import os import gradio as gr from groq import Groq from datetime import datetime import time class RealTimeFactChecker: def __init__(self): self.client = None self.model_options = ["compound-beta", "compound-beta-mini"] def initialize_client(self, api_key): """Initialize Groq client with API key""" try: self.client = Groq(api_key=api_key) return True, "✅ API Key validated successfully!" except Exception as e: return False, f"❌ Error initializing client: {str(e)}" def get_system_prompt(self): """Get the system prompt for consistent behavior""" return """You are a Real-time Fact Checker and News Agent. Your primary role is to provide accurate, up-to-date information by leveraging web search when needed. CORE RESPONSIBILITIES: 1. **Fact Verification**: Always verify claims with current, reliable sources 2. **Real-time Information**: Use web search for any information that changes frequently (news, stocks, weather, current events) 3. **Source Transparency**: When using web search, mention the sources or indicate that you've searched for current information 4. **Accuracy First**: If information is uncertain or conflicting, acknowledge this clearly RESPONSE GUIDELINES: - **Structure**: Start with a clear, direct answer, then provide supporting details - **Recency**: Always prioritize the most recent, reliable information - **Clarity**: Use clear, professional language while remaining accessible - **Completeness**: Provide comprehensive answers but stay focused on the query - **Source Awareness**: When you've searched for information, briefly indicate this (e.g., "Based on current reports..." or "Recent data shows...") WHEN TO SEARCH: - Breaking news or current events - Stock prices, market data, or financial information - Weather conditions or forecasts - Recent scientific discoveries or research - Current political developments - Real-time statistics or data - Verification of recent claims or rumors RESPONSE FORMAT: - Lead with key facts - Include relevant context - Mention timeframe when relevant (e.g., "as of today", "this week") - If multiple sources conflict, acknowledge this - End with a clear summary for complex topics Remember: Your goal is to be the most reliable, up-to-date source of information possible.""" def query_compound_model(self, query, model, temperature=0.7): """Query the compound model and return response with tool execution info""" if not self.client: return "❌ Please set a valid API key first.", None, None try: start_time = time.time() system_prompt = self.get_system_prompt() chat_completion = self.client.chat.completions.create( messages=[ { "role": "system", "content": system_prompt }, { "role": "user", "content": query, } ], model=model, temperature=temperature, max_tokens=1500 ) end_time = time.time() response_time = round(end_time - start_time, 2) response_content = chat_completion.choices[0].message.content executed_tools = getattr(chat_completion.choices[0].message, 'executed_tools', None) tool_info = self.format_tool_info(executed_tools) return response_content, tool_info, response_time except Exception as e: return f"❌ Error querying model: {str(e)}", None, None def format_tool_info(self, executed_tools): """Format executed tools information for display""" if not executed_tools: return "🔍 Tools Used: None (Used existing knowledge)" tool_info = "🔍 Tools Used:\n" for i, tool in enumerate(executed_tools, 1): try: if hasattr(tool, 'name'): tool_name = tool.name elif hasattr(tool, 'tool_name'): tool_name = tool.tool_name elif isinstance(tool, dict): tool_name = tool.get('name', 'Unknown') else: tool_name = str(tool) tool_info += f"{i}. {tool_name}\n" if hasattr(tool, 'parameters'): params = tool.parameters if isinstance(params, dict): for key, value in params.items(): tool_info += f" - {key}: {value}\n" elif hasattr(tool, 'input'): tool_info += f" - Input: {tool.input}\n" except Exception as e: tool_info += f"{i}. Tool {i} (Error parsing details)\n" return tool_info def create_interface(): fact_checker = RealTimeFactChecker() def validate_api_key(api_key): if not api_key or api_key.strip() == "": return "❌ Please enter a valid API key", False success, message = fact_checker.initialize_client(api_key.strip()) return message, success def process_query(query, model, temperature, api_key): if not api_key or api_key.strip() == "": return "❌ Please set your API key first", "", "" if not query or query.strip() == "": return "❌ Please enter a query", "", "" if not fact_checker.client: success, message = fact_checker.initialize_client(api_key.strip()) if not success: return message, "", "" response, tool_info, response_time = fact_checker.query_compound_model( query.strip(), model, temperature ) timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S") formatted_response = f"**Query:** {query}\n\n**Response:**\n{response}\n\n---\n*Generated at {timestamp} in {response_time}s*" return formatted_response, tool_info or "", f"Response time: {response_time}s" with gr.Blocks(title="Real-time Fact Checker & News Agent", theme=gr.themes.Ocean()) as demo: gr.Markdown("# Real-time Fact Checker & News Agent") gr.Markdown("Powered by Groq's Compound Models with Built-in Web Search") with gr.Row(): with gr.Column(): api_key_input = gr.Textbox( label="Groq API Key", placeholder="Enter your Groq API key here...", type="password", info="Get your free API key from https://console.groq.com/" ) api_status = gr.Textbox( label="Status", value="Please enter your API key", interactive=False ) validate_btn = gr.Button("Validate API Key") query_input = gr.Textbox( label="Query", placeholder="e.g., What are the latest AI developments today?", lines=4 ) with gr.Row(): model_choice = gr.Dropdown( choices=fact_checker.model_options, value="compound-beta", label="Model", info="compound-beta: More capable | compound-beta-mini: Faster" ) temperature = gr.Slider( minimum=0.0, maximum=1.0, value=0.7, step=0.1, label="Temperature", info="Higher = more creative, Lower = more focused" ) with gr.Row(): submit_btn = gr.Button("Get Real-time Information") clear_btn = gr.Button("Clear") with gr.Column(): response_output = gr.Markdown( label="Response", value="*Your response will appear here...*" ) tool_info_output = gr.Markdown( label="Tool Execution Info", value="*Tool execution details will appear here...*" ) performance_output = gr.Textbox( label="Performance", value="", interactive=False ) validate_btn.click( fn=validate_api_key, inputs=[api_key_input], outputs=[api_status, gr.State()] ) submit_btn.click( fn=process_query, inputs=[query_input, model_choice, temperature, api_key_input], outputs=[response_output, tool_info_output, performance_output] ) clear_btn.click( fn=lambda: ("", "*Your response will appear here...*", "*Tool execution details will appear here...*", ""), outputs=[query_input, response_output, tool_info_output, performance_output] ) gr.Markdown(""" ### Useful Links - [Groq Console](https://console.groq.com/) - Get your free API key - [Groq Documentation](https://console.groq.com/docs/quickstart) - Learn more about Groq models - [Compound Models Info](https://console.groq.com/docs/models) - Details about compound models ### Tips - The compound models automatically use web search when real-time information is needed - Try different temperature settings: 0.1 for factual queries, 0.7-0.9 for creative questions - compound-beta is more capable but slower, compound-beta-mini is faster but less capable - Check the Tool Execution Info to see when web search was used """) return demo if __name__ == "__main__": demo = create_interface() demo.launch( share=True )