+                gr.Markdown("""
+                ## About FinEE
+                
+                **FinEE (Financial Entity Extractor)** is a specialized NLP tool for extracting 
+                structured information from Indian banking messages.
+                
+                ### Features
+                
+                - ✅ **Multi-Bank Support**: HDFC, ICICI, SBI, Axis, Kotak, and 20+ banks
+                - ✅ **All Transaction Types**: UPI, NEFT, IMPS, Credit Card, EMI
+                - ✅ **Multilingual**: English, Hindi, Tamil, Telugu, Bengali, Kannada
+                - ✅ **RAG Enhanced**: Context-aware extraction with merchant knowledge base
+                - ✅ **High Accuracy**: 95%+ on standard benchmarks
+                
+                ### Output Schema
+                
+                | Field | Type | Description |
+                |-------|------|-------------|
+                | amount | float | Transaction amount in INR |
+                | type | string | "debit" or "credit" |
+                | merchant | string | Business name (P2M) |
+                | beneficiary | string | Person name (P2P) |
+                | category | string | Transaction category |
+                | bank | string | Bank name |
+                | reference | string | UPI/NEFT reference |
+                | vpa | string | UPI VPA address |
+                
+                ### Links
+                
+                - 📦 **PyPI**: `pip install finee`
+                - 🤗 **Dataset**: [Ranjit0034/finee-dataset](https://huggingface.co/datasets/Ranjit0034/finee-dataset)
+                - 💻 **GitHub**: [Ranjitbehera0034/Finance-Entity-Extractor](https://github.com/Ranjitbehera0034/Finance-Entity-Extractor)
+                
+                ### Author
+                
+                Built by **Ranjit Behera** | MIT License
+                """)
+        
+        # Footer
+        gr.Markdown("""
+        ---
+        <center>
+        Made with ❤️ for the Indian fintech ecosystem
+        </center>
+        """)
+    
+    return demo
+
+
+# ============================================================================
+# MAIN
+# ============================================================================
+
+def launch(share: bool = False, port: int = 7860):
+    """Launch the Gradio app."""
+    demo = create_ui()
+    demo.launch(share=share, server_port=port)
+
+
+if __name__ == "__main__":
+    import argparse
+    
+    parser = argparse.ArgumentParser(description="FinEE Gradio UI")
+    parser.add_argument("--share", action="store_true", help="Create public link")
+    parser.add_argument("--port", type=int, default=7860, help="Port to run on")
+    
+    args = parser.parse_args()
+    launch(share=args.share, port=args.port)