Enhance financial agent capabilities: Integrate OpenAI function calling for improved extraction of stock symbols, budget data, and portfolio information from natural language inputs. Update example queries in app.py and README.md to reflect new features and support for additional stocks.

README.md

@@ -19,7 +19,7 @@ short_description: Financial analysis, investments, budget planning, and more.
 
 # AI Financial Advisory Agent
 
-An intelligent financial advisory agent powered by OpenAI's GPT-4.1 and specialized financial tools. This agent provides comprehensive financial analysis, investment recommendations, budget planning, and market insights through an intuitive web interface.
+An intelligent financial advisory agent powered by OpenAI's GPT-4.1 and specialized financial tools. This agent provides comprehensive financial analysis, investment recommendations, budget planning, and market insights through an intuitive web interface with **advanced natural language processing**.
 
 **Watch the demo video:** [Financial Advisory Agent Demo (Hugging Face Hackathon)](https://youtu.be/vGOSAXyrQgM)
 
@@ -29,16 +29,18 @@ An intelligent financial advisory agent powered by OpenAI's GPT-4.1 and speciali
 ## 🚀 Features
 
 ### Core Financial Tools
-- **📊 Investment Analyzer**: Real-time stock analysis with technical indicators, risk assessment, and investment recommendations
-- **💰 Budget Planner**: Personalized budget creation using the 50/30/20 rule with optimization suggestions
-- **📈 Portfolio Analyzer**: Portfolio diversification analysis and performance optimization
-- **📰 Market Trends Analyzer**: Real-time market news, sector analysis, and economic insights
+- **📊 Investment Analyzer**: Analyze ANY stock using natural language - "Tell me about NVIDIA" or "How is Tesla doing?"
+- **💰 Budget Planner**: Create budgets from conversational input - "I make 6k monthly, rent is 1800, food 600"
+- **📈 Portfolio Analyzer**: Analyze portfolios from natural descriptions - "My portfolio: Apple 100 shares, Google 50 shares"
+- **📰 Market Trends Analyzer**: Get optimized market research from simple queries - "What's happening in tech?"
 
 ### Advanced Features
+- **🧠 OpenAI Function Calling**: All tools use intelligent input extraction from natural language
 - **🎯 Smart Response Type Detection**: Automatically detects if users want short summaries or detailed reports
 - **⚡ Real-time LLM Streaming**: Live token-by-token response generation from OpenAI API
 - **📋 Interactive Tool Results**: Collapsible sections for clean data presentation
 - **🔍 Comprehensive Data Analysis**: Enhanced insights for all financial data
+- **🌐 Universal Stock Support**: Supports analysis of ANY publicly traded stock, not just predefined ones
 
 ## 🛠️ Installation
 
@@ -83,37 +85,48 @@ The system automatically detects your preference:
 
 - **Frontend**: Gradio web interface with real-time streaming
 - **Backend**: LangChain agents with OpenAI GPT-4.1
+- **AI Integration**: OpenAI function calling for intelligent input processing across all tools
 - **Data Sources**: 
-  - Yahoo Finance API for stock data
+  - Yahoo Finance API for stock data (supports ALL publicly traded stocks)
   - Tavily Search API for market news
   - Real-time financial calculations
 - **Streaming**: Direct LLM token streaming from OpenAI API
-- **Tools**: Specialized financial analysis functions
+- **Tools**: Specialized financial analysis functions with natural language processing
 
 ## 📊 Tool Capabilities
 
-### Investment Analyzer
+### Investment Analyzer 🧠
+- **Natural Language Input**: "Tell me about NVIDIA", "How is Tesla doing?", "Analyze AMZN"
+- **Universal Stock Support**: Works with ANY publicly traded stock symbol
+- **OpenAI-Powered Extraction**: Intelligently extracts stock symbols from company names
 - Real-time stock prices and historical data
 - Technical indicators (RSI, MACD, Bollinger Bands, Moving Averages)
 - Risk assessment (volatility, VaR, beta analysis)
 - Fundamental analysis (P/E, P/B ratios, dividend yield)
 - Buy/Hold/Sell recommendations with confidence scores
 
-### Budget Planner
+### Budget Planner 🧠
+- **Conversational Input**: "I make 6k monthly, rent 1800, food 600", "Help budget my 4500 salary"
+- **OpenAI Data Extraction**: Automatically parses income and expenses from natural language
 - 50/30/20 rule implementation
 - Emergency fund calculations
 - Debt-to-income ratio analysis
 - Savings optimization recommendations
 - Expense category warnings
 
-### Market Trends Analyzer
+### Market Trends Analyzer 🧠
+- **Query Optimization**: Converts "What's happening in tech?" to optimized search queries
+- **OpenAI-Enhanced Search**: Automatically improves search terms for better results
 - Real-time market news via Tavily Search
 - Major index tracking (S&P 500, NASDAQ)
 - Market sentiment analysis
 - Key theme extraction
 - Economic trend identification
 
-### Portfolio Analyzer
+### Portfolio Analyzer 🧠
+- **Natural Portfolio Input**: "My portfolio: Apple 100 shares, Google 50 shares"
+- **Smart Format Detection**: Handles both share counts and percentages automatically
+- **OpenAI JSON Conversion**: Converts natural descriptions to structured data
 - Diversification scoring
 - Sector allocation analysis
 - Concentration risk assessment
@@ -123,6 +136,17 @@ The system automatically detects your preference:
 
 ## 🚀 Recent Updates
 
+### Latest (v2.0) - OpenAI Function Calling Integration 🧠
+- ✅ **Universal Stock Support**: Removed hardcoded stock limitations - now supports ANY publicly traded stock
+- ✅ **OpenAI-Powered Input Extraction**: All tools now use OpenAI function calling for intelligent data extraction
+- ✅ **Natural Language Processing**: 
+  - Investment: "Tell me about NVIDIA" → Automatically extracts "NVDA"
+  - Budget: "I make 6k monthly, rent 1800" → Extracts structured JSON
+  - Portfolio: "Apple 100 shares, Google 50 shares" → Converts to holdings format
+  - Market: "What's happening in tech?" → Optimizes to targeted search query
+- ✅ **Enhanced Example Queries**: Updated with NVIDIA and natural language examples
+
+### Previous Updates
 - ✅ **Smart Response Detection**: Automatically detects if users want short or detailed responses
 - ✅ **Real LLM Streaming**: Genuine token-by-token streaming from OpenAI API
 - ✅ **Improved Data Analysis**: Comprehensive insights
@@ -134,12 +158,30 @@ The system automatically detects your preference:
 - **OpenAI API Key**: For language model
 - **Tavily API Key**: For real-time market news and trends
 
-## Issues to Resolve
-The problem is in the current flow:
-1. Agent executor runs all tools
-2. We collect ALL results
-3. Then display everything at once
-4. Then stream the final response
+## 💡 Example Natural Language Queries
+
+### Investment Analysis
+- "Tell me about NVIDIA stocks"
+- "How is Tesla performing?"
+- "Should I invest in Apple?"
+- "Analyze AMZN for me"
+- "What's the latest on Microsoft stock?"
+
+### Budget Planning
+- "I make $6000 monthly, rent is $1800, food $600, help me budget"
+- "Help budget my $4500 salary, utilities cost $150"
+- "Create a budget: income 5k, rent 1500, groceries 400"
+
+### Portfolio Analysis
+- "My portfolio: Apple 100 shares, Google 50 shares, Microsoft 25 shares"
+- "I have 40% AAPL, 30% MSFT, 30% TSLA"
+- "Analyze portfolio with Tesla 200 shares and Amazon 25%"
+
+### Market Trends
+- "What's happening in tech stocks?"
+- "Tell me about today's market"
+- "How is the crypto market doing?"
+- "What are the latest NVIDIA trends?"
 
 ## 🤝 Contributing
 

agents/financial_agent.py

@@ -76,46 +76,166 @@ When a user asks a question:
     def _prepare_tool_input(self, message: str, tool_name: str) -> str:
         """Prepare input for specific tools based on the message"""
         if tool_name == "investment_analyzer":
-            # Extract stock symbols
+            # Use OpenAI to extract stock symbol from natural language
+            extraction_prompt = f"""Extract the stock symbol from this message: "{message}"
+
+If the user mentions a company name, return the corresponding stock ticker symbol.
+If they mention a ticker symbol directly, return that symbol.
+If no stock or company is mentioned, return "UNKNOWN".
+
+Examples:
+- "Tell me about NVIDIA" -> "NVDA"
+- "Analyze AAPL stock" -> "AAPL" 
+- "How is Tesla doing?" -> "TSLA"
+- "What about Microsoft stock?" -> "MSFT"
+
+Return only the stock symbol, nothing else."""
+
+            try:
+                response = self.llm.invoke([
+                    SystemMessage(content="You are a stock symbol extraction assistant. Return only the ticker symbol."),
+                    HumanMessage(content=extraction_prompt)
+                ])
+                
+                extracted_symbol = response.content.strip().upper()
+                if extracted_symbol and extracted_symbol != "UNKNOWN":
+                    return extracted_symbol
+                
+            except Exception:
+                pass
+            
+            # Fallback to regex if LLM fails
             symbols = re.findall(r"\b[A-Z]{2,5}\b", message)
-            if symbols:
-                return symbols[0]
-            return "AAPL"  # Default
+            return symbols[0] if symbols else ""
 
         elif tool_name == "budget_planner":
-            # Try to extract income and expenses from message
-            income_match = re.search(
-                r"\$?(\d+(?:,\d{3})*(?:\.\d{2})?)\s*(?:monthly\s*)?income",
-                message,
-                re.I,
-            )
-            income = (
-                float(income_match.group(1).replace(",", "")) if income_match else 5000
-            )
-
-            # Extract expenses
+            # Use OpenAI to extract budget information from natural language
+            extraction_prompt = f"""Extract budget information from this message: "{message}"
+
+Extract:
+1. Monthly income (if mentioned)
+2. Expenses by category (rent, food, utilities, transportation, etc.)
+
+Return as JSON format:
+{{"income": 5000, "expenses": {{"rent": 1500, "food": 500, "utilities": 200}}}}
+
+If income is not mentioned, use 5000 as default.
+If no expenses are mentioned, return empty expenses object.
+
+Examples:
+- "I make $6000 monthly, rent is $1800, food $600" -> {{"income": 6000, "expenses": {{"rent": 1800, "food": 600}}}}
+- "Help with budget, income 4500, utilities 150" -> {{"income": 4500, "expenses": {{"utilities": 150}}}}
+
+Return only valid JSON, nothing else."""
+
+            try:
+                response = self.llm.invoke([
+                    SystemMessage(content="You are a budget data extraction assistant. Return only valid JSON."),
+                    HumanMessage(content=extraction_prompt)
+                ])
+                
+                # Try to parse the JSON response
+                extracted_data = response.content.strip()
+                # Remove any markdown formatting
+                if extracted_data.startswith("```"):
+                    extracted_data = extracted_data.split("\n")[1:-1]
+                    extracted_data = "\n".join(extracted_data)
+                
+                # Validate JSON
+                json.loads(extracted_data)
+                return extracted_data
+                
+            except Exception:
+                pass
+            
+            # Fallback to regex extraction
+            income_match = re.search(r"\$?(\d+(?:,\d{3})*(?:\.\d{2})?)\s*(?:monthly\s*)?income", message, re.I)
+            income = float(income_match.group(1).replace(",", "")) if income_match else 5000
+            
             expenses = {}
             expense_patterns = [
                 (r"rent:?\s*\$?(\d+(?:,\d{3})*(?:\.\d{2})?)", "rent"),
                 (r"food:?\s*\$?(\d+(?:,\d{3})*(?:\.\d{2})?)", "food"),
                 (r"utilities:?\s*\$?(\d+(?:,\d{3})*(?:\.\d{2})?)", "utilities"),
-                (
-                    r"transportation:?\s*\$?(\d+(?:,\d{3})*(?:\.\d{2})?)",
-                    "transportation",
-                ),
+                (r"transportation:?\s*\$?(\d+(?:,\d{3})*(?:\.\d{2})?)", "transportation"),
             ]
-
+            
             for pattern, category in expense_patterns:
                 match = re.search(pattern, message, re.I)
                 if match:
                     expenses[category] = float(match.group(1).replace(",", ""))
-
+            
             return json.dumps({"income": income, "expenses": expenses})
 
         elif tool_name == "portfolio_analyzer":
+            # Use OpenAI to extract portfolio information from natural language
+            extraction_prompt = f"""Extract portfolio holdings from this message: "{message}"
+
+Convert the portfolio information to JSON format with holdings array.
+Each holding should have symbol and either shares or percentage.
+
+Return format:
+{{"holdings": [{{"symbol": "AAPL", "shares": 100}}, {{"symbol": "GOOGL", "percentage": 30}}]}}
+
+Examples:
+- "My portfolio: AAPL 100 shares, GOOGL 50 shares" -> {{"holdings": [{{"symbol": "AAPL", "shares": 100}}, {{"symbol": "GOOGL", "shares": 50}}]}}
+- "I have 40% AAPL, 30% MSFT, 30% TSLA" -> {{"holdings": [{{"symbol": "AAPL", "percentage": 40}}, {{"symbol": "MSFT", "percentage": 30}}, {{"symbol": "TSLA", "percentage": 30}}]}}
+- "Portfolio with Apple 200 shares and Microsoft 25%" -> {{"holdings": [{{"symbol": "AAPL", "shares": 200}}, {{"symbol": "MSFT", "percentage": 25}}]}}
+
+If no clear portfolio data is found, return: {{"holdings": []}}
+
+Return only valid JSON, nothing else."""
+
+            try:
+                response = self.llm.invoke([
+                    SystemMessage(content="You are a portfolio data extraction assistant. Return only valid JSON with holdings array."),
+                    HumanMessage(content=extraction_prompt)
+                ])
+                
+                # Try to parse the JSON response
+                extracted_data = response.content.strip()
+                # Remove any markdown formatting
+                if extracted_data.startswith("```"):
+                    extracted_data = extracted_data.split("\n")[1:-1]
+                    extracted_data = "\n".join(extracted_data)
+                
+                # Validate JSON
+                json.loads(extracted_data)
+                return extracted_data
+                
+            except Exception:
+                pass
+            
+            # Fallback to returning the original message
             return message
 
         elif tool_name == "market_trends":
+            # Use OpenAI to extract and refine market research query
+            extraction_prompt = f"""Convert this user message into an optimized market research query: "{message}"
+
+Create a focused search query that will get the best market trends and financial news results.
+
+Examples:
+- "What's happening in tech stocks?" -> "technology stocks market trends latest news 2025"
+- "Tell me about the market today" -> "stock market trends today financial news latest"
+- "How is the crypto market?" -> "cryptocurrency market trends bitcoin ethereum latest news"
+- "What about NVIDIA trends?" -> "NVIDIA NVDA stock market trends analysis latest news"
+
+Return only the optimized search query, nothing else."""
+
+            try:
+                response = self.llm.invoke([
+                    SystemMessage(content="You are a search query optimization assistant. Return only the optimized search query."),
+                    HumanMessage(content=extraction_prompt)
+                ])
+                
+                optimized_query = response.content.strip()
+                return optimized_query if optimized_query else message
+                
+            except Exception:
+                pass
+            
+            # Fallback to returning the original message
             return message
 
         return message

app.py

@@ -477,10 +477,16 @@ def determine_intended_tool(message):
             "sell",
             "analyze",
             "AAPL",
-            "GOOGL",
+            "GOOGL", 
             "TSLA",
+            "NVDA",
+            "NVIDIA",
+            "MSFT",
+            "AMZN",
+            "META",
             "share",
             "equity",
+            "ticker",
         ],
         "portfolio_analyzer": [
             "portfolio",
@@ -821,8 +827,9 @@ with gr.Blocks(theme=gr.themes.Base(), title="Financial Advisory Agent") as demo
 
     # Example queries
     example_queries = [
-        "Analyze AAPL stock and tell me if it's a good investment",
-        "Help me create a budget with $5000 monthly income and expenses: rent $1500, food $500, utilities $200",
+        "Analyze NVDA stock and tell me if it's a good investment",
+        "Tell me more about NVIDIA stocks",
+        "Help me create a budget with $5000 monthly income and expenses: rent $1500, food $500, utilities $200", 
         "What are the latest market trends in tech stocks?",
         "Analyze my portfolio: {'holdings': [{'symbol': 'AAPL', 'shares': 100}, {'symbol': 'GOOGL', 'shares': 50}]}",
     ]