Update app.py

app.py

@@ -6,207 +6,249 @@ import pandas as pd
 import json
 from duckduckgo_search import DDGS
 from dotenv import load_dotenv
+import ast # For safely evaluating literal structures if needed, though JSON is preferred
 
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+OPENROUTER_API_URL = "https://openrouter.ai/api/v1/chat/completions"
+MODEL_NAME = "deepseek/deepseek-chat-v3-0324" # Or your preferred OpenRouter model
 
 # --- Basic Agent Definition ---
-# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 class BasicAgent:
     def __init__(self):
         print("BasicAgent initialized.")
-        # Load environment variables
         load_dotenv()
         self.api_key = os.getenv("TEST_AGENT_KEY")
-        self.api_url = DEFAULT_API_URL
+        if not self.api_key:
+            raise ValueError("TEST_AGENT_KEY (OpenRouter API Key) not found in environment variables.")
+        self.scorer_api_url = DEFAULT_API_URL # For fetching files/submitting
+        self.llm_api_url = OPENROUTER_API_URL
+        self.model_name = MODEL_NAME
+
         # Load system prompt
-        with open("prompt.txt", "r") as file:
-            self.system_prompt = file.read().strip()
+        try:
+            with open("prompt.txt", "r") as file:
+                self.system_prompt = file.read().strip()
+        except FileNotFoundError:
+            print("Error: prompt.txt not found. Using a default system prompt.")
+            self.system_prompt = "You are a helpful AI assistant. Please answer the user's questions. Use tools if necessary by outputting TOOL: {\"name\": \"tool_name\", \"args\": {\"arg_name\": \"value\"}}. When you have the final answer, output ANSWER: your_final_answer."
+        except Exception as e:
+            print(f"Error loading prompt.txt: {e}. Using a default system prompt.")
+            self.system_prompt = "You are a helpful AI assistant. Please answer the user's questions. Use tools if necessary by outputting TOOL: {\"name\": \"tool_name\", \"args\": {\"arg_name\": \"value\"}}. When you have the final answer, output ANSWER: your_final_answer."
+
+
         # Define tools
         self.tools = {
-            "web_search_tool": lambda search_terms: web_search_tool(search_terms),
-            "decimal_approximation_tool": lambda number, decimals: decimal_approximation_tool(number, decimals),
-            "get_files_task_id_tool": lambda task_id: get_files_task_id_tool(task_id),
-            "image_processing_tool": lambda file_content: image_processing_tool(file_content)
+            "web_search_tool": web_search_tool,
+            "decimal_approximation_tool": decimal_approximation_tool,
+            "get_files_task_id_tool": get_files_task_id_tool
+            # image_processing_tool removed as requested
         }
-    def __call__(self, question: dict) -> str:
-        print(f"Agent received question (first 50 chars): {str(question)[:50]}...")
-        task_id = question["task_id"]
-        question_text = question["question"]
-        conversation = [{"role": "system", "content": self.system_prompt},
-                        {"role": "user", "content": question_text}]
-        while True:
-            prompt = "\n".join([msg["content"] for msg in conversation])
-            response = self._call_llm(prompt)
-            if response.startswith("ANSWER:"):
-                answer = response[len("ANSWER:"):].strip()
-                break
-            elif response.startswith("TOOL:"):
-                tool_call = response[len("TOOL:"):].strip()
+        print(f"Agent tools initialized: {list(self.tools.keys())}")
+
+    def _call_llm(self, conversation_history: list) -> str:
+        print(f"Calling LLM. Conversation history length: {len(conversation_history)}")
+        headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json",
+            "HTTP-Referer": os.getenv("SPACE_ID", "http://localhost"), # Recommended by OpenRouter
+            "X-Title": os.getenv("SPACE_TITLE", "Test Agent")      # Recommended by OpenRouter
+        }
+        payload = {
+            "model": self.model_name,
+            "messages": conversation_history,
+            "temperature": 0.7, # Adjust as needed
+            # "max_tokens": 1000 # Adjust as needed
+        }
+        try:
+            response = requests.post(self.llm_api_url, headers=headers, json=payload, timeout=120)
+            response.raise_for_status()
+            llm_response_data = response.json()
+            if llm_response_data.get("choices") and llm_response_data["choices"][0].get("message"):
+                content = llm_response_data["choices"][0]["message"].get("content", "").strip()
+                print(f"LLM raw response: {content[:200]}...")
+                return content
+            else:
+                print(f"LLM response malformed: {llm_response_data}")
+                return "Error: LLM response was malformed."
+        except requests.exceptions.Timeout:
+            print("Error: LLM API call timed out.")
+            return "Error: LLM call timed out."
+        except requests.exceptions.RequestException as e:
+            print(f"Error calling LLM API: {e}")
+            if e.response is not None:
+                print(f"LLM Error Response Status: {e.response.status_code}")
+                print(f"LLM Error Response Body: {e.response.text}")
+            return f"Error: Failed to communicate with LLM. {str(e)}"
+        except Exception as e:
+            print(f"An unexpected error occurred during LLM call: {e}")
+            return f"Error: An unexpected error occurred communicating with LLM. {str(e)}"
+
+    def __call__(self, question_data: dict) -> str:
+        task_id = question_data.get("task_id")
+        question_text = question_data.get("question")
+        print(f"Agent received task_id: {task_id}, question (first 50 chars): {str(question_text)[:50]}...")
+
+        if not task_id or question_text is None:
+            print("Error: Missing task_id or question in agent input.")
+            return "Error: Invalid input to agent."
+
+        conversation = [
+            {"role": "system", "content": self.system_prompt},
+            {"role": "user", "content": question_text}
+        ]
+
+        max_loops = 10 # Prevent infinite loops
+        for loop_count in range(max_loops):
+            print(f"\nAgent Loop: {loop_count + 1}")
+            llm_response = self._call_llm(conversation)
+
+            if llm_response.startswith("ANSWER:"):
+                answer = llm_response[len("ANSWER:"):].strip()
+                print(f"Agent returning final answer: {answer}")
+                return answer
+            elif llm_response.startswith("TOOL:"):
+                tool_call_str = llm_response[len("TOOL:"):].strip()
+                print(f"Attempting tool call: {tool_call_str}")
                 try:
-                    tool_name, args_str = tool_call.split("(", 1)
-                    args_str = args_str.rstrip(")")
-                    if tool_name == "get_files_task_id_tool":
-                        observation = self.tools[tool_name](task_id)
+                    tool_data = json.loads(tool_call_str) # Parse the JSON string
+                    tool_name = tool_data.get("name")
+                    tool_args_dict = tool_data.get("args", {})
+
+                    if tool_name in self.tools:
+                        print(f"Executing tool: {tool_name} with args: {tool_args_dict}")
+                        # Special handling for get_files_task_id_tool if it doesn't take generic args
+                        if tool_name == "get_files_task_id_tool":
+                            # Ensure task_id is passed correctly, not from LLM args unless intended
+                            observation = self.tools[tool_name](task_id)
+                        else:
+                            observation = self.tools[tool_name](**tool_args_dict)
+
+                        print(f"Tool observation: {str(observation)[:200]}...")
+                        conversation.append({"role": "assistant", "content": llm_response}) # LLM's tool request
+                        conversation.append({"role": "user", "content": f"Observation: {observation}"}) # Tool result
                     else:
-                        args = eval(args_str)
-                        observation = self.tools[tool_name](args)
-                    conversation.append({"role": "assistant", "content": response})
-                    conversation.append({"role": "user", "content": f"Observation: {observation}"})
+                        print(f"Error: Unknown tool name: {tool_name}")
+                        conversation.append({"role": "user", "content": f"Error: Unknown tool '{tool_name}'. Available tools are: {', '.join(self.tools.keys())}."})
+                except json.JSONDecodeError as e:
+                    print(f"Error decoding JSON for tool call: {e} - String was: {tool_call_str}")
+                    conversation.append({"role": "user", "content": f"Error: Invalid tool call format. Expected JSON. {e}"})
                 except Exception as e:
-                    conversation.append({"role": "user", "content": f"Error executing tool: {e}"})
+                    print(f"Error executing tool or processing its call: {e}")
+                    conversation.append({"role": "assistant", "content": llm_response}) # LLM's tool request
+                    conversation.append({"role": "user", "content": f"Error executing tool {tool_name}: {e}"})
             else:
-                conversation.append({"role": "user", "content": "Please provide a tool call or the final answer."})
-        print(f"Agent returning answer: {answer}")
-        return answer
-    
+                # If the LLM doesn't use the specified prefixes, treat its response as a potential direct answer or a misstep.
+                # Could also be a clarification question from the LLM.
+                print(f"LLM response did not start with ANSWER: or TOOL:. Treating as intermediate thought or error. Response: {llm_response[:100]}")
+                # Adding it as an assistant message and prompting for a structured response
+                conversation.append({"role": "assistant", "content": llm_response})
+                conversation.append({"role": "user", "content": "Please respond with either 'TOOL: {\"name\": \"tool_name\", \"args\": {}}' or 'ANSWER: your_final_answer'."})
+
+            if loop_count == max_loops - 1:
+                print("Agent reached max loops. Returning last LLM response or error.")
+                return f"Error: Agent reached maximum iteration limit. Last response: {llm_response}"
+
+        return "Error: Agent loop completed without returning an answer."
+
+
+# --- Tool Definitions ---
 def web_search_tool(search_terms: str) -> str:
     """
     Retrieves information from the internet using DuckDuckGo Search and returns results in JSON format.
-    
-    Args:
-        search_terms (str): The search query to look up.
-    
-    Returns:
-        str: JSON string containing search results.
-    
-    Example:
-        >>> web_search_tool("H. pylori trial NIH")
-        '{"results": [{"title": "H. pylori Trial", "snippet": "90 patients enrolled Jan-May 2018"}]}'
+    Args: search_terms (str): The search query to look up.
+    Returns: str: JSON string containing search results.
     """
+    print(f"Web search tool called with terms: {search_terms}")
     try:
         with DDGS() as ddgs:
             results = [r for r in ddgs.text(search_terms, max_results=3)]
-            return str({"results": results})
+        return json.dumps({"results": results}) # Ensure it's a JSON string
     except Exception as e:
-        return str({"error": f"Search failed: {str(e)}"})
+        print(f"Web search failed: {e}")
+        return json.dumps({"error": f"Search failed: {str(e)}"})
 
 def decimal_approximation_tool(number: float, decimals: int = 1) -> float:
     """
     Adjusts a numerical answer to the specified number of decimal places.
-    
     Args:
         number (float): The number to round.
         decimals (int): Number of decimal places to round to (default is 1).
-    
-    Returns:
-        float: The rounded number.
-    
-    Example:
-        >>> decimal_approximation_tool(4.567, 1)
-        4.6
+    Returns: float: The rounded number.
     """
-    return round(float(number), decimals)
+    print(f"Decimal approximation tool called with number: {number}, decimals: {decimals}")
+    try:
+        return round(float(number), int(decimals))
+    except Exception as e:
+        print(f"Decimal approximation failed: {e}")
+        return f"Error in decimal_approximation_tool: {str(e)}" # Return error as string
 
 def get_files_task_id_tool(task_id: str) -> str:
     """
     Downloads the file associated with the given task_id by making an API call.
-    
-    Args:
-        task_id (str): The ID of the task to fetch the file for.
-    
-    Returns:
-        str: The file content as a string (e.g., text or image description), or an error message.
-    
-    Example:
-        >>> get_files_task_id_tool("task_1")
-        'Nutrition facts: Calories 390, Butterfat 11%'
+    Args: task_id (str): The ID of the task to fetch the file for.
+    Returns: str: The file content as a string or an error message.
     """
+    print(f"Get files tool called with task_id: {task_id}")
     try:
-        response = requests.get(f"{DEFAULT_API_URL}/files/{task_id}")
+        # Assuming DEFAULT_API_URL is the base for the /files endpoint
+        file_url = f"{DEFAULT_API_URL}/files/{task_id}"
+        response = requests.get(file_url, timeout=30)
         if response.status_code == 200:
-            return response.text  # Assuming the API returns the file content as a string
+            return response.text
         else:
-            return f"Error fetching file for task_id {task_id}: Status {response.status_code}"
+            return f"Error fetching file for task_id {task_id}: Status {response.status_code}, Response: {response.text}"
     except Exception as e:
+        print(f"Error in get_files_task_id_tool: {e}")
         return f"Error fetching file for task_id {task_id}: {str(e)}"
 
-def image_processing_tool(file_content: str) -> str:
-    """
-    Processes an image file (or its description) and extracts text using OCR.
-    
-    Args:
-        file_content (str): The file content or description (e.g., from get_files_task_id_tool).
-    
-    Returns:
-        str: Extracted text from the image or description.
-    
-    Example:
-        >>> image_processing_tool("Nutrition facts: Calories 390, Butterfat 11%")
-        'Calories: 390, Butterfat: 11%'
-    """
-    # Since API returns a string, we simulate OCR on the description for now
-    # Placeholder for real OCR: If file_content were an image path or bytes, we'd use pytesseract
-    try:
-        # Simulated OCR processing on the string (as a placeholder)
-        # In a real scenario, file_content would be an image file path or bytes
-        extracted_text = file_content.replace("Nutrition facts: ", "")  # Mock extraction
-        return f"Extracted: {extracted_text}"
-    except Exception as e:
-        return f"OCR error: {str(e)}"
-    # Uncomment below for actual OCR when file_content is an image path
-    """
-    try:
-        from PIL import Image
-        import pytesseract
-        image = Image.open(file_content)  # Assuming file_content is a path to an image
-        extracted_text = pytesseract.image_to_string(image)
-        return f"Extracted: {extracted_text.strip()}"
-    except Exception as e:
-        return f"OCR error: {str(e)}"
-    """
-
-def run_and_submit_all( profile: gr.OAuthProfile | None):
+# --- Gradio App ---
+def run_and_submit_all(profile: gr.OAuthProfile | None):
     """
     Fetches all questions, runs the BasicAgent on them, submits all answers,
     and displays the results.
     """
-    # --- Determine HF Space Runtime URL and Repo URL ---
-    space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
+    space_id = os.getenv("SPACE_ID")
 
     if profile:
-        username= f"{profile.username}"
+        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
 
-    api_url = DEFAULT_API_URL
-    questions_url = f"{api_url}/questions"
-    submit_url = f"{api_url}/submit"
+    scorer_api_url = DEFAULT_API_URL
+    questions_url = f"{scorer_api_url}/questions"
+    submit_url = f"{scorer_api_url}/submit"
 
-    # 1. Instantiate Agent ( modify this part to create your agent)
     try:
         agent = BasicAgent()
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
-    # In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)
-    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
-    print(agent_code)
 
-    # 2. Fetch Questions
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else "local_run_code_link_not_available"
+    print(f"Agent code link: {agent_code}")
+
     print(f"Fetching questions from: {questions_url}")
     try:
         response = requests.get(questions_url, timeout=15)
         response.raise_for_status()
         questions_data = response.json()
         if not questions_data:
-             print("Fetched questions list is empty.")
-             return "Fetched questions list is empty or invalid format.", None
+            print("Fetched questions list is empty.")
+            return "Fetched questions list is empty or invalid format.", None
         print(f"Fetched {len(questions_data)} questions.")
     except requests.exceptions.RequestException as e:
         print(f"Error fetching questions: {e}")
         return f"Error fetching questions: {e}", None
     except requests.exceptions.JSONDecodeError as e:
-         print(f"Error decoding JSON response from questions endpoint: {e}")
-         print(f"Response text: {response.text[:500]}")
-         return f"Error decoding server response for questions: {e}", None
+        print(f"Error decoding JSON response from questions endpoint: {e}. Response text: {response.text[:500]}")
+        return f"Error decoding server response for questions: {e}", None
     except Exception as e:
         print(f"An unexpected error occurred fetching questions: {e}")
         return f"An unexpected error occurred fetching questions: {e}", None
 
-    # 3. Run your Agent
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
@@ -217,23 +259,24 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
-            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})
+            # Pass the whole item dictionary to the agent
+            submitted_answer = agent(item)
+            answers_payload.append({"task_id": task_id, "submitted_answer": str(submitted_answer)}) # Ensure answer is string
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": str(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}"})
+            print(f"Error running agent on task {task_id}: {e}")
+            import traceback
+            traceback.print_exc() # Print full traceback for agent errors
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
 
     if not answers_payload:
         print("Agent did not produce any answers to submit.")
         return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
 
-    # 4. Prepare Submission 
     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)
 
-    # 5. Submit
     print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
     try:
         response = requests.post(submit_url, json=submission_data, timeout=60)
@@ -247,8 +290,6 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             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:
@@ -256,63 +297,54 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             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
+        final_status = f"Submission Failed: {error_detail}"
+        print(final_status)
     except requests.exceptions.Timeout:
-        status_message = "Submission Failed: The request timed out."
-        print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
+        final_status = "Submission Failed: The request timed out."
+        print(final_status)
     except requests.exceptions.RequestException as e:
-        status_message = f"Submission Failed: Network error - {e}"
-        print(status_message)
-        results_df = pd.DataFrame(results_log)
-        return status_message, results_df
+        final_status = f"Submission Failed: Network error - {e}"
+        print(final_status)
     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
+        final_status = f"An unexpected error occurred during submission: {e}"
+        print(final_status)
+
+    results_df = pd.DataFrame(results_log)
+    return final_status, results_df
 
 
-# --- Build Gradio Interface using Blocks ---
 with gr.Blocks() as demo:
     gr.Markdown("# Basic Agent Evaluation Runner")
     gr.Markdown(
         """
         **Instructions:**
 
-        1.  Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
-        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.
+        1.  Ensure your `TEST_AGENT_KEY` (OpenRouter API Key) is set in your Hugging Face Space secrets or `.env` file.
+        2.  Modify `prompt.txt` to guide the agent, especially for tool use and answer formatting. Remove references to the old `image_processing_tool`.
+        3.  Log in to your Hugging Face account using the button below.
+        4.  Click 'Run Evaluation & Submit All Answers'.
 
         ---
         **Disclaimers:**
-        Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
-        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
+        Agent execution can take time. This setup is a starting point.
         """
     )
 
     gr.LoginButton()
-
     run_button = gr.Button("Run Evaluation & Submit All Answers")
-
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
-    # Removed max_rows=10 from DataFrame constructor
     results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
 
     run_button.click(
         fn=run_and_submit_all,
-        outputs=[status_output, results_table]
+        outputs=[status_output, results_table],
+        api_name="run_evaluation" # Added api_name for programmatic access if needed
     )
 
 if __name__ == "__main__":
     print("\n" + "-"*30 + " App Starting " + "-"*30)
-    # Check for SPACE_HOST and SPACE_ID at startup for information
     space_host_startup = os.getenv("SPACE_HOST")
-    space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
+    space_id_startup = os.getenv("SPACE_ID")
 
     if space_host_startup:
         print(f"✅ SPACE_HOST found: {space_host_startup}")
@@ -320,7 +352,7 @@ if __name__ == "__main__":
     else:
         print("ℹ️  SPACE_HOST environment variable not found (running locally?).")
 
-    if space_id_startup: # Print repo URLs if SPACE_ID is found
+    if space_id_startup:
         print(f"✅ SPACE_ID found: {space_id_startup}")
         print(f"   Repo URL: https://huggingface.co/spaces/{space_id_startup}")
         print(f"   Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
@@ -328,6 +360,5 @@ if __name__ == "__main__":
         print("ℹ️  SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
 
     print("-"*(60 + len(" App Starting ")) + "\n")
-
     print("Launching Gradio Interface for Basic Agent Evaluation...")
     demo.launch(debug=True, share=False)