app.py

import os
import gradio as gr
import requests
import inspect
import pandas as pd
from smolagents import tool, Tool, CodeAgent, DuckDuckGoSearchTool, HfApiModel, 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 duckduckgo_search import DDGS
import wikipedia
import arxiv
from transformers import pipeline
import os
import re
import ast
import subprocess
import sys

# ===== Search Tools =====
class DuckDuckGoSearchTool:
    def __init__(self, max_results=3):
        self.description = "Search web using DuckDuckGo. Input: search query"
        self.max_results = max_results
        
    def run(self, query: str) -> str:
        try:
            with DDGS() as ddgs:
                results = [r for r in ddgs.text(query, max_results=self.max_results)]
                return "\n\n".join(
                    f"Title: {res['title']}\nURL: {res['href']}\nSnippet: {res['body']}" 
                    for res in results
                )
        except Exception as e:
            return f"Search error: {str(e)}"

class WikiSearchTool:
    def __init__(self, sentences=3):
        self.description = "Get Wikipedia summaries. Input: search phrase"
        self.sentences = sentences
        
    def run(self, query: str) -> str:
        try:
            return wikipedia.summary(query, sentences=self.sentences)
        except wikipedia.DisambiguationError as e:
            return f"Disambiguation error. Options: {', '.join(e.options[:5])}"
        except wikipedia.PageError:
            return "Page not found"
        except Exception as e:
            return f"Wikipedia error: {str(e)}"

class ArxivSearchTool:
    def __init__(self, max_results=3):
        self.description = "Search academic papers on arXiv. Input: search query"
        self.max_results = max_results
        
    def run(self, query: str) -> str:
        try:
            results = arxiv.Search(
                query=query, 
                max_results=self.max_results,
                sort_by=arxiv.SortCriterion.Relevance
            ).results()
            
            output = []
            for r in results:
                output.append(
                    f"Title: {r.title}\n"
                    f"Authors: {', '.join(a.name for a in r.authors)}\n"
                    f"Published: {r.published.strftime('%Y-%m-%d')}\n"
                    f"Summary: {r.summary[:250]}...\n"
                    f"URL: {r.entry_id}"
                )
            return "\n\n".join(output)
        except Exception as e:
            return f"arXiv error: {str(e)}"

# ===== QA Tools =====
class HuggingFaceDocumentQATool:
    def __init__(self):
        self.description = "Answer questions from documents. Input: 'document_text||question'"
        self.model = pipeline(
            'question-answering', 
            model='deepset/roberta-base-squad2',
            tokenizer='deepset/roberta-base-squad2'
        )
    
    def run(self, input_str: str) -> str:
        try:
            if '||' not in input_str:
                return "Invalid format. Use: 'document_text||question'"
                
            context, question = input_str.split('||', 1)
            result = self.model(question=question, context=context)
            return result['answer']
        except Exception as e:
            return f"QA error: {str(e)}"

# ===== Code Execution =====
class PythonCodeExecutionTool:
    def __init__(self):
        self.description = "Execute Python code. Input: valid Python code"
        
    def run(self, code: str) -> str:
        try:
            # Isolate code in a clean environment
            env = {}
            exec(f"def __temp_func__():\n    {indent_code(code)}", env)
            output = env['__temp_func__']()
            return str(output)
        except Exception as e:
            return f"Execution error: {str(e)}"

def indent_code(code: str) -> str:
    """Add proper indentation for multiline code"""
    return '\n    '.join(code.splitlines())

# ===== Answer Formatting =====
class FinalAnswerTool:
    def __init__(self):
        self.description = "Format final answer. Input: answer content"
        
    def run(self, answer: str) -> str:
        return f"FINAL ANSWER: {answer}"

#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}"

@tool
def parse_excel_to_json(task_id: str) -> dict:
    """
    For a given task_id fetch and parse an Excel file and save parsed data in structured JSON file.
    Args:
        task_id: An task ID to fetch.
        
    Returns:
        {
            "task_id": str,
            "sheets": {
                "SheetName1": [ {col1: val1, col2: val2, ...}, ... ],
                ...
            },
            "status": "Success" | "Error"
        }
    """
    url = f"https://agents-course-unit4-scoring.hf.space/files/{task_id}"

    try:
        response = requests.get(url, timeout=100)
        if response.status_code != 200:
            return {"task_id": task_id, "sheets": {}, "status": f"{response.status_code} - Failed"}

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

        for sheet in xls_content.sheet_names:
            df = xls_content.parse(sheet)
            df = df.dropna(how="all")  
            rows = df.head(20).to_dict(orient="records")
            json_sheets[sheet] = rows

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

    except Exception as e:
        return {
            "task_id": task_id,
            "sheets": {},
            "status": f"Error in parsing Excel file: {str(e)}"
        }



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):
        token = os.environ.get("HF_API_TOKEN")
        model = HfApiModel(
            temperature=0.1,
            token=token
        )

        # Existing tools
        search_tool = DuckDuckGoSearchTool()
        wiki_search_tool = WikiSearchTool()
        str_reverse_tool = StringReverseTool()
        keywords_extract_tool = KeywordsExtractorTool()
        speech_to_text_tool = SpeechToTextTool()
        visit_webpage_tool = VisitWebpageTool()
        final_answer_tool = FinalAnswerTool()
        video_transcription_tool = VideoTranscriptionTool()

        # ✅ New Llama Tool
        code_llama_tool = CodeLlamaTool()
        arxiv_search_tool = ArxivSearchTool()
        doc_qa_tool = HuggingFaceDocumentQATool()
        image_qa_tool = HuggingFaceImageQATool()
        translation_tool = HuggingFaceTranslationTool()
        python_tool = PythonCodeExecutionTool()

        system_prompt = f"""
You are my general AI assistant. Your primary goal is to answer the user's question accurately and concisely.

Here's a detailed plan for answering:
1. **Understand the Question:** Carefully parse the question to identify key entities, relationships, and the type of information requested.
2. **Reasoning Steps (Chain-of-Thought):** Before attempting to answer, outline a step-by-step reasoning process. This helps in breaking down complex questions.
3. **Tool Selection and Usage:** Based on your reasoning, select the most appropriate tool(s) to gather information or perform operations.
   - Use `search_tool` (DuckDuckGoSearchTool) for general web searches.
   - Use `wiki_search_tool` for encyclopedic knowledge.
   - Use `arxiv_search_tool` for scientific papers.
   - Use `visit_webpage_tool` to read content from URLs found via search.
   - Use `doc_qa_tool` for answering questions about specific documents (if provided).
   - Use `image_qa_tool` for questions about images.
   - Use `translation_tool` for language translation.
   - Use `python_tool` or `code_llama_tool` for code generation, execution, or complex calculations/data manipulation.
   - Use `keywords_extract_tool` to identify important terms from text.
   - Use `str_reverse_tool` for string manipulation if needed (less common for Q&A).
   - Use `speech_to_text_tool` or `video_transcription_tool` if audio/video input is part of the question.
   - Use `parse_excel_to_json` if the question involves data from Excel.
4. **Information Synthesis:** Combine and process the information obtained from tools. Cross-reference if necessary to ensure accuracy.
5. **Formulate Final Answer:** Construct the final answer according to the specified format.

**Final Answer Format:**
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 explicitly specified in the question.
- If the answer is a string, do not use articles (a, an, the) or common abbreviations (e.g., "NY" for "New York") unless specified. 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.
- If you cannot find a definitive answer, state "FINAL ANSWER: I don't know."

Let's think step by step.
"""
self.agent.prompt_templates["system_prompt"] = self.agent.prompt_templates["system_prompt"] + system_prompt


        self.agent = CodeAgent(
            model=model,
            tools=[
                search_tool, wiki_search_tool, str_reverse_tool,
                keywords_extract_tool, speech_to_text_tool,
                visit_webpage_tool, final_answer_tool,
                parse_excel_to_json, video_transcription_tool,
                arxiv_search_tool,
                doc_qa_tool, image_qa_tool,
                translation_tool, python_tool, 
                code_llama_tool  # 🔧 Add here
            ],
            add_base_tools=True
        )
        self.agent.prompt_templates["system_prompt"] = self.agent.prompt_templates["system_prompt"] + system_prompt

    def __call__(self, question: str) -> str:
        print(f"Agent received question (first 50 chars): {question[:50]}...")
        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)