updated prompt

app.py

@@ -62,6 +62,15 @@ class GeminiModel(Model):
 - Be precise about dates, numbers, and classifications
 - If information seems unclear or contradictory, search for additional sources
 - When counting items, be explicit about what you're counting and why
+- **UNDERSTAND CONTEXT AND SYNONYMS**: When searching for specific terms, also look for related terms, synonyms, and contextual descriptions
+- **READ FOR CONTEXT**: Don't just search for exact phrases - read the content to understand who or what is being described
+- **CONSIDER ALTERNATIVE TERMINOLOGY**: "Horse doctor" = "equine veterinarian", "physician" = "doctor", etc.
+- **BREAK DOWN COMPLEX QUESTIONS**: For questions with multiple steps, identify each step clearly:
+  1. What is the first thing I need to find?
+  2. What is the second thing I need to find using the first result?
+  3. What is the final answer I need to provide?
+- **FOLLOW THE LOGICAL CHAIN**: Don't jump to conclusions - follow each step in order
+- Don't take the final answera from web results preview, always read the full content of the page
 
 **YOUR TOOLS**
 - `web_search(query: str)`: Finds URLs and information.
@@ -90,20 +99,20 @@ web_search(query="Mercedes Sosa studio albums discography 2000-2009")
 *Observation:* Search results show various sources including Wikipedia, AllMusic, and other music databases.
 
 **Your Turn 2:**
-Thought: I need to check multiple sources to ensure accuracy. Let me visit a reliable music database.
+Thought: I need to check multiple sources to ensure accuracy. Let me visit Wikipedia.
 Code:
 ```py
-visit_webpage(url="https://www.allmusic.com/artist/mercedes-sosa-mn0000869068")
+visit_webpage(url="https://en.wikipedia.org/wiki/Mercedes_Sosa")
 ```<end_code>
-*Observation:* The AllMusic page shows studio albums with release dates. I can see: Corazón Libre (2005), Cantora 1 (2009), Cantora 2 (2009).
+*Observation:* The Wikipedia page shows studio albums with release dates. I can see: Corazón Libre (2005), Cantora 1 (2009), Cantora 2 (2009).
 
 **Your Turn 3:**
 Thought: I found 3 studio albums in the specified period. Let me verify this with another source to be sure.
 Code:
 ```py
-visit_webpage(url="https://en.wikipedia.org/wiki/Mercedes_Sosa")
+visit_webpage(url="https://www.allmusic.com/artist/mercedes-sosa-mn0000869068#discography")
 ```<end_code>
-*Observation:* The Wikipedia page confirms the same 3 studio albums in the discography section.
+*Observation:* AllMusic shows more albums in the discography section between 2000 and 2009: Acustico (2002), Maestros del Folklore (2003), Argentina Quiere Cantar (2003), Para Cantar He Nacido (2005), Corazón Libre (2005), Acustico Na Suica (2007), Acustico en Suiza Inedito (2009), Cantora, Vol. 2 (2009). But it doesn't show the album type - studio, live or something else. I can assume that the 3 studio albums are the albums mentionCorazón Libre (2005), Cantora 1 (2009), and Cantora 2 (2009).
 
 **Your Turn 4:**
 Thought: I have verified from multiple sources that there are exactly 3 studio albums: Corazón Libre (2005), Cantora 1 (2009), and Cantora 2 (2009).
@@ -111,6 +120,34 @@ Code:
 ```py
 final_answer("3")
 ```<end_code>
+
+---
+**EXAMPLE 3: Understanding Context and Alternative Terms**
+*Question:* What is the surname of the equine veterinarian mentioned in the LibreTexts chemistry exercises?
+
+**Your Turn 1:**
+Thought: I need to find the LibreTexts chemistry exercises and look for mentions of an equine veterinarian. I should also search for related terms like "horse doctor" or "veterinarian" since the exact phrase might not be used.
+Code:
+```py
+web_search(query="LibreTexts chemistry exercises equine veterinarian horse doctor")
+```<end_code>
+*Observation:* Search results show the LibreTexts exercises page.
+
+**Your Turn 2:**
+Thought: I found the exercises page. Now I need to read the content carefully, looking not just for "equine veterinarian" but also for related terms and contextual descriptions of someone who treats horses.
+Code:
+```py
+visit_webpage(url="https://chem.libretexts.org/Bookshelves/Introductory_Chemistry/Introductory_Chemistry_(LibreTexts)/01%3A_The_Chemical_World/1.E%3A_Exercises")
+```<end_code>
+*Observation:* The page contains a story about Pasteur and a "horse doctor in eastern France named Louvrier" who claimed to have a cure for anthrax.
+
+**Your Turn 3:**
+Thought: I found the answer! The "horse doctor" mentioned is Louvrier. Even though the text says "horse doctor" instead of "equine veterinarian", this is clearly the same profession and the person being asked about.
+Code:
+```py
+final_answer("Louvrier")
+```<end_code>
+
 """
 
     def generate(

tools.py

@@ -1,4 +1,5 @@
 import re
+import time
 from typing import Any
 
 from smolagents import tool, Tool
@@ -37,11 +38,32 @@ class DuckDuckGoSearchTool(Tool):
         self.ddgs = DDGS(**kwargs)
 
     def forward(self, query: str) -> str:
-        results = self.ddgs.text(query, max_results=self.max_results)
-        if len(results) == 0:
-            raise Exception("No results found! Try a less restrictive/shorter query.")
-        postprocessed_results = [f"[{result['title']}]({result['href']})\n{result['body']}" for result in results]
-        return "## Search Results\n\n" + "\n\n".join(postprocessed_results)
+        max_retries = 3
+        retry_delay = 2  # seconds
+        
+        for attempt in range(max_retries):
+            try:
+                results = self.ddgs.text(query, max_results=self.max_results)
+                if len(results) == 0:
+                    raise Exception("No results found! Try a less restrictive/shorter query.")
+                postprocessed_results = [f"[{result['title']}]({result['href']})\n{result['body']}" for result in results]
+                return "## Search Results\n\n" + "\n\n".join(postprocessed_results)
+                
+            except Exception as e:
+                error_msg = str(e)
+                if "202" in error_msg or "rate" in error_msg.lower() or "ratelimit" in error_msg.lower():
+                    if attempt < max_retries - 1:
+                        print(f"Rate limit hit, retrying in {retry_delay} seconds... (attempt {attempt + 1}/{max_retries})")
+                        time.sleep(retry_delay)
+                        retry_delay *= 2  # Exponential backoff
+                        continue
+                    else:
+                        return "## Search Results\n\nRate limit exceeded. Please try again later or use a different search strategy."
+                else:
+                    # For non-rate-limit errors, don't retry
+                    raise e
+        
+        return "## Search Results\n\nFailed to get search results after multiple attempts."
     
 class VisitWebpageTool(Tool):
     name = "visit_webpage"