app.py

import gradio as gr
import pandas as pd
import numpy as np
import json
import re
from transformers import AutoTokenizer, AutoModelForCausalLM, AutoModelForSequenceClassification
import torch

class FinancialAnalyzer:
    def __init__(self):
        print("Initializing Financial Analyzer...")
        self.initialize_models()

    def initialize_models(self):
        print("Loading models...")
        self.tiny_tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
        self.tiny_model = AutoModelForCausalLM.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
        
        self.finbert_tokenizer = AutoTokenizer.from_pretrained("ProsusAI/finbert")
        self.finbert_model = AutoModelForSequenceClassification.from_pretrained("ProsusAI/finbert")
        print("Models loaded successfully!")

    def parse_markdown_table(self, markdown_content):
        """Parse markdown table into pandas DataFrame"""
        # Split content into lines
        lines = markdown_content.strip().split('\n')
        
        # Find table start (line with |)
        table_lines = []
        headers = None
        current_table = []
        
        for line in lines:
            if '|' in line:
                # Skip separator lines (contains ---)
                if '-|-' in line:
                    continue
                # Clean and split the line
                row = [cell.strip() for cell in line.split('|')[1:-1]]
                if headers is None:
                    headers = row
                else:
                    current_table.append(row)
                    
        # Create DataFrame
        df = pd.DataFrame(current_table, columns=headers)
        return df

    def extract_financial_data(self, markdown_content):
        """Convert markdown content to a structured text format"""
        # Remove markdown formatting
        clean_text = markdown_content.replace('#', '').replace('*', '')
        
        # Extract tables
        tables = {}
        current_section = "General"
        
        for line in clean_text.split('\n'):
            if line.strip() and not line.startswith('|'):
                current_section = line.strip()
            elif '|' in line:
                if current_section not in tables:
                    tables[current_section] = []
                tables[current_section].append(line)
                
        # Convert to text format
        structured_text = []
        for section, content in tables.items():
            structured_text.append(f"\n{section}:")
            if content:
                df = self.parse_markdown_table('\n'.join(content))
                structured_text.append(df.to_string())
                
        return '\n'.join(structured_text)

    def analyze_financials(self, balance_sheet_file, income_stmt_file):
        """Main analysis function"""
        try:
            # Read markdown files
            with open(balance_sheet_file, 'r') as f:
                balance_sheet_content = f.read()
            with open(income_stmt_file, 'r') as f:
                income_stmt_content = f.read()

            # Convert to structured text
            structured_balance = self.extract_financial_data(balance_sheet_content)
            structured_income = self.extract_financial_data(income_stmt_content)

            # Create analysis prompt
            prompt = f"""<human>Please analyze these financial statements and provide detailed insights:

Financial Statements Analysis (2021-2025)

Balance Sheet Summary:
{structured_balance}

Income Statement Summary:
{structured_income}

Please provide a detailed analysis including:
1. Financial Health Assessment
   - Liquidity position
   - Capital structure
   - Asset efficiency

2. Profitability Analysis
   - Revenue trends
   - Cost management
   - Profit margins

3. Growth Analysis
   - Year-over-year growth rates
   - Market position
   - Future growth potential

4. Risk Assessment
   - Operating risks
   - Financial risks
   - Strategic risks

5. Recommendations
   - Short-term actions
   - Medium-term strategy
   - Long-term planning

6. Future Outlook
   - Market conditions
   - Company positioning
   - Growth opportunities</human>"""

            # Generate AI analysis
            inputs = self.tiny_tokenizer(prompt, return_tensors="pt", truncation=True)
            outputs = self.tiny_model.generate(
                inputs["input_ids"],
                max_length=2048,
                temperature=0.7,
                top_p=0.95,
                do_sample=True
            )
            analysis = self.tiny_tokenizer.decode(outputs[0], skip_special_tokens=True)

            # Generate sentiment
            sentiment = self.analyze_sentiment(structured_balance + structured_income)

            # Compile results
            results = {
                "Financial Analysis": analysis,
                "Sentiment Analysis": sentiment,
                "Analysis Period": "2021-2025",
                "Note": "All values in millions ($M)"
            }

            return json.dumps(results, indent=2)

        except Exception as e:
            return f"Error in analysis: {str(e)}\n\nDetails: {type(e).__name__}"

    def analyze_sentiment(self, text):
        inputs = self.finbert_tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
        outputs = self.finbert_model(**inputs)
        probs = torch.nn.functional.softmax(outputs.logits, dim=1)
        sentiment_labels = ['negative', 'neutral', 'positive']
        
        return {
            'sentiment': sentiment_labels[probs.argmax().item()],
            'confidence': f"{probs.max().item():.2f}"
        }

def create_interface():
    analyzer = FinancialAnalyzer()
    
    iface = gr.Interface(
        fn=analyzer.analyze_financials,
        inputs=[
            gr.File(label="Balance Sheet (Markdown)", type="filepath"),
            gr.File(label="Income Statement (Markdown)", type="filepath")
        ],
        outputs=gr.Textbox(label="Analysis Results", lines=25),
        title="Financial Statement Analyzer",
        description="Upload financial statements in Markdown format for comprehensive AI-powered analysis."
    )
    
    return iface

if __name__ == "__main__":
    iface = create_interface()
    iface.launch()