import os import gradio as gr from langchain_groq import ChatGroq from langchain_community.tools import DuckDuckGoSearchRun from langchain_community.tools import TavilySearchResults # Added this import from langgraph.prebuilt import create_react_agent # --- 1. SECURE KEY --- # --- 1. Fetch the Secrets --- groq_api_key = os.environ.get("GROQ_API_KEY") tavily_api_key = os.environ.get("TAVILY_API_KEY") # --- 2. Security Check (Governance) --- if not groq_api_key: raise ValueError("❌ CRITICAL ERROR: 'GROQ_API_KEY' not found. Please add it to your Space Secrets.") if not tavily_api_key: # Optional: You can just warn if Tavily is not critical print("⚠️ Warning: 'TAVILY_API_KEY' not found. Search may not work.") # --- 2. INITIALIZE THE BRAIN (Groq + Llama 3) --- try: llm = ChatGroq( api_key=os.environ["GROQ_API_KEY"], model="llama-3.3-70b-versatile", temperature=0 ) except Exception as e: raise ValueError(f"❌ Failed to connect to Groq: {e}") # --- 3. INITIALIZE TOOLS --- # Requires 'tavily-python' in requirements.txt # If you don't have a key, this line will fail. try: search = TavilySearchResults(max_results=2) tools = [search] except Exception: raise ValueError("❌ TAVILY_API_KEY missing. Please add it to Secrets or switch back to DuckDuckGo.") # search = DuckDuckGoSearchRun() # tools = [search] # --- 4. CREATE THE AGENT --- agent_executor = create_react_agent(llm, tools) # --- 5. DEFINE THE INTERACTION --- def chat_function(message, history): """ Wrapper function for Gradio. Takes user message, runs agent, returns answer. """ print(f"\n[GOVERNANCE LOG] User Query: {message}") # Audit Log try: # Run the agent logic response = agent_executor.invoke( {"messages": [("user", message)]}, config={"recursion_limit": 10} # Cost Control ) # Extract the final answer text final_answer = response['messages'][-1].content # Log the tool usage to the internal console (Governance) for msg in response['messages']: if msg.type == 'tool': print(f"[GOVERNANCE LOG] Tool Used: {str(msg.content)[:100]}...") return final_answer except Exception as e: error_msg = f"❌ System Error: {str(e)}" print(f"[GOVERNANCE LOG] {error_msg}") return error_msg # --- 6. BUILD THE UI --- demo = gr.ChatInterface( fn=chat_function, title="AI Financial Analyst", description="**Governance-Focused AI Agent** | Powered by Llama 3.3 (Groq) & LangGraph.\n\nAsk me about stock prices, market trends, or financial concepts.", examples=[ "What is the current price of NVDA?", "Compare the P/E ratio of Ford and GM.", "What are the latest risks for Tesla?" ], ) # --- 7. LAUNCH --- if __name__ == "__main__": demo.launch()