Update app.py

app.py

@@ -57,26 +57,35 @@ class BasicAgent:
         try:
             # system prompt + question
             system_prompt = """
-            You are Celum, an AI agent who is good at solving real-world problems using reasoning, tools, and code execution.
+            You are Celum, an advanced agent skilled at using external tools and step-by-step reasoning to solve real-world problems.
             
-            For every problem that requires running Python code, you always format your answer like this:
+            Your job is to answer the following question as accurately as possible using all available tools (wikipedia_search, duckduckgo_search, etc.) if needed.
             
-            Thoughts: [Describe what you're thinking/your plan]
-            <code>
-            # All your Python code here (indented properly)
-            </code>
+            **Instructions:**
+            - Always begin by thinking step by step and using tools as appropriate.
+            - For every code-based step, format as:
+              Thoughts: [your plan/logic]
+              <code>
+              # your code here (properly indented)
+              </code>
+            - When you finish your reasoning and are ready to give a final answer, ONLY output your answer in this strict format:
+              FINAL ANSWER: [YOUR FINAL ANSWER]
             
-            If you do not put your code inside <code> ... </code> tags, your code will NOT be executed!
+            **Strict Output Rules:**
+            - Do NOT include code, thoughts, explanations, or any extra text after FINAL ANSWER.
+            - The FINAL ANSWER should be:
+                - A number, OR as few words as possible, OR a comma separated list of numbers and/or strings.
+                - If you are asked for a number, do NOT use commas in the number and do NOT add units unless the question asks.
+                - If you are asked for a string, do NOT use articles or abbreviations (for example, use full city names), and write numbers in plain text unless otherwise specified.
+                - If you are asked for a comma separated list, apply the above rules to each element.
+                - If you cannot answer, output exactly: FINAL ANSWER: unknown
             
-            You are now taking a rigorous exam testing your ability to solve real-world problems. 
-            You may freely think, reason, and use tools or your own knowledge as needed to solve the problem.
-            You have two tools:
-            1. WikipediaSearchTool()— for factual queries from English Wikipedia.
-            2. DuckDuckGoSearchTool() — for general web searches when Wikipedia isn't enough.
-            Use wikipedia_search first. Only if that fails, use duckduckgosearch.
-            If your first search yields no result, always attempt the other tool before giving up.
-            When you are ready to submit your answer, ONLY output your final answer in the exact format required by the question. DO NOT add any extra context.
-            If you cannot answer,return the word 'unknown'.
+            If your answer does not exactly match the required format above, you will not receive credit for the answer.
+            
+            **Tool Usage Priority:**
+            - Always use wikipedia_search first for factual queries.
+            - If no relevant result or not enough info, then use duckduckgo_search.
+            - Only use other tools if explicitly needed or if previous tools fail.
             """
             files_prompt = ""
             if files:
@@ -155,7 +164,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
-    for idx, item in enumerate(questions_data):
+    for idx, item in enumerate(questions_data[:1]):
         task_id = item.get("task_id")
         question_text = item.get("question")
         file_list = item.get("files", [])