Updated with Serper API tool capabilities

README.md

@@ -11,4 +11,93 @@ hf_oauth: true
 hf_oauth_expiration_minutes: 480
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+# Agent Assignment - LlamaIndex Agent
+
+This project implements an intelligent agent using LlamaIndex with enhanced search capabilities.
+
+## Features
+
+- **Multiple Information Sources**: The agent can search Wikipedia, search Google using Serper API, and extract content from web pages
+- **Mathematical Operations**: Simple calculator for basic arithmetic operations
+- **Text Processing**: Includes a text reverser tool for handling reversed text queries
+- **Structured Output**: All responses include a clear "FINAL ANSWER" in a consistent format
+
+## Installation
+
+1. Clone this repository
+2. Install dependencies:
+
+```bash
+pip install -r requirements.txt
+```
+
+3. Set up environment variables:
+
+Create a `.env` file in the root directory with:
+
+```
+# Google API key for Gemini LLM
+GOOGLE_API_KEY=your_google_api_key
+
+# Or HuggingFace API key
+# HUGGINGFACE_API_KEY=your_huggingface_api_key
+# DEFAULT_LLM_PROVIDER=huggingface
+
+# Serper API key for Google search
+SERPER_API_KEY=your_serper_api_key
+```
+
+## Setting up Serper API
+
+The agent uses [Serper.dev](https://serper.dev/) to perform Google searches. To set this up:
+
+1. Sign up for an account at [Serper.dev](https://serper.dev/)
+2. Get your API key from the dashboard
+3. Add your API key to the `.env` file or set it using the utility script:
+
+```bash
+python set_env_var.py SERPER_API_KEY your_api_key_here
+```
+
+## Usage
+
+### Testing Individual Tools
+
+```bash
+# Test Wikipedia search
+python test_wikipedia.py
+
+# Test Google search (requires SERPER_API_KEY)
+python test_google_search.py
+
+# Test the full agent with various questions
+python test_agent_with_google.py
+```
+
+### Running the Agent
+
+```bash
+# Run the main application
+python app.py
+```
+
+## Troubleshooting
+
+- **Missing API Keys**: Make sure all required API keys are set in your `.env` file
+- **Wikipedia Issues**: The agent will try to use direct Wikipedia API first and fall back to LlamaIndex WikipediaReader
+- **Google Search Issues**: Verify your SERPER_API_KEY is correct and has sufficient quota
+
+## Testing Scripts
+
+- `test_wikipedia.py`: Tests the Wikipedia search functionality
+- `test_google_search.py`: Tests the Google search functionality using Serper API
+- `test_agent_output.py`: Tests the agent's output format
+- `test_agent_with_google.py`: Tests the agent with questions using Google search
+- `mini_test.py`: Minimal test script for checking the agent's initialization
+- `standalone_debug.py`: Standalone debug script without Gradio dependencies
+
+## Environment Utilities
+
+- `set_env_var.py`: Utility to easily set environment variables in the `.env` file

agent.py

@@ -1,6 +1,8 @@
 """Tool definitions and utility functions for the agent"""
 from typing import List, Dict, Any
 import os
+import json
+import requests
 from dotenv import load_dotenv
 
 from llama_index.core.tools import BaseTool, FunctionTool
@@ -54,6 +56,77 @@ def simple_calculator(operation: str, a: float, b: float) -> float:
         raise ValueError(f"Unknown operation: {operation}")
 
 # --- Information Retrieval Tools ---
+def google_search(query: str, num_results: int = 5) -> str:
+    """
+    Search Google using the Serper API.
+    
+    Args:
+        query: The search query
+        num_results: Number of results to return (default: 5)
+    
+    Returns:
+        A formatted string with the search results
+    """
+    api_key = os.getenv("SERPER_API_KEY")
+    if not api_key:
+        return "SERPER_API_KEY not found in environment variables. Cannot perform Google search."
+    
+    headers = {
+        "X-API-KEY": api_key,
+        "Content-Type": "application/json"
+    }
+    
+    payload = {
+        "q": query,
+        "num": num_results
+    }
+    
+    try:
+        response = requests.post(
+            "https://google.serper.dev/search",
+            headers=headers,
+            json=payload
+        )
+        response.raise_for_status()
+        search_results = response.json()
+        
+        # Format the results
+        formatted_results = []
+        
+        # Add organic results
+        if "organic" in search_results:
+            for i, result in enumerate(search_results["organic"][:num_results], 1):
+                title = result.get("title", "No Title")
+                snippet = result.get("snippet", "No snippet available")
+                link = result.get("link", "No link available")
+                formatted_results.append(f"Result {i}:\nTitle: {title}\nSnippet: {snippet}\nURL: {link}\n")
+        
+        # Add answer box if available
+        if "answerBox" in search_results:
+            answer_box = search_results["answerBox"]
+            answer_title = answer_box.get("title", "")
+            answer_snippet = answer_box.get("snippet", "")
+            answer_answer = answer_box.get("answer", "")
+            
+            answer_text = "Featured Answer:\n"
+            if answer_title:
+                answer_text += f"Title: {answer_title}\n"
+            if answer_answer:
+                answer_text += f"Answer: {answer_answer}\n"
+            if answer_snippet:
+                answer_text += f"Snippet: {answer_snippet}\n"
+                
+            formatted_results.insert(0, answer_text)  # Put answer box at the top
+        
+        # Return formatted results
+        if formatted_results:
+            return "\n".join(formatted_results)
+        else:
+            return f"No results found for query: '{query}'"
+            
+    except Exception as e:
+        return f"Error performing Google search: {str(e)}"
+
 def wikipedia_search(query: str, num_results: int = 2) -> str:
     """
     Search Wikipedia for information.
@@ -162,14 +235,21 @@ def get_tools() -> List[BaseTool]:
     web_tool = FunctionTool.from_defaults(
         fn=web_search,
         name="web_search",
-        description="Fetches and extracts content from a specific web page.",
+        description="Fetches and extracts content from a specific web page. Requires a full URL.",
+    )
+    
+    google_tool = FunctionTool.from_defaults(
+        fn=google_search,
+        name="google_search",
+        description="Searches Google for information. Use this for recent events, current information, or when Wikipedia doesn't have enough information.",
     )
     
     return [
         text_reverser_tool,
         calculator_tool,
         wikipedia_tool,
-        web_tool
+        web_tool,
+        google_tool
     ]
 
 if __name__ == "__main__":

system_prompt.txt

@@ -3,8 +3,10 @@ You are an intelligent assistant designed to handle a wide variety of questions
 Here are your capabilities and instructions:
 
 1. INFORMATION RETRIEVAL: 
+   - Use the google_search tool for most up-to-date information, recent events, or specific questions needing comprehensive search
    - Use the wikipedia_search tool for factual information, historical data, or general knowledge
    - Use the web_search tool when you have a specific URL to extract content from
+   - ALWAYS prefer google_search for recent events or when you need the most current information
 
 2. MATHEMATICAL REASONING: 
    - Use the calculator tool for basic arithmetic (add, subtract, multiply, divide)
@@ -16,14 +18,15 @@ Here are your capabilities and instructions:
    - After processing unusual formats, make sure to answer the actual question
 
 4. GENERAL KNOWLEDGE:
-   - Rely on your built-in knowledge for questions about specialized topics
+   - For factual questions, use google_search first, then wikipedia_search if needed
+   - For recent events or current information, always use google_search
    - Use your reasoning capabilities for logical problems
    - Apply your understanding of various subjects without requiring specialized tools
 
 When using tools:
 - Choose the most appropriate tool for each question
-- For factual questions, prefer Wikipedia for well-established facts
-- For recent events or specialized topics, use web search
+- For factual questions, prefer Google Search for up-to-date information, then fall back to Wikipedia if needed
+- For recent events or specialized topics, always use Google Search
 - If a question is unclear, ask for clarification
 
 Always provide concise, accurate answers that directly address the question. Format your response according to any specific instructions in the question.