test3

app.py

@@ -5,28 +5,27 @@ import pandas as pd
 import gradio as gr
 from openai import OpenAI
 
-# --- Agent yang memakai GPT-4 dengan Chain-of-Thought dan Tools ---
 class ToolEnhancedAgent:
     def __init__(self):
         api_key = os.getenv("OPENAI_API_KEY")
         if not api_key:
-            raise ValueError("OPENAI_API_KEY is not set in environment variables.")
+            raise ValueError("OPENAI_API_KEY not found in environment variables.")
         self.client = OpenAI(api_key=api_key)
-        print("ToolEnhancedAgent initialized with GPT-4 + CoT + Tools.")
+        print("ToolEnhancedAgent initialized.")
 
     def use_tool(self, tool_name: str, input_text: str) -> str:
         try:
             if tool_name == "calculator":
-                # Hati-hati menggunakan eval, sebaiknya batasi ekspresi jika produksi
+                # Sangat dasar dan raw eval, hati2 di produksi
                 return str(eval(input_text))
             elif tool_name == "date":
                 return str(datetime.datetime.now().date())
             elif tool_name == "wikipedia":
                 return self.search_wikipedia(input_text)
             else:
-                return "[Tool Error: Unknown Tool]"
+                return "[Unknown tool]"
         except Exception as e:
-            return f"[Tool Error: {e}]"
+            return f"[Tool error: {e}]"
 
     def search_wikipedia(self, query: str) -> str:
         try:
@@ -35,36 +34,38 @@ class ToolEnhancedAgent:
             if res.status_code == 200:
                 return res.json().get("extract", "No summary found.")
             else:
-                return f"No Wikipedia summary found for '{query}'."
+                return f"No Wikipedia summary for '{query}'."
         except Exception as e:
-            return f"Wikipedia error: {e}"
+            return f"Wikipedia API error: {e}"
 
     def __call__(self, question: str) -> str:
+        # Prompt dengan chain of thought agar GPT berpikir langkah demi langkah
         prompt = (
-            "You are a helpful AI assistant. Use tools if needed. Think step-by-step before answering.\n"
+            "You are an assistant that solves problems step-by-step, and you can use tools like calculator, date, and wikipedia if needed.\n"
             f"Question: {question}\n"
-            "Answer (show your thinking steps):"
+            "Answer (think step-by-step and use tools if helpful):"
         )
+
         try:
             response = self.client.chat.completions.create(
-                model="gpt-4",
+                model="gpt-4o-mini",  # model yang lebih ringan dan biasanya cukup
                 messages=[
-                    {"role": "system", "content": "You are a smart assistant that can use tools and think step-by-step."},
+                    {"role": "system", "content": "You are a helpful assistant that thinks step-by-step and can use tools."},
                     {"role": "user", "content": prompt}
                 ],
-                temperature=0.3,
+                temperature=0.2,
                 max_tokens=700,
             )
             answer = response.choices[0].message.content.strip()
-            print(f"Answer generated (truncated): {answer[:150]}...")
+            print(f"Generated answer (preview): {answer[:100]}...")
             return answer
         except Exception as e:
-            print(f"[Agent Error]: {e}")
-            return f"[Agent Error: {e}]"
+            print(f"Agent error: {e}")
+            return f"[Agent error: {e}]"
+
 
-# --- Fungsi utama untuk fetch soal, jalankan agent, dan submit jawaban ---
 def run_and_submit_all(profile: gr.OAuthProfile | None):
-    if not profile:
+    if profile is None:
         return "Please login to Hugging Face to submit answers.", None
 
     username = profile.username
@@ -75,14 +76,14 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     try:
         agent = ToolEnhancedAgent()
     except Exception as e:
-        return f"Error initializing agent: {e}", None
+        return f"Agent initialization error: {e}", None
 
     try:
         response = requests.get(questions_url, timeout=15)
         response.raise_for_status()
         questions = response.json()
-        if not questions:
-            return "No questions fetched or invalid response.", None
+        if not isinstance(questions, list) or len(questions) == 0:
+            return "Failed to fetch questions or empty list.", None
     except Exception as e:
         return f"Error fetching questions: {e}", None
 
@@ -106,11 +107,11 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             results_log.append({
                 "Task ID": task_id,
                 "Question": question_text,
-                "Submitted Answer": f"[Agent Error: {e}]"
+                "Submitted Answer": f"[Agent error: {e}]"
             })
 
     if not answers:
-        return "Agent did not produce any answers.", pd.DataFrame(results_log)
+        return "Agent did not produce answers.", pd.DataFrame(results_log)
 
     agent_code_url = f"https://huggingface.co/spaces/{os.getenv('SPACE_ID')}/tree/main"
 
@@ -123,38 +124,43 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     try:
         submit_response = requests.post(submit_url, json=submission_payload, timeout=60)
         submit_response.raise_for_status()
-        result_data = submit_response.json()
+        result = submit_response.json()
 
         final_status = (
-            f"Submission Successful!\n"
-            f"User: {result_data.get('username')}\n"
-            f"Score: {result_data.get('score', 'N/A')}%\n"
-            f"Correct: {result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')}\n"
-            f"Message: {result_data.get('message', 'No message received.')}"
+            f"Submission Success!\n"
+            f"User: {result.get('username')}\n"
+            f"Score: {result.get('score', 'N/A')}%\n"
+            f"Correct: {result.get('correct_count', '?')}/{result.get('total_attempted', '?')}\n"
+            f"Message: {result.get('message', 'No message')}"
         )
         return final_status, pd.DataFrame(results_log)
 
     except Exception as e:
-        return f"Submission failed: {e}", pd.DataFrame(results_log)
+        return f"Submission error: {e}", pd.DataFrame(results_log)
+
 
-# --- Gradio UI ---
 with gr.Blocks() as demo:
-    gr.Markdown("# Tool Enhanced Agent for GAIA Benchmark")
+    gr.Markdown("# GAIA Benchmark - Tool Enhanced Agent")
 
     gr.Markdown(
         """
         1. Login with your Hugging Face account.
-        2. Click 'Run & Submit All Answers' to evaluate the agent on GAIA tasks.
-        3. View your score and answers below.
+        2. Click 'Run Evaluation & Submit All Answers' to run the agent on GAIA tasks.
+        3. View your results and submission status.
         """
     )
 
-    login = gr.LoginButton()
+    login_button = gr.LoginButton()
     run_button = gr.Button("Run Evaluation & Submit All Answers")
-    status_output = gr.Textbox(label="Status / Submission Result", lines=7)
-    results_table = gr.DataFrame(label="Questions and Answers")
+    status_box = gr.Textbox(label="Status / Submission Result", lines=7)
+    results_table = gr.DataFrame(label="Questions and Agent Answers")
+
+    run_button.click(
+        fn=run_and_submit_all,
+        inputs=login_button,
+        outputs=[status_box, results_table]
+    )
 
-    run_button.click(fn=run_and_submit_all, inputs=login, outputs=[status_output, results_table])
 
 if __name__ == "__main__":
     demo.launch()