Update app.py

app.py

@@ -1,8 +1,101 @@
 import os
 import gradio as gr
-import requests
 import inspect
 import pandas as pd
+from agents import Agent, Runner, function_tool
+from duckduckgo_search import DDGS
+from agents import Agent, Runner
+from markdownify import markdownify
+from duckduckgo_search import DDGS
+from bs4 import BeautifulSoup
+from pydantic import BaseModel, Field
+import nest_asyncio
+import requests
+import os, re
+import litellm
+
+os.getenv("OPENAI_API_KEY")
+os.getenv("GEMINI_API_KEY")
+os.environ["LITELLM_PROVIDER"] = "gemini"
+
+# add this
+nest_asyncio.apply()
+
+
+#Tools
+
+@function_tool
+def web_search(query: str) -> str:
+    """
+    Perform a web search.
+    Args:
+        query (str): The search query string.
+    Returns:
+        str: The search results formatted in markdown.
+    """
+
+    try:
+        results = DDGS().text(query, max_results=10)
+
+        if not results:
+            raise Exception("No search results found.")
+
+        formatted_results = []
+        for i, result in enumerate(results, 1):
+            title = result.get("title", "No title available.")
+            link = result.get("href", "No link available.")
+            snippet = result.get("body", "No description available.")
+
+            entry = "  \n".join([
+                f"**Title**: {title}",
+                f"**Link**: {link}",
+                f"**Snippet**: {snippet}"
+            ])
+
+            formatted_results.append(entry)
+
+        return "\n\n".join(formatted_results)
+
+    except Exception as e:
+        return f"Error executing the query: {e}"
+
+@function_tool
+def visit_website(url: str) -> str:
+    """
+    Extract the contents of a website.
+    Args:
+        url (str): The URL of the website to visit.
+    Returns:
+        str: Formatted markdown ready for LLM consumption.
+    """
+
+    headers = {
+        "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36"
+    }
+
+    try:
+        response = requests.get(url, headers=headers, timeout=10)
+        response.raise_for_status()
+
+        html_content = response.text
+        soup = BeautifulSoup(html_content, 'html.parser')
+
+        for tag in soup(['script', 'style', 'nav', 'header', 'footer', 'aside', 'meta']):
+            tag.decompose()
+
+        main_content = soup.body
+        markdown_text = markdownify(str(main_content), strip=['img', 'iframe', 'script', 'meta', 'button', 'input', 'svg'])
+
+        max_length = 10000
+        markdown_text = re.sub(r'\n\s*\n', '\n\n', markdown_text[:max_length])
+
+        return markdown_text
+
+    except requests.RequestException as e:
+        return f"Error fetching the website: {e}"
+
+
+
 
 # (Keep Constants as is)
 # --- Constants ---
@@ -15,9 +108,57 @@ class BasicAgent:
         print("BasicAgent initialized.")
     def __call__(self, question: str) -> str:
         print(f"Agent received question (first 50 chars): {question[:50]}...")
-        fixed_answer = "This is a default answer."
-        print(f"Agent returning fixed answer: {fixed_answer}")
-        return fixed_answer
+
+        instructions = """
+You are a ReAct (Reason-Act-Observe) agent that searches the internet to find accurate answers to questions.
+## Available Tools
+- **web_search**: Search the web for information
+- **visit_website**: Visit specific webpages for detailed content
+## Output Format Rules
+Your final answer must be **exactly one** of these formats:
+- **Single number**: No commas, units, or symbols (unless explicitly requested)
+- **Single word/phrase**: No abbreviations (write "Los Angeles" not "LA")  
+- **Comma-separated list**: Each item follows the above rules
+**Important**: Provide ONLY the final answer - no explanations, markdown, or extra text.
+## ReAct Process
+Follow this cycle until you find the answer:
+**Thought**: [Internal reasoning about your next step]
+**Action**: [Single tool call]
+**Observation**: [Tool result will appear here]
+## Quality Guidelines
+- Use multiple sources when possible to verify accuracy
+- For recent events, prioritize newer sources
+- If information conflicts between sources, use the most authoritative source
+- For numerical data, ensure you're using the most current figures
+## Before Final Answer
+- Internally verify: "Does my answer violate format rules (extra text, wrong units, abbreviations)?"
+- Before providing a final answer, always ensure it contains the minimal amount of text possible.
+## Examples
+- Q: What is 15 + 27? → 42
+- Q: What is the capital of France? → Paris
+- Q: What are the top 3 most populous US states? → California, Texas, Florida
+"""
+
+            my_agent = Agent(
+                name="Expert Question Answering Agent",
+                instructions=instructions,
+                tools = [
+                    web_search,
+                    visit_website
+                ],
+                model="gpt-4o-mini"
+            )
+
+        result = Runner.run_sync(
+                my_agent,
+                input=question,
+                max_turns=25
+            )
+
+        
+
+        print(f"Agent returning fixed answer(first 50 chars): {result.final_output[:50]}...")
+        return result.final_output
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """