app.py

import os
import gradio as gr
import requests
import inspect
import pandas as pd
from smolagents import tool, Tool, CodeAgent, DuckDuckGoSearchTool, HfApiModel, ToolCallingAgent, VisitWebpageTool, SpeechToTextTool, FinalAnswerTool
from dotenv import load_dotenv
import heapq
from collections import Counter
import re
from io import BytesIO
from youtube_transcript_api import YouTubeTranscriptApi
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_community.document_loaders import WikipediaLoader
from langchain_community.utilities import WikipediaAPIWrapper
from langchain_community.document_loaders import ArxivLoader

# (Keep Constants as is)
# --- Constants ---
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"

#Load environment variables
load_dotenv()

import io
import contextlib
import traceback
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from smolagents import Tool, CodeAgent, DuckDuckGoSearchTool, FinalAnswerTool, HfApiModel


class CodeLlamaTool(Tool):
    name = "code_llama_tool"
    description = "Solves reasoning/code questions using Meta Code Llama 7B Instruct"

    inputs = {
        "question": {
            "type": "string",
            "description": "The question requiring code-based or reasoning-based solution"
        }
    }
    output_type = "string"

    def __init__(self):
        self.model_id = "codellama/CodeLlama-7b-Instruct-hf"
        token = os.getenv("HF_TOKEN")

        self.tokenizer = AutoTokenizer.from_pretrained(self.model_id, token=token)
        self.model = AutoModelForCausalLM.from_pretrained(
            self.model_id,
            device_map="auto",
            torch_dtype="auto",
            token=token
        )
        self.pipeline = pipeline(
            "text-generation",
            model=self.model,
            tokenizer=self.tokenizer,
            max_new_tokens=512,
            temperature=0.2,
            truncation=True
        )

    def forward(self, question: str) -> str:
        prompt = f"""You are an AI that uses Python code to answer questions.

Question: {question}

Instructions:
- If solving requires code, use a block like <tool>code</tool>.
- Always end with <final>FINAL ANSWER</final> containing the final number or string.

Example:
Question: What is 5 * sqrt(36)?
Answer:
<tool>
import math
print(5 * math.sqrt(36))
</tool>
<final>30.0</final>

Answer:"""

        response = self.pipeline(prompt)[0]["generated_text"]
        return self.parse_and_execute(response)

    def parse_and_execute(self, response: str) -> str:
        try:
            # Extract and run code if exists
            if "<tool>" in response and "</tool>" in response:
                code = response.split("<tool>")[1].split("</tool>")[0].strip()
                result = self._run_code(code)
                return f"FINAL ANSWER (code output): {result}"

            # Extract final result directly
            elif "<final>" in response and "</final>" in response:
                final = response.split("<final>")[1].split("</final>")[0].strip()
                return f"FINAL ANSWER: {final}"

            return f"Could not extract final answer.\n\n{response}"

        except Exception as e:
            return f"Error in parse_and_execute: {str(e)}\n\nFull response:\n{response}"

    def _run_code(self, code: str) -> str:
        buffer = io.StringIO()
        try:
            with contextlib.redirect_stdout(buffer):
                exec(code, {})
            return buffer.getvalue().strip()
        except Exception:
            return f"Error executing code:\n{traceback.format_exc()}"



#from smolagents import Tool
#from langchain_community.document_loaders import WikipediaLoader

class WikiSearchTool(Tool):
    name = "wiki_search"
    description = "Search Wikipedia for a query and return up to 2 results."
    inputs = {
        "query": {"type": "string", "description": "The search term for Wikipedia."}
    }
    output_type = "string"

    def forward(self, query: str) -> str:
        search_docs = WikipediaLoader(query=query, load_max_docs=2).load()

        formatted_search_docs = "\n\n---\n\n".join(
            [
                f'<Document source="{doc.metadata.get("source", "Wikipedia")}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
                for doc in search_docs
            ]
        )
        return formatted_search_docs

    


class StringReverseTool(Tool):
    name = "reverse_message"
    description = "When you received a strange text, try to reverse it and perform action described in reversed message."
    inputs = {
        "message": {
            "type": "string",
            "description": "A message, which looks like strange and can be reversed to get actions to execute."
        }
    }
    output_type = "string"

    def __init__(self):
        return

    def forward(self, message: str):
        return message[::-1]

class KeywordsExtractorTool(Tool):
    """Extracts top 5 keywords from a given text based on frequency."""

    name = "keywords_extractor"
    description = "This tool returns the 5 most frequent keywords occur in provided block of text."

    inputs = {
        "text": {
            "type": "string",
            "description": "Text to analyze for keywords.",
        }
    }
    output_type = "string"

    def forward(self, text: str) -> str:
        try:
            all_words = re.findall(r'\b\w+\b', text.lower())
            conjunctions = {'a', 'and', 'of', 'is', 'in', 'to', 'the'}
            filtered_words = []
            for w in all_words:
                if w not in conjunctions:
                    filtered_words.push(w)
            word_counts = Counter(filtered_words)
            k = 5
            return heapq.nlargest(k, word_counts.items(), key=lambda x: x[1])
        except Exception as e:
            return f"Error during extracting most common words: {e}"

import requests
import pandas as pd
from io import BytesIO
from smolagents import Tool # Make sure to import the Tool base class

class ParseExcelToJsonTool(Tool):
    """
    A tool for fetching and parsing an Excel file into structured JSON data.
    """
    name = "parse_excel_to_json"
    description = (
        "For a given task_id, fetches an Excel file from a remote URL, "
        "parses its sheets, and returns the data as a structured JSON object. "
        "Each sheet's data is returned as a list of dictionaries, with each dictionary "
        "representing a row (limited to the first 20 rows). "
        "Useful for extracting structured information from Excel files."
    )
    inputs = {
        "task_id": {
            "type": "string",
            "description": "The task ID used to construct the URL for fetching the Excel file.",
        }
    }
    output_type = "json" # The tool returns a dictionary, so "json" is the appropriate output_type

    def _run(self, task_id: str) -> dict:
        """
        Fetches and parses an Excel file from a URL based on the task_id.
        """
        url = f"https://agents-course-unit4-scoring.hf.space/files/{task_id}"

        try:
            response = requests.get(url, timeout=100)
            response.raise_for_status() # Raise an HTTPError for bad responses (4xx or 5xx)

            xls_content = pd.ExcelFile(BytesIO(response.content))
            json_sheets = {}

            for sheet_name in xls_content.sheet_names:
                df = xls_content.parse(sheet_name)
                df = df.dropna(how="all") # Drop rows that are entirely NaN
                
                # Limit to the first 20 rows for efficiency and to prevent overwhelming context
                rows = df.head(20).to_dict(orient="records") 
                json_sheets[sheet_name] = rows

            return {
                "task_id": task_id,
                "sheets": json_sheets,
                "status": "Success"
            }

        except requests.exceptions.RequestException as e:
            return {
                "task_id": task_id,
                "sheets": {},
                "status": f"Network or HTTP error: {str(e)}"
            }
        except Exception as e:
            return {
                "task_id": task_id,
                "sheets": {},
                "status": f"Error in parsing Excel file: {str(e)}"
            }

    # Optional: You can keep __call__ for direct instance calling, but it's handled by Tool base class
    # def __call__(self, task_id: str) -> dict:
    #     return self._run(task_id)


import os
from langchain_community.document_loaders import PyMuPDFLoader
from docx import Document as DocxDocument
import openpyxl

class AnalyseAttachmentTool(Tool):
    """
    A tool for analyzing various attachment types (PY, PDF, TXT, DOCX, XLSX)
    and extracting their text content.
    """

    name = "analyze_attachment"
    description = (
        "Analyzes attachments including PY, PDF, TXT, DOCX, and XLSX files and returns text content. "
        "Useful for understanding the content of various document types. "
        "The output is limited to the first 3000 characters for readability."
    )
    inputs = {
        "file_path": {
            "type": "string",
            "description": "Local path to the attachment file (e.g., 'documents/report.pdf').",
        }
    }
    output_type = "string"

    def forward(self, file_path: str) -> str:
        """
        Executes the attachment analysis. This method is called internally by the tool.
        """
        if not os.path.exists(file_path):
            return f"File not found: {file_path}"

        try:
            ext = os.path.splitext(file_path)[1].lower()
            content = ""

            if ext == ".pdf":
                loader = PyMuPDFLoader(file_path)
                documents = loader.load()
                content = "\n\n".join([doc.page_content for doc in documents])

            elif ext == ".txt" or ext == ".py":
                with open(file_path, "r", encoding="utf-8") as file:
                    content = file.read()

            elif ext == ".docx":
                doc = DocxDocument(file_path)
                content = "\n".join([para.text for para in doc.paragraphs])

            elif ext == ".xlsx":
                wb = openpyxl.load_workbook(file_path, data_only=True)
                for sheet in wb:
                    content += f"Sheet: {sheet.title}\n"
                    for row in sheet.iter_rows(values_only=True):
                        content += "\t".join([str(cell) if cell is not None else "" for cell in row]) + "\n"
            else:
                return "Unsupported file format. Please use PY, PDF, TXT, DOCX, or XLSX."

            return content[:3000]

        except Exception as e:
            return f"An error occurred while processing the file: {str(e)}"

    def __call__(self, file_path: str) -> str:
        """
        Makes the instance callable directly, invoking the _run method.
        """
        return self._run(file_path)



import os
import base64
import requests
from PIL import Image
from io import BytesIO


# Define image analysis tool
import requests

class ImageAnalysisTool(Tool):
    """
    A tool for analyzing images using a hosted Hugging Face model.
    """

    name = "image_analysis"
    description = (
        "Analyzes an image provided via a URL and returns a textual description of its content. "
        "This tool is useful for understanding the visual content of an image."
    )
    inputs = {
        "image_url": {
            "type": "string",
            "description": "The URL of the image to be analyzed (e.g., 'https://example.com/image.jpg').",
        }
    }
    output_type = "string"
    # You might consider making API_URL a class attribute if it's constant
    # or an instance attribute if it could vary per instance.
    # For this example, we'll keep it within the _run method for directness.

    def forward(self, image_url: str) -> str:
        """
        Executes the image analysis by sending the image URL to the Hugging Face API.
        """
        API_URL = "https://api-inference.huggingface.co/models/llava-hf/llava-1.5-7b-hf"
        
        try:
            response = requests.post(API_URL, json={"inputs": image_url})
            response.raise_for_status()  # Raise an HTTPError for bad responses (4xx or 5xx)
            
            # Assuming the response structure is always a list with a dictionary
            # and 'generated_text' is the key for the description.
            if response.json() and isinstance(response.json(), list) and 'generated_text' in response.json()[0]:
                return response.json()[0]['generated_text']
            else:
                return f"Unexpected API response format: {response.text}"
                
        except requests.exceptions.RequestException as e:
            return f"An error occurred during the API request: {e}"
        except IndexError:
            return "API response did not contain expected 'generated_text'."
        except Exception as e:
            return f"An unexpected error occurred: {e}"

    def __call__(self, image_url: str) -> str:
        """
        Makes the instance callable directly, invoking the _run method for convenience.
        """
        return self._run(image_url)


class VideoTranscriptionTool(Tool):
    """Fetch transcripts from YouTube videos"""
    name = "transcript_video"
    description = "Fetch text transcript from YouTube movies with optional timestamps"
    inputs = {
        "url": {"type": "string", "description": "YouTube video URL or ID"},
        "include_timestamps": {"type": "boolean", "description": "If timestamps should be included in output", "nullable": True}
    }
    output_type = "string"

    def forward(self, url: str, include_timestamps: bool = False) -> str:

        if "youtube.com/watch" in url:
            video_id = url.split("v=")[1].split("&")[0]
        elif "youtu.be/" in url:
            video_id = url.split("youtu.be/")[1].split("?")[0]
        elif len(url.strip()) == 11:  # Direct ID
            video_id = url.strip()
        else:
            return f"YouTube URL or ID: {url} is invalid!"

        try:
            transcription = YouTubeTranscriptApi.get_transcript(video_id)

            if include_timestamps:
                formatted_transcription = []
                for part in transcription:
                    timestamp = f"{int(part['start']//60)}:{int(part['start']%60):02d}"
                    formatted_transcription.append(f"[{timestamp}] {part['text']}")
                return "\n".join(formatted_transcription)
            else:
                return " ".join([part['text'] for part in transcription])

        except Exception as e:
            return f"Error in extracting YouTube transcript: {str(e)}"



class BasicAgent:
    def __init__(self):
        try:
            token = os.environ.get("HF_API_TOKEN")
            self.model = HfApiModel(temperature=0.1, token=token)
        except Exception as e:
            print("Error initializing model:")
            traceback.print_exc()
            raise

        try:
        # Initialize all tool instances
            self.search_tool = DuckDuckGoSearchTool()
            self.wiki_search_tool = WikiSearchTool() # Ensure this class is defined/imported
            self.str_reverse_tool = StringReverseTool() # Ensure this class is defined/imported
            self.keywords_extract_tool = KeywordsExtractorTool() # Ensure this class is defined/imported
            self.speech_to_text_tool = SpeechToTextTool() # Ensure this class is defined/imported
            self.visit_webpage_tool = VisitWebpageTool() # Ensure this class is defined/imported
            self.final_answer_tool = FinalAnswerTool()
        except Exception as e:
            print("Error initializing model:")
            traceback.print_exc()
            raise

        
        # Custom tools - ensure these classes are defined and imported
        self.video_transcription_tool = VideoTranscriptionTool()
        self.image_analysis_tool = ImageAnalysisTool() # Renamed for clarity
        self.analyse_attachment_tool = AnalyseAttachmentTool() # Renamed for clarity
        self.code_llama_tool = CodeLlamaTool() # Ensure this class is defined/imported
        self.parse_excel_to_json_tool = ParseExcelToJsonTool()
        
        
        system_prompt_template = """
You are my general AI assistant. Your task is to answer the question I asked.
First, provide an explanation of your reasoning, step by step, to arrive at the answer.
Then, return your final answer in a single line, formatted as follows: "FINAL ANSWER: [YOUR FINAL ANSWER]".
[YOUR FINAL ANSWER] should be a number, a string, or a comma-separated list of numbers and/or strings, depending on the question.
If the answer is a number, do not use commas or units (e.g., $, %) unless specified.
If the answer is a string, do not use articles or abbreviations (e.g., for cities), and write digits in plain text unless specified.
If the answer is a comma-separated list, apply the above rules for each element based on whether it is a number or a string.
"""

        # Create web agent with image analysis capability
        try:
            self.web_agent = ToolCallingAgent(
                tools=[
                    self.search_tool, # Use the initialized DuckDuckGoSearchTool instance
                    self.visit_webpage_tool,
                    self.image_analysis_tool
                ],
                model=self.model, # Use self.model
                max_steps=10,
                name="web_search_agent",
                description="Runs web searches and analyzes images",
            )
        except Exception as e:
            print("Error initializing model:")
            traceback.print_exc()
            raise

        # Create main agent with all capabilities
        try:
            self.agent = CodeAgent(
                model=self.model, # Use self.model
                tools=[
                    self.search_tool,
                    self.wiki_search_tool,
                    self.str_reverse_tool,
                    self.keywords_extract_tool,
                    self.speech_to_text_tool,
                    self.visit_webpage_tool,
                    self.final_answer_tool,
                    self.video_transcription_tool,
                    self.code_llama_tool,
                    self.parse_excel_to_json_tool,
                    self.image_analysis_tool, # Use the initialized instance
                    self.analyse_attachment_tool # Add the initialized attachment analysis tool
                ],
                add_base_tools=True # Consider what this adds, ensure it doesn't duplicate.
            )
        except Exception as e:
            print("Error initializing model:")
            traceback.print_exc()
            raise

        # Update system prompt
        # It's generally better to pass the system prompt directly if possible
        # or manage it through prompt templates defined by smolagents.
        # If smolagents adds its own system prompt, this appends to it.
        # self.agent.prompt_templates["system_prompt"] = self.agent.prompt_templates["system_prompt"] + system_prompt_template
        try:
            if hasattr(self.agent, "prompt_templates") and "system_prompt" in self.agent.prompt_templates:
                self.agent.prompt_templates["system_prompt"] += system_prompt_template
        except Exception as e:
            print("Error updating prompt:")
            traceback.print_exc()
            raise


    def __call__(self, question: str) -> str:
        print(f"Agent received question (first 50 chars): {question[:50]}...")
        
        # First try web agent for image-based queries
        if any(keyword in question.lower() for keyword in ["image", "picture", "photo", "screenshot", "diagram"]):
            print("Using web agent for image-related query")
            answer = self.web_agent.run(question)
        else:
            print("Using main agent")
            answer = self.agent.run(question)
            
        print(f"Agent returning answer: {answer}")
        return answer



        
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)
    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
    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...")
    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:
            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})
        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)