Update app.py

app.py

@@ -1,129 +1,164 @@
 import gradio as gr
 import pandas as pd
-from transformers import pipeline
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 import torch
 
-class FinancialAnalyzer:
-    def __init__(self):
-        """Initialize models"""
-        # 1. Llama 2 for strategic analysis
-        self.strategic_analyzer = pipeline(
-            "text-generation",
-            model="meta-llama/Llama-2-7b-chat-hf",
-            device_map="auto"
-        )
-        
-        # 2. FinBERT for financial sentiment
-        self.financial_analyzer = pipeline(
+def load_models():
+    """Initialize models with proper error handling"""
+    models = {}
+    try:
+        # 1. Load FinBERT (always works, no approval needed)
+        models['finbert'] = pipeline(
             "text-classification",
             model="ProsusAI/finbert",
             return_all_scores=True
         )
+        print("✓ FinBERT loaded successfully")
         
-        # 3. Falcon for recommendations
-        self.recommendation_generator = pipeline(
+        # 2. Try loading Llama 2 (requires approval)
+        try:
+            models['llama2'] = pipeline(
+                "text-generation",
+                model="meta-llama/Llama-2-7b-chat-hf",
+                torch_dtype=torch.float16,
+                device_map="auto"
+            )
+            print("✓ Llama 2 loaded successfully")
+        except Exception as e:
+            print(f"⚠️ Llama 2 not available: {str(e)}")
+            # Fallback to Falcon
+            models['llama2'] = pipeline(
+                "text-generation",
+                model="tiiuae/falcon-7b",
+                torch_dtype=torch.float16,
+                device_map="auto"
+            )
+            print("✓ Using Falcon as fallback for analysis")
+        
+        # 3. Load Falcon (always works, no approval needed)
+        models['falcon'] = pipeline(
             "text-generation",
-            model="tiiuae/falcon-7b-instruct",
+            model="tiiuae/falcon-7b",
+            torch_dtype=torch.float16,
             device_map="auto"
         )
-    
-    def generate_strategic_analysis(self, financial_data):
-        """Generate strategic analysis using Llama 2"""
-        prompt = f"""[INST] As a senior financial analyst, analyze these financial statements:
-
-        Financial Data:
-        {financial_data}
-
-        Provide:
-        1. Business Health Assessment
-        2. Key Strategic Insights
-        3. Market Position Analysis
-        4. Growth Opportunities
-        5. Risk Factors [/INST]"""
+        print("✓ Falcon loaded successfully")
         
-        response = self.strategic_analyzer(
-            prompt,
-            max_length=1000,
-            temperature=0.7
-        )
-        return response[0]['generated_text']
-    
-    def analyze_sentiment(self, text):
-        """Analyze financial sentiment using FinBERT"""
-        return self.financial_analyzer(text)
-    
-    def generate_recommendations(self, analysis):
-        """Generate recommendations using Falcon"""
-        prompt = f"""Based on this financial analysis:
-        {analysis}
-        
-        Provide specific, actionable recommendations covering:
-        1. Strategic Initiatives
-        2. Operational Improvements
-        3. Financial Management
-        4. Risk Mitigation
-        5. Growth Strategy"""
-        
-        response = self.recommendation_generator(
-            prompt,
-            max_length=800,
-            temperature=0.6
-        )
-        return response[0]['generated_text']
+        return models
+    except Exception as e:
+        print(f"Error loading models: {str(e)}")
+        return None
 
 def analyze_financial_statements(income_statement, balance_sheet):
-    """Main analysis function"""
+    """Analyze financial statements using AI models"""
     try:
-        # Read files
+        # Load data
         income_df = pd.read_csv(income_statement.name)
         balance_df = pd.read_csv(balance_sheet.name)
         
-        # Initialize analyzer
-        analyzer = FinancialAnalyzer()
-        
-        # Prepare financial data
-        financial_data = f"""
-        Income Statement Summary:
-        {income_df.to_string()}
-        
-        Balance Sheet Summary:
-        {balance_df.to_string()}
-        """
-        
-        # Generate strategic analysis
-        strategic_analysis = analyzer.generate_strategic_analysis(financial_data)
+        # Prepare financial data summary
+        financial_data = prepare_financial_summary(income_df, balance_df)
         
-        # Analyze sentiment
-        sentiment = analyzer.analyze_sentiment(strategic_analysis)
+        # Load models
+        models = load_models()
+        if not models:
+            return "Error: Could not load analysis models."
         
-        # Generate recommendations
-        recommendations = analyzer.generate_recommendations(strategic_analysis)
+        # Generate analyses
+        analysis = generate_analysis(models, financial_data)
         
-        # Format output
-        output = format_results(strategic_analysis, sentiment, recommendations)
-        
-        return output
+        # Format and return results
+        return format_results(analysis)
         
     except Exception as e:
         return f"Error analyzing files: {str(e)}"
 
-def format_results(analysis, sentiment, recommendations):
-    """Format analysis results"""
+def prepare_financial_summary(income_df, balance_df):
+    """Prepare financial data summary"""
+    latest_year = income_df['Period'].iloc[-1]
+    previous_year = income_df['Period'].iloc[-2]
+    
+    summary = f"""Financial Analysis for {latest_year}:
+
+Income Statement Summary:
+- Revenue: {income_df['Revenue'].iloc[-1]}
+- Gross Profit: {income_df['Gross Profit'].iloc[-1]}
+- Net Income: {income_df['Net Income'].iloc[-1]}
+- YoY Revenue Growth: {(income_df['Revenue'].iloc[-1] / income_df['Revenue'].iloc[-2] - 1) * 100:.1f}%
+- YoY Net Income Growth: {(income_df['Net Income'].iloc[-1] / income_df['Net Income'].iloc[-2] - 1) * 100:.1f}%
+
+Balance Sheet Summary:
+- Total Assets: {balance_df['Total Assets'].iloc[-1]}
+- Total Liabilities: {balance_df['Total Liabilities'].iloc[-1]}
+- Shareholder's Equity: {balance_df["Shareholder's Equity"].iloc[-1]}
+- Current Ratio: {balance_df['Total current assets'].iloc[-1] / balance_df['Total current liabilities'].iloc[-1]:.2f}"""
+    
+    return summary
+
+def generate_analysis(models, financial_data):
+    """Generate comprehensive analysis using all models"""
+    
+    # 1. Strategic Analysis (Llama 2 or Falcon)
+    strategy_prompt = f"""[INST] As a senior financial analyst, analyze these financial metrics and provide strategic insights:
+
+{financial_data}
+
+Provide analysis covering:
+1. Financial Health Assessment
+2. Key Performance Indicators
+3. Growth Analysis
+4. Risk Factors
+5. Strategic Position [/INST]"""
+
+    strategic_analysis = models['llama2'](
+        strategy_prompt,
+        max_length=1000,
+        temperature=0.7
+    )[0]['generated_text']
+
+    # 2. Financial Sentiment (FinBERT)
+    sentiment = models['finbert'](strategic_analysis)
+
+    # 3. Recommendations (Falcon)
+    recommendations_prompt = f"""Based on this financial analysis:
+{strategic_analysis}
+
+Provide specific, actionable recommendations for:
+1. Immediate Actions (0-6 months)
+2. Short-term Strategy (6-12 months)
+3. Medium-term Initiatives (1-2 years)
+4. Risk Mitigation
+5. Growth Opportunities"""
+
+    recommendations = models['falcon'](
+        recommendations_prompt,
+        max_length=800,
+        temperature=0.6
+    )[0]['generated_text']
+
+    return {
+        'strategic_analysis': strategic_analysis,
+        'sentiment': sentiment,
+        'recommendations': recommendations
+    }
+
+def format_results(analysis):
+    """Format analysis results into readable report"""
     output = "# Financial Analysis Report\n\n"
     
     # Strategic Analysis
     output += "## Strategic Analysis\n\n"
-    output += analysis + "\n\n"
+    output += analysis['strategic_analysis'].split('[/INST]')[-1].strip() + "\n\n"
     
-    # Sentiment Analysis
+    # Market Sentiment
     output += "## Market Sentiment\n\n"
-    for score in sentiment[0]:
+    for score in analysis['sentiment'][0]:
         output += f"- {score['label']}: {score['score']:.2%}\n"
     output += "\n"
     
     # Recommendations
     output += "## Strategic Recommendations\n\n"
-    output += recommendations
+    output += analysis['recommendations']
     
     return output
 
@@ -137,9 +172,11 @@ iface = gr.Interface(
     outputs=gr.Markdown(),
     title="AI-Powered Financial Statement Analysis",
     description="""Upload your financial statements for comprehensive analysis using:
-    - Llama 2: Strategic Analysis
+    - Llama 2 / Falcon: Strategic Analysis
     - FinBERT: Financial Sentiment Analysis
-    - Falcon: Strategic Recommendations""",
+    - Falcon: Strategic Recommendations
+    
+    Note: The system will automatically use available models based on access permissions.""",
     examples=[
         [
             "OFINTECH-Income Statement-template.csv",