better score

app.py

@@ -7,6 +7,7 @@ import io
 import contextlib
 from typing import TypedDict, Annotated
 import torch
+import json # For robust tool call parsing/generation if needed
 
 # --- Multimodal & Web Tool Imports ---
 from transformers import pipeline
@@ -16,7 +17,7 @@ from bs4 import BeautifulSoup
 
 # --- LangChain & LangGraph Imports ---
 from langgraph.graph.message import add_messages
-from langchain_core.messages import AnyMessage, HumanMessage, AIMessage, ToolMessage
+from langchain_core.messages import AnyMessage, HumanMessage, AIMessage, ToolMessage, SystemMessage
 from langgraph.prebuilt import ToolNode
 from langgraph.graph import START, StateGraph
 from langgraph.prebuilt import tools_condition
@@ -27,13 +28,13 @@ from langchain_core.tools import tool
 
 # (Keep Constants as is)
 # --- Constants ---
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+DEFAULT_API_URL = "[https://agents-course-unit4-scoring.hf.space](https://agents-course-unit4-scoring.hf.space)"
 
 # --- Initialize ASR Pipeline (for Audio Tool) ---
 # Load the model once when the app starts for efficiency
 try:
     asr_pipeline = pipeline(
-        "automatic-speech-recognition", 
+        "automatic-speech-recognition",
         model="openai/whisper-base",
         torch_dtype=torch.float16, # Use float16 for faster inference
         device_map="auto" # Use GPU if available
@@ -47,7 +48,7 @@ except Exception as e:
 
 @tool
 def search_tool(query: str) -> str:
-    """Calls DuckDuckGo search and returns the results."""
+    """Calls DuckDuckGo search and returns the results. Use this for recent information or general web searches."""
     print(f"--- Calling Search Tool with query: {query} ---")
     try:
         search = DuckDuckGoSearchRun()
@@ -59,61 +60,78 @@ def search_tool(query: str) -> str:
 def code_interpreter(code: str) -> str:
     """
     Executes a string of Python code and returns its stdout, stderr, and any error.
-    Use this for calculations, data manipulation, or any other Python operation.
-    The code runs in a sandboxed environment.
-    Note: 'pandas' and 'openpyxl' are available.
+    Use this for calculations, data manipulation (including pandas on dataframes read from files), list operations, string manipulations, or any other Python operation.
+    The code runs in a sandboxed environment. 'pandas' (as pd) and 'openpyxl' are available.
+    Ensure the code is complete and executable. If printing, use print().
     """
     print(f"--- Calling Code Interpreter with code:\n{code}\n---")
     output_stream = io.StringIO()
     error_stream = io.StringIO()
-    
+
     try:
         # Use contextlib to redirect stdout and stderr
         with contextlib.redirect_stdout(output_stream), contextlib.redirect_stderr(error_stream):
             # Execute the code. Provide 'pd' (pandas) in the globals
             exec(code, {"pd": pd}, {})
-        
+
         stdout = output_stream.getvalue()
         stderr = error_stream.getvalue()
-        
+
         if stderr:
             return f"Error: {stderr}\nStdout: {stdout}"
-        return f"Success:\n{stdout}"
-    
+        if stdout:
+             return f"Success:\n{stdout}"
+        return "Success: Code executed without error and produced no stdout."
+
     except Exception as e:
         # Capture any exception during exec
         return f"Execution failed with error: {str(e)}"
 
 @tool
 def read_file(path: str) -> str:
-    """Reads the content of a file at the specified path."""
+    """Reads the content of a file at the specified path. Use this to examine files provided in the question."""
     print(f"--- Calling Read File Tool at path: {path} ---")
     try:
-        with open(path, 'r', encoding='utf-8') as f:
+        # Try finding the file relative to the app directory first
+        script_dir = os.path.dirname(__file__)
+        full_path = os.path.join(script_dir, path)
+        if not os.path.exists(full_path):
+             # If not found, try the direct path (might be absolute or relative to cwd)
+             full_path = path
+             if not os.path.exists(full_path):
+                 # Try basename for GAIA questions providing just the filename
+                 if os.path.exists(os.path.basename(path)):
+                      full_path = os.path.basename(path)
+                 else:
+                     return f"Error: File not found at '{path}', '{os.path.join(script_dir, path)}', or '{os.path.basename(path)}'"
+
+        with open(full_path, 'r', encoding='utf-8') as f:
             return f.read()
     except Exception as e:
         return f"Error reading file {path}: {str(e)}"
 
 @tool
 def write_file(path: str, content: str) -> str:
-    """Writes the given content to a file at the specified path."""
+    """Writes the given content to a file at the specified path. Creates directories if they don't exist."""
     print(f"--- Calling Write File Tool at path: {path} ---")
     try:
         # Ensure the directory exists
-        os.makedirs(os.path.dirname(path), exist_ok=True)
-        
-        with open(path, 'w', encoding='utf-8') as f:
+        full_path = os.path.join(os.path.dirname(__file__), path) # Write relative to script dir
+        os.makedirs(os.path.dirname(full_path), exist_ok=True)
+
+        with open(full_path, 'w', encoding='utf-8') as f:
             f.write(content)
-        return f"Successfully wrote to file {path}."
+        return f"Successfully wrote to file {path} (relative to app)."
     except Exception as e:
         return f"Error writing to file {path}: {str(e)}"
 
 @tool
 def list_directory(path: str = ".") -> str:
-    """Lists the contents of a directory at the specified path."""
+    """Lists the contents (files and directories) of a directory at the specified path relative to the app."""
     print(f"--- Calling List Directory Tool at path: {path} ---")
     try:
-        files = os.listdir(path)
+        full_path = os.path.join(os.path.dirname(__file__), path) # List relative to script dir
+        files = os.listdir(full_path)
         return "\n".join(files) if files else "Directory is empty."
     except Exception as e:
         return f"Error listing directory {path}: {str(e)}"
@@ -121,21 +139,28 @@ def list_directory(path: str = ".") -> str:
 @tool
 def audio_transcription_tool(file_path: str) -> str:
     """
-    Transcribes an audio file (like .mp3 or .wav) and returns the text.
+    Transcribes an audio file (like .mp3 or .wav) using Whisper and returns the text content.
+    Use this for questions involving audio file analysis.
     """
     print(f"--- Calling Audio Transcription Tool at path: {file_path} ---")
     if not asr_pipeline:
         return "Error: Audio transcription pipeline is not available."
     try:
-        if not os.path.exists(file_path):
-            # GAIA questions might provide relative paths, so we check
-            if os.path.exists(os.path.basename(file_path)):
-                file_path = os.path.basename(file_path)
-            else:
-                return f"Error: File not found at {file_path}"
-        
+        # Try finding the file relative to the app directory first
+        script_dir = os.path.dirname(__file__)
+        full_path = os.path.join(script_dir, file_path)
+        if not os.path.exists(full_path):
+             # If not found, try the direct path
+             full_path = file_path
+             if not os.path.exists(full_path):
+                 # Try basename for GAIA questions
+                 if os.path.exists(os.path.basename(file_path)):
+                     full_path = os.path.basename(file_path)
+                 else:
+                     return f"Error: Audio file not found at '{file_path}', '{os.path.join(script_dir, file_path)}', or '{os.path.basename(file_path)}'"
+
         # The pipeline handles file loading
-        transcription = asr_pipeline(file_path)
+        transcription = asr_pipeline(full_path)
         print("--- Transcription Complete ---")
         return transcription["text"]
     except Exception as e:
@@ -144,47 +169,72 @@ def audio_transcription_tool(file_path: str) -> str:
 @tool
 def get_youtube_transcript(video_url: str) -> str:
     """
-    Fetches the transcript for a given YouTube video URL.
+    Fetches the transcript for a given YouTube video URL. Use this for questions about YouTube video content.
     """
     print(f"--- Calling YouTube Transcript Tool for URL: {video_url} ---")
     try:
-        # Extract video ID from URL
-        video_id = video_url.split("v=")[1].split("&")[0]
+        # Extract video ID from URL more robustly
+        video_id = None
+        if "watch?v=" in video_url:
+            video_id = video_url.split("v=")[1].split("&")[0]
+        elif "youtu.be/" in video_url:
+            video_id = video_url.split("youtu.be/")[1].split("?")[0]
+
+        if not video_id:
+             return f"Error: Could not extract video ID from URL: {video_url}"
+
         transcript_list = YouTubeTranscriptApi.get_transcript(video_id)
-        
+
         # Combine all transcript parts into one string
         full_transcript = " ".join([item["text"] for item in transcript_list])
         print("--- Transcript Fetched ---")
-        return full_transcript
+        # Return a limited amount to avoid overwhelming the context
+        return full_transcript[:8000]
     except Exception as e:
         return f"Error fetching YouTube transcript: {str(e)}"
 
 @tool
 def scrape_web_page(url: str) -> str:
     """
-    Fetches the full text content of a given web page URL.
+    Fetches the primary text content of a given web page URL, removing navigation, footer, scripts, and styles.
+    Use this when you need the full content of a webpage found via search.
     """
     print(f"--- Calling Web Scraper Tool for URL: {url} ---")
     try:
-        response = requests.get(url, headers={'User-Agent': 'Mozilla/5.0'}, timeout=10)
-        response.raise_for_status() # Raise an error for bad responses
-        
+        headers = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.3'}
+        response = requests.get(url, headers=headers, timeout=15) # Increased timeout
+        response.raise_for_status() # Raise an error for bad responses (4xx or 5xx)
+
+        # Check content type to avoid parsing non-HTML
+        if 'html' not in response.headers.get('Content-Type', '').lower():
+            return f"Error: URL {url} did not return HTML content."
+
         soup = BeautifulSoup(response.text, 'html.parser')
-        
-        # Remove script/style tags
-        for script in soup(["script", "style", "nav", "footer", "aside"]):
-            script.extract()
-            
-        text = soup.get_text()
-        
-        # Clean up whitespace
+
+        # Remove common non-content tags
+        for tag in soup(["script", "style", "nav", "footer", "aside", "header", "form"]):
+            tag.extract()
+
+        # Attempt to find the main content area (heuristics, may not always work)
+        main_content = soup.find('main') or soup.find('article') or soup.find('div', role='main') or soup.body
+        if not main_content:
+             main_content = soup # Fallback to the whole soup if no main area found
+
+        text = main_content.get_text(separator='\n', strip=True)
+
+        # Clean up excessive whitespace
         lines = (line.strip() for line in text.splitlines())
         chunks = (phrase.strip() for line in lines for phrase in line.split("  "))
         text = '\n'.join(chunk for chunk in chunks if chunk)
+
         print("--- Web Page Scraped ---")
-        return text[:8000] # Return first 8000 chars to avoid overload
+        # Limit context size
+        return text[:8000]
+
+    except requests.exceptions.RequestException as e:
+         return f"Error fetching web page {url}: {str(e)}"
     except Exception as e:
-        return f"Error scraping web page: {str(e)}"
+        return f"Error scraping web page {url}: {str(e)}"
 
 # --- End of Tool Definitions ---
 
@@ -197,12 +247,11 @@ class AgentState(TypedDict):
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 class BasicAgent:
-    
+
     def __init__(self):
         print("BasicAgent (LangGraph) initialized.")
-        
+
         # 1. Get API Token from Space Secrets
-        # Go to your Space's Settings -> Secrets and add HUGGINGFACEHUB_API_TOKEN
         HUGGINGFACEHUB_API_TOKEN = os.getenv("HUGGINGFACEHUB_API_TOKEN")
         if not HUGGINGFACEHUB_API_TOKEN:
             raise ValueError("HUGGINGFACEHUB_API_TOKEN secret is not set! Please add it to your Space secrets.")
@@ -218,264 +267,27 @@ class BasicAgent:
             get_youtube_transcript,
             scrape_web_page
         ]
-        
-        # 3. Initialize the LLM
-        # We wrap HuggingFaceEndpoint in ChatHuggingFace for LangChain compatibility
-        llm = HuggingFaceEndpoint(
-            repo_id="HuggingFaceH4/zephyr-7b-beta", # A good, fast model for tool use
-            # repo_id="Qwen/Qwen2.5-Coder-32B-Instruct", # Your chosen model
-            huggingfacehub_api_token=HUGGINGFACEHUB_API_TOKEN,
-            max_new_tokens=1500,
-            temperature=0.1,
-        )
-        chat_llm = ChatHuggingFace(llm=llm)
-
-        # 4. Bind tools to the LLM
-        self.llm_with_tools = chat_llm.bind_tools(self.tools)
-
-        # 5. Define the Agent Node
-        def agent_node(state: AgentState):
-            print("--- Running Agent Node ---")
-            ai_message = self.llm_with_tools.invoke(state["messages"])
-            print(f"AI Message: {ai_message.pretty_repr()}")
-            return {"messages": [ai_message]}
-
-        # 6. Define the Tool Node
-        tool_node = ToolNode(self.tools)
-        
-        # 7. Create the Graph
-        graph_builder = StateGraph(AgentState)
-        
-        # Add the nodes
-        graph_builder.add_node("agent", agent_node)
-        graph_builder.add_node("tools", tool_node)
-        
-        # Define the edges
-        graph_builder.add_edge(START, "agent")
-        
-        # Add the conditional edge
-        graph_builder.add_conditional_edges(
-            "agent",
-            tools_condition,
-            {
-                "tools": "tools",
-                "__end__": "__end__",
-            },
-        )
-        graph_builder.add_edge("tools", "agent")
-        
-        # 8. Compile the graph and store it
-        self.graph = graph_builder.compile()
-        print("Graph compiled successfully with all tools.")
-
-    def __call__(self, question: str) -> str:
-        print(f"Agent received question (first 50 chars): {question[:50]}...")
-        
-        # Prepare the input for the graph
-        graph_input = {"messages": [HumanMessage(content=question)]}
-        
-        final_answer = ""
-        
-        # Stream the graph's execution
-        try:
-            # We use stream_mode="values" to get the full state at each step
-            for event in self.graph.stream(graph_input, stream_mode="values"):
-                last_message = event["messages"][-1]
-                
-                # Update the final answer with the latest AI message
-                if isinstance(last_message, AIMessage):
-                    if last_message.content:
-                        print(f"AI: {last_message.content[:200]}...")
-                        final_answer = last_message.content
-                elif isinstance(last_message, ToolMessage):
-                     print(f"Tool Result: {last_message.content[:200]}...")
-
-            print(f"Agent returning final answer: {final_answer}")
-            return final_answer
-        
-        except Exception as e:
-            print(f"Error running agent graph: {e}")
-            return f"AGENT ERROR: {e}"
-
-
-# --- (Original Template Code Starts Here) ---
-
-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
-    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
-    
-    api_url = DEFAULT_API_URL
-    questions_url = f"{api_url}/questions"
-    submit_url = f"{api_url}/submit"
-
-    # 1. Instantiate Agent ( modify this part to create your agent)
-    print("Initializing agent...")
-    try:
-        agent = BasicAgent()
-    except Exception as e:
-        print(f"Error instantiating agent: {e}")
-        return f"Error initializing agent: {e}", None
-    print("Agent initialized successfully.")
-
-    # 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(f"Agent code URL: {agent_code}")
-
-    # 2. Fetch Questions
-    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(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
-    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...")
-    
-    # Set a limit for testing. Remove '[:question_limit]' for the full submission.
-    # question_limit = 10 
-    
-    for item in questions_data: # [:question_limit]: # Using limit here
-        task_id = item.get("task_id")
-        question_text = item.get("question")
-        if not task_id or question_text is None:
-            print(f"Skipping item with missing task_id or question: {item}")
-            continue
-        
-        print(f"\n--- Running Task {task_id} ---")
-        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})
-            print(f"--- Task {task_id} Complete ---")
-        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}"})
-    
-    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)
-        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 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
-    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
-    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
-
-# --- 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.
-        ---
-        **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.
-        """
-    )
-    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]
-    )
-
-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
-    
-    if space_host_startup:
-        print(f"✅ SPACE_HOST found: {space_host_startup}")
-        print(f"   Runtime URL should be: https://{space_host_startup}.hf.space")
-    else:
-        print("ℹ️  SPACE_HOST environment variable not found (running locally?).")
-    
-    if space_id_startup: # Print repo URLs if SPACE_ID is found
-        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")
-    else:
-        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)
 
+        # 3. Define the Improved System Prompt
+        tool_descriptions = "\n".join([f"- {tool.name}: {tool.description}" for tool in self.tools])
+        self.system_prompt = f"""You are a highly intelligent and meticulous AI assistant built to answer questions from the GAIA benchmark.
+Your primary goal is to provide **only the concise, factual, and direct answer** to the user's question, exactly matching the format required by the benchmark (e.g., a name, a number, a specific string format, a comma-separated list).
+
+**CRITICAL INSTRUCTIONS:**
+* **DO NOT** include conversational filler (e.g., "Sure, I can help...", "The answer is...", "Here is the information...").
+* **DO NOT** explain your reasoning or the steps you took unless the question *explicitly* asks for it.
+* **DO NOT** repeat the question in your final answer.
+* **FINAL ANSWER FORMAT:** Your final response must contain *only* the answer itself.
+
+You have access to the following tools to gather information and perform actions:
+{tool_descriptions}
+
+**TOOL USAGE PROTOCOL:**
+* To use a tool, you MUST respond ONLY with a single JSON object formatted exactly like this:
+    ```json
+    {{
+      "tool": "tool_name",
+      "tool_input": {{ "arg_name1": "value1", "arg_name2": "value2", ... }}
+    }}
+    ```
+    """

requirements.txt

@@ -12,4 +12,5 @@ torchaudio
 librosa
 youtube-transcript-api
 beautifulsoup4
-openpyxl
+openpyxl
+accelerate