Update app.py

app.py

@@ -8,181 +8,134 @@ import re
 import time
 from google.api_core import exceptions
 
-# --- Constants ---
+# --- Constants (No Changes) ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-MAX_ITERATIONS = 7 # Max steps in the ReAct loop
-MAX_RETRIES = 5 # NEW: Max retries for API calls
-
-# --- Tool Definitions (No changes here, kept for completeness) ---
+MAX_ITERATIONS = 7
+MAX_RETRIES = 5
 
+# --- Tool Definitions (No Changes) ---
 class WebSearchTool:
-    """
-    A tool to search the web using the Perplexity API.
-    It returns a concise answer to a given query.
-    """
     def __init__(self, api_key):
         self.api_key = api_key
         self.url = "https://api.perplexity.ai/chat/completions"
-        print("WebSearchTool initialized.")
-
     def execute(self, query: str) -> str:
         print(f"Executing WebSearchTool with query: {query}")
-        payload = {
-            "model": "llama-3-sonar-small-32k-online",
-            "messages": [
-                # MODIFIED: A slightly better prompt for GAIA-style questions
-                {"role": "system", "content": "You are a world-class research assistant. Answer the user's query based on verifiable public information. Be precise and comprehensive."},
-                {"role": "user", "content": query}
-            ]
-        }
-        headers = {
-            "accept": "application/json",
-            "content-type": "application/json",
-            "Authorization": f"Bearer {self.api_key}"
-        }
+        payload = {"model": "llama-3-sonar-small-32k-online", "messages": [{"role": "system", "content": "You are a world-class research assistant. Answer the user's query based on verifiable public information. Be precise and comprehensive."}, {"role": "user", "content": query}]}
+        headers = {"accept": "application/json", "content-type": "application/json", "Authorization": f"Bearer {self.api_key}"}
         try:
             response = requests.post(self.url, json=payload, headers=headers, timeout=30)
             response.raise_for_status()
-            result = response.json()
-            answer = result['choices'][0]['message']['content']
-            print(f"WebSearchTool result: {answer[:150]}...")
-            return answer
+            return response.json()['choices'][0]['message']['content']
         except requests.exceptions.RequestException as e:
-            print(f"Error calling Perplexity API: {e}")
             return f"Error: Could not get a response from the web search tool. {e}"
 
 class FileDownloaderTool:
-    """
-    A tool to download and read the content of a file associated with a task.
-    The input should be the task_id.
-    """
     def __init__(self, api_url: str):
         self.api_url = api_url
-        print("FileDownloaderTool initialized.")
-
     def execute(self, task_id: str) -> str:
-        print(f"Executing FileDownloaderTool for task_id: {task_id}")
         file_url = f"{self.api_url}/files/{task_id}"
         try:
             response = requests.get(file_url, timeout=20)
             response.raise_for_status()
             content = response.text
-            print(f"FileDownloaderTool successfully read file for task {task_id}. Content length: {len(content)}")
-            if len(content) > 5000:
-                return f"File content (first 5000 chars):\n{content[:5000]}"
+            if len(content) > 5000: return f"File content (first 5000 chars):\n{content[:5000]}"
             return f"File content:\n{content}"
         except requests.exceptions.HTTPError as e:
-            if e.response.status_code == 404:
-                print(f"No file found for task_id: {task_id}")
-                return "No file is associated with this task."
-            else:
-                 print(f"HTTP error downloading file for task_id {task_id}: {e}")
-                 return f"Error: Failed to download file due to an HTTP error: {e}"
+            if e.response.status_code == 404: return "No file is associated with this task."
+            return f"Error: Failed to download file due to an HTTP error: {e}"
         except requests.exceptions.RequestException as e:
-            print(f"Network error downloading file for task_id {task_id}: {e}")
             return f"Error: Failed to download file due to a network error: {e}"
 
-
 # --- GAIA Agent Definition ---
 class GAIAAgent:
     def __init__(self, gemini_api_key: str, pplx_api_key: str, api_url: str):
         print("Initializing GAIAAgent...")
         genai.configure(api_key=gemini_api_key)
         self.model = genai.GenerativeModel('gemini-1.5-flash-latest')
+        self.tools = {"WebSearch": WebSearchTool(api_key=pplx_api_key), "FileDownloader": FileDownloaderTool(api_url=api_url)}
 
-        self.tools = {
-            "WebSearch": WebSearchTool(api_key=pplx_api_key),
-            "FileDownloader": FileDownloaderTool(api_url=api_url),
-        }
-        
-        # MODIFIED: A simpler prompt for the initial zero-shot check
+        # MODIFIED: Made the zero-shot prompt even stricter to prevent conversational filler.
         self.zero_shot_prompt_template = """
-You are a helpful assistant. Your job is to answer the user's question directly and concisely.
-Do not explain your reasoning.
-Do not use tools.
-If you can answer the question with high confidence, provide the answer.
-If the question requires browsing the web, accessing a file, or performing complex calculations, respond with the single word: "UNSURE".
+You are a highly intelligent question-answering bot. Follow these instructions precisely.
+1. Analyze the user's question.
+2. If the question is simple and you are 100% certain of the answer without needing any tools, provide ONLY the answer and nothing else.
+3. If the question requires web searches, file access, or complex reasoning, respond with the single word: UNSURE.
+Do not add any explanations or introductory phrases.
 
 Question: {question}
-Answer:
-"""
+Answer:"""
 
+        # MODIFIED: Added a new, explicit rule for how to fail gracefully.
         self.react_prompt_template = """
 You are a helpful assistant designed to answer questions accurately.
 
 To solve the user's question, you must use a sequence of thoughts and actions.
 You have access to the following tools:
 
-- **WebSearch[query]**: Use this to search the internet for up-to-date information, facts, or general knowledge.
-- **FileDownloader[task_id]**: Use this to download and read a file associated with the current task. The task_id is '{task_id}'.
+- **WebSearch[query]**: Use this to search the internet for up-to-date information.
+- **FileDownloader[task_id]**: Use this to download a file associated with the current task. The task_id is '{task_id}'.
 
 Your reasoning process should follow this format:
 
 Thought: I need to figure out what information is missing. I will use a tool to find it.
 Action: ToolName[input for the tool]
 Observation: [The result from the tool will be inserted here]
-... (this Thought/Action/Observation cycle can repeat multiple times)
+... (this Thought/Action/Observation cycle can repeat)
 
 Thought: I have now gathered enough information to answer the user's question.
 Final Answer: The final answer to the original question.
 
 **Important Rules:**
-1. The `Action` line must be *exactly* in the format `ToolName[input]`. For example: `WebSearch[When was the Eiffel Tower built?]`.
+1. The `Action` line must be *exactly* in the format `ToolName[input]`.
 2. The `task_id` for the current question is '{task_id}'. Use it ONLY with the FileDownloader tool.
-3. Once you have the final answer, do not use any more tools. State the final answer clearly after "Final Answer:". Your entire response should end here.
+3. **CRITICAL RULE:** If you determine that the question cannot be answered with your tools (e.g., a required file is missing, the information is not on the web), you MUST conclude with: `Final Answer: I am unable to answer this question.` Do not make up an answer.
 
 Here is the question:
-{question}
-"""
+{question}"""
         print("GAIAAgent initialized successfully.")
 
-    # NEW: Function to handle API calls with exponential backoff
     def _call_gemini_api_with_backoff(self, prompt_text):
         retries = 0
         while retries < MAX_RETRIES:
             try:
-                print(f"Attempt {retries + 1} to call Gemini API...")
+                time.sleep(1) # Add a small base delay
                 response = self.model.generate_content(prompt_text)
                 return response.text
             except exceptions.ResourceExhausted as e:
-                print(f"API Rate Limit Exceeded (429). Waiting to retry... ({e.message})")
-                wait_time = (2 ** retries) + 1 # Exponential backoff: 2, 3, 5, 9, 17 seconds
+                wait_time = (2 ** retries)
+                print(f"API Rate Limit Exceeded (429). Waiting for {wait_time}s to retry...")
                 time.sleep(wait_time)
                 retries += 1
             except Exception as e:
-                print(f"An unexpected error occurred with Gemini API: {e}")
                 return f"AGENT_ERROR: An unexpected error occurred: {e}"
-        
-        print("Max retries reached. Failing.")
         return "AGENT_ERROR: API rate limit exceeded after multiple retries."
 
     def __call__(self, question: str, task_id: str) -> str:
         print(f"\n{'='*20}\nProcessing Task ID: {task_id}\nQuestion: {question[:100]}...")
 
-        # === NEW: Step 1 - Zero-Shot Attempt ===
         print("--- Step 1: Zero-Shot Attempt ---")
         zero_shot_prompt = self.zero_shot_prompt_template.format(question=question)
         zero_shot_answer = self._call_gemini_api_with_backoff(zero_shot_prompt).strip()
 
-        if "AGENT_ERROR" in zero_shot_answer:
-            return zero_shot_answer # Propagate API failure
+        if "AGENT_ERROR" in zero_shot_answer: return zero_shot_answer
         
         if "UNSURE" not in zero_shot_answer.upper():
             print(f"Zero-shot successful! Answer: {zero_shot_answer}")
             return zero_shot_answer
 
-        # === MODIFIED: Step 2 - ReAct Loop ===
         print("--- Step 2: Zero-shot failed, starting ReAct loop ---")
-        react_prompt = self.react_prompt_template.format(question=question, task_id=task_id)
+        
+        # CRITICAL FIX: Reset the prompt history for each question to prevent context bleed.
+        # This was the cause of the botany/bird video mix-up.
+        current_prompt_history = self.react_prompt_template.format(question=question, task_id=task_id)
         
         for i in range(MAX_ITERATIONS):
             print(f"\n--- ReAct Iteration {i+1} ---")
             
-            response_text = self._call_gemini_api_with_backoff(react_prompt)
+            response_text = self._call_gemini_api_with_backoff(current_prompt_history)
             print(f"LLM Response:\n{response_text}")
 
-            if "AGENT_ERROR" in response_text:
-                return response_text # Propagate API failure
+            if "AGENT_ERROR" in response_text: return response_text
 
             final_answer_match = re.search(r"Final Answer:\s*(.*)", response_text, re.DOTALL)
             if final_answer_match:
@@ -196,152 +149,86 @@ Here is the question:
                 tool_input = action_match.group(2).strip()
 
                 if tool_name in self.tools:
-                    print(f"Executing tool '{tool_name}' with input '{tool_input}'")
                     tool = self.tools[tool_name]
                     try:
-                        if tool_name == "FileDownloader":
-                            observation = tool.execute(task_id)
-                        else:
-                            observation = tool.execute(tool_input)
+                        observation = tool.execute(task_id if tool_name == "FileDownloader" else tool_input)
                     except Exception as e:
                         observation = f"Error executing tool: {e}"
-                    
-                    react_prompt += f"{response_text}\nObservation: {observation}\n"
+                    # Append the whole thought/action/observation cycle
+                    current_prompt_history += f"\n{response_text}\nObservation: {observation}"
                 else:
-                    print(f"Error: Agent tried to use an unknown tool: {tool_name}")
-                    react_prompt += f"{response_text}\nObservation: Error - The tool '{tool_name}' does not exist.\n"
+                    current_prompt_history += f"\n{response_text}\nObservation: Error - The tool '{tool_name}' does not exist."
             else:
                 print("Error: Agent did not provide a valid Action or Final Answer. Returning last response.")
                 return response_text.strip()
 
-        print("Agent reached max iterations without finding a final answer.")
-        return "AGENT_ERROR: Agent could not determine the answer within the allowed number of steps."
-
+        return "AGENT_ERROR: Agent reached max iterations."
 
+# --- Main run_and_submit_all function (No significant changes needed here, only added a longer sleep) ---
 def run_and_submit_all(profile: gr.OAuthProfile | None):
-    # This function is mostly the same, with one key change added.
     space_id = os.getenv("SPACE_ID")
-
-    if profile:
-        username = f"{profile.username}"
-        print(f"User logged in: {username}")
-    else:
-        print("User not logged in.")
-        return "Please Login to Hugging Face with the button.", None
+    if not profile: return "Please Login to Hugging Face with the button.", None
+    username = f"{profile.username}"
+    print(f"User logged in: {username}")
 
     pplx_key = os.getenv("PPLX_API_KEY")
     gemini_key = os.getenv("GEMINI_API_KEY")
-
-    if not pplx_key or not gemini_key:
-        error_msg = "API keys not found in Space secrets. Please set PPLX_API_KEY and GEMINI_API_KEY."
-        print(error_msg)
-        return error_msg, None
+    if not pplx_key or not gemini_key: return "API keys not found in Space secrets.", None
     
     api_url = DEFAULT_API_URL
-    questions_url = f"{api_url}/questions"
-    submit_url = f"{api_url}/submit"
-
     try:
         agent = GAIAAgent(gemini_api_key=gemini_key, pplx_api_key=pplx_key, api_url=api_url)
-    except Exception as e:
-        print(f"Error instantiating agent: {e}")
-        return f"Error initializing agent: {e}", None
-    
-    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
-    print(f"Agent code link: {agent_code}")
-
-    try:
-        response = requests.get(questions_url, timeout=15)
-        response.raise_for_status()
-        questions_data = response.json()
-        if not questions_data:
-             return "Fetched questions list is empty or invalid format.", None
-        print(f"Fetched {len(questions_data)} questions.")
-    except Exception as e:
-        return f"Error fetching questions: {e}", None
+        questions_data = requests.get(f"{api_url}/questions", timeout=15).json()
+    except Exception as e: return f"Error during setup: {e}", None
 
-    results_log = []
-    answers_payload = []
-    print(f"Running agent on {len(questions_data)} questions...")
+    results_log, answers_payload = [], []
     for item in questions_data:
-        task_id = item.get("task_id")
-        question_text = item.get("question")
-        if not task_id or question_text is None:
-            continue
+        task_id, question_text = item.get("task_id"), item.get("question")
+        if not task_id or question_text is None: continue
         try:
             submitted_answer = agent(question_text, task_id)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e:
-             print(f"Error running agent on task {task_id}: {e}")
              results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
         
-        # NEW: Add a delay between each question to respect rate limits
-        print(f"--- Waiting for 5 seconds before next question... ---")
-        time.sleep(5)
-
-    if not answers_payload:
-        return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
+        print(f"--- Waiting for 8 seconds before next question to respect rate limits ---")
+        time.sleep(8) # Increased delay to be safer
 
+    if not answers_payload: return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
+    
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
     submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
-    status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
-    print(status_update)
-
-    print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
+    
     try:
-        response = requests.post(submit_url, json=submission_data, timeout=120)
+        response = requests.post(f"{api_url}/submit", json=submission_data, timeout=120)
         response.raise_for_status()
         result_data = response.json()
-        final_status = (
-            f"Submission Successful!\n"
-            f"User: {result_data.get('username')}\n"
-            f"Overall Score: {result_data.get('score', 'N/A')}% "
-            f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
-            f"Message: {result_data.get('message', 'No message received.')}"
-        )
-        print("Submission successful.")
-        results_df = pd.DataFrame(results_log)
-        return final_status, results_df
-    except requests.exceptions.HTTPError as e:
-        error_detail = f"Server responded with status {e.response.status_code}."
-        try:
-            error_json = e.response.json()
-            error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
-        except requests.exceptions.JSONDecodeError:
-            error_detail += f" Response: {e.response.text[:500]}"
-        status_message = f"Submission Failed: {error_detail}"
-        print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
-    except Exception as e:
-        status_message = f"An unexpected error occurred during submission: {e}"
-        print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
-
-# --- Gradio Interface (No changes here) ---
+        final_status = (f"Submission Successful!\nUser: {result_data.get('username')}\n"
+                        f"Overall Score: {result_data.get('score', 'N/A')}% "
+                        f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
+                        f"Message: {result_data.get('message', 'No message received.')}")
+        return final_status, pd.DataFrame(results_log)
+    except requests.exceptions.RequestException as e:
+        return f"Submission Failed: {e}", pd.DataFrame(results_log)
+
+# --- Gradio Interface (No changes) ---
 with gr.Blocks() as demo:
     gr.Markdown("# GAIA Agent Evaluation Runner")
-    gr.Markdown(
-        """
+    gr.Markdown("""
         **Instructions:**
         1.  Ensure you have added your `PPLX_API_KEY` and `GEMINI_API_KEY` to this Space's **Settings > Secrets**.
         2.  Log in to your Hugging Face account using the button below. This uses your HF username for submission.
         3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
         ---
         **Disclaimers:**
-        This process will now be slower due to the added delays to respect API rate limits, but it should be much more reliable. Please be patient.
-        """
-    )
+        This process is slow due to the added delays to respect API rate limits. Please be patient.
+        """)
     gr.LoginButton()
     run_button = gr.Button("Run Evaluation & Submit All Answers")
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
     results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
-    run_button.click(
-        fn=run_and_submit_all,
-        outputs=[status_output, results_table]
-    )
+    run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
 
 if __name__ == "__main__":
-    print("Launching Gradio Interface for GAIA Agent Evaluation...")
     demo.launch(debug=True, share=False)