app.py

import os
import gradio as gr
import requests
import inspect
import pandas as pd
import asyncio
import aiohttp
import time
import random
import json
import re
from smolagents import FinalAnswerTool, Tool, tool, OpenAIServerModel, DuckDuckGoSearchTool, CodeAgent, VisitWebpageTool
from gemini_agent import GeminiAgent  # Assuming you have a GeminiAgent class defined in gemini_agent.py

from dotenv import load_dotenv

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


OPENAI_TOKEN = os.getenv("OPENAI_API_KEY")

# --- Custom Tools for Better Reasoning ---

class TrickQuestionDetector(Tool):
    """Detects and handles trick questions"""
    
    def __init__(self):
        super().__init__()
        self.name = "trick_detector"
        self.description = "Analyze if a question is a trick question and provide guidance"
        self.inputs = {"question": {"type": "string", "description": "The question to analyze"}}
    
    def detect_trick(self, question: str) -> str:
        """Detect common trick question patterns"""
        q_lower = question.lower()
        
        # Reverse text tricks - check if question might be reversed
        reversed_q = question[::-1]
        if len(question) > 5 and any(c.isalpha() for c in question):
            # Simple heuristic: if reversed version has common English patterns
            if any(word in reversed_q.lower() for word in ['the', 'and', 'what', 'how', 'when', 'where']):
                return f"TRICK DETECTED: This appears to be reversed text. Decoded: '{reversed_q}'"
        
        # Word puzzles
        if 'rewsna' in question or 'tfel' in question:
            return "TRICK DETECTED: Contains reversed words. Try reading backwards."
        
        # Contradictory statements
        contradiction_words = ['impossible', 'never', 'always', 'none', 'all']
        if sum(word in q_lower for word in contradiction_words) >= 2:
            return "TRICK DETECTED: Contains contradictory terms. Look for logical impossibilities."
        
        # Mathematical tricks
        if any(phrase in q_lower for phrase in ['how many', 'total', 'sum']) and 'zero' in q_lower:
            return "TRICK DETECTED: Mathematical trick involving zero or impossible calculations."
        
        return "No obvious trick detected. Proceed with normal analysis."

class StepByStepReasoner(Tool):
    """Breaks down complex questions into steps"""
    
    def __init__(self):
        super().__init__()
        self.name = "step_reasoner"
        self.description = "Break down complex questions into logical steps"
        self.inputs = {"question": {"type": "string", "description": "The question to break down"}}
    
    def reason_steps(self, question: str) -> str:
        """Break question into reasoning steps"""
        steps = []
        q_lower = question.lower()
        
        # Identify question components
        if any(word in q_lower for word in ['who', 'what', 'when', 'where', 'why', 'how']):
            steps.append("1. Identify the specific information being requested")
        
        if any(word in q_lower for word in ['between', 'from', 'to', 'during']):
            steps.append("2. Note the time period or range specified")
        
        if any(word in q_lower for word in ['calculate', 'count', 'how many', 'total']):
            steps.append("3. Determine what needs to be calculated or counted")
        
        if any(word in q_lower for word in ['wikipedia', 'article', 'featured']):
            steps.append("4. Consider Wikipedia-specific processes and history")
        
        if any(word in q_lower for word in ['only', 'single', 'one', 'unique']):
            steps.append("5. Focus on finding the single/unique answer requested")
        
        steps.append("6. Verify the answer makes logical sense")
        
        return "REASONING STEPS:\n" + "\n".join(steps)

class FactChecker(Tool):
    """Validates factual claims and provides confidence levels"""
    
    def __init__(self):
        super().__init__()
        self.name = "fact_checker"
        self.description = "Check factual accuracy and provide confidence assessment"
        self.inputs = {"claim": {"type": "string", "description": "The claim to fact-check"}}
    
    def check_facts(self, claim: str) -> str:
        """Assess factual accuracy of a claim"""
        confidence_indicators = {
            'high': ['wikipedia', 'well-known', 'documented', 'official', 'verified'],
            'medium': ['likely', 'probably', 'appears', 'seems', 'reported'],
            'low': ['unclear', 'uncertain', 'possibly', 'might', 'could be']
        }
        
        claim_lower = claim.lower()
        
        # Check for confidence indicators
        high_conf = sum(1 for word in confidence_indicators['high'] if word in claim_lower)
        medium_conf = sum(1 for word in confidence_indicators['medium'] if word in claim_lower)
        low_conf = sum(1 for word in confidence_indicators['low'] if word in claim_lower)
        
        if high_conf > medium_conf and high_conf > low_conf:
            return f"CONFIDENCE: HIGH - Claim appears to be well-documented: '{claim}'"
        elif low_conf > high_conf:
            return f"CONFIDENCE: LOW - Claim contains uncertainty markers: '{claim}'"
        else:
            return f"CONFIDENCE: MEDIUM - Standard factual claim: '{claim}'"

class AnswerValidator(Tool):
    """Validates if an answer makes sense for the question"""
    
    def __init__(self):
        super().__init__()
        self.name = "answer_validator"
        self.description = "Validate if an answer is reasonable for the given question"
        self.inputs = {"question": {"type": "string", "description": "The question"}, "answer": {"type": "string", "description": "The answer to validate"}}
    
    def validate_answer(self, question: str, answer: str) -> str:
        """Check if answer is reasonable for the question"""
        q_lower = question.lower()
        a_lower = answer.lower()
        
        # Check for question-answer type matching
        if 'who' in q_lower and not any(indicator in a_lower for indicator in ['person', 'user', 'editor', 'author', 'name']):
            return "WARNING: 'Who' question but answer doesn't seem to identify a person"
        
        if 'when' in q_lower and not any(indicator in a_lower for indicator in ['year', 'date', 'time', '20', '19']):
            return "WARNING: 'When' question but answer doesn't contain time information"
        
        if 'how many' in q_lower and not any(char.isdigit() for char in answer):
            return "WARNING: 'How many' question but answer contains no numbers"
        
        if len(answer.strip()) < 3:
            return "WARNING: Answer seems too short"
        
        if len(answer.strip()) > 200:
            return "WARNING: Answer seems too long - may need to be more concise"
        
        return "VALIDATION: Answer format appears appropriate for question type"

# --- Enhanced Agent with Tools ---
class SlpMultiAgent:
    def __init__(self):
        print("Enhanced Agent initialized with reasoning tools.")
        self.trick_detector = TrickQuestionDetector()
        self.step_reasoner = StepByStepReasoner()
        self.fact_checker = FactChecker()
        self.answer_validator = AnswerValidator()
        
    async def __call__(self, question: str) -> str:
        print(f"Agent received question (first 50 chars): {question[:50]}...")
        
        # Step 1: Check for tricks
        trick_analysis = self.trick_detector.detect_trick(question)
        print(f"Trick analysis: {trick_analysis}")
        
        # Step 2: Break down reasoning steps
        reasoning_steps = self.step_reasoner.reason_steps(question)
        print(f"Reasoning steps: {reasoning_steps}")
        
        # Step 3: Enhanced model call with tool insights
        model = OpenAIServerModel(
            model_id="gpt-4o-mini",
            temperature=0.1,
            max_tokens=1000
        )
        
        try:
            enhanced_prompt = f"""You are an expert problem solver. Analyze this question carefully:

QUESTION: {question}

TRICK ANALYSIS: {trick_analysis}

{reasoning_steps}

Instructions:
1. If a trick was detected, handle it appropriately
2. Follow the reasoning steps systematically
3. Think through each step carefully
4. Provide a clear, direct answer
5. If unsure, state your uncertainty clearly

Be precise and thorough in your analysis."""

            messages = [
                {
                    "role": "system",
                    "content": "You are an expert at solving complex and trick questions. Always think step by step and be very careful about the exact wording of questions."
                },
                {
                    "role": "user", 
                    "content": enhanced_prompt
                }
            ]
            
            result = model(messages)
            
            if result:
                # Step 4: Validate the answer
                validation = self.answer_validator.validate_answer(question, result)
                print(f"Answer validation: {validation}")
                
                # Clean up the result
                lines = result.strip().split('\n')
                for line in reversed(lines):
                    line = line.strip()
                    if line and len(line) > 5 and not line.startswith(('Step', 'Analysis', 'TRICK', 'REASONING')):
                        # Remove common prefixes
                        line = re.sub(r'^(Answer:|Final answer:|The answer is:?)\s*', '', line, flags=re.IGNORECASE)
                        if line:
                            return line
                
                return result
            else:
                return "I don't have enough information to answer this question accurately."
                
        except Exception as e:
            print(f"Model call failed: {e}")
            return "I apologize, but I'm currently experiencing technical difficulties."

def check_reasoning(final_answer, agent_memory):
    return True


async def run_and_submit_all(profile):
    """
    Fetches all questions, runs the BasicAgent on them, submits all answers,
    and displays the results asynchronously.
    """
    # --- Determine HF Space Runtime URL and Repo URL ---
    space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code

    # Handle different profile types
    if profile:
        if hasattr(profile, 'username'):
            # It's an OAuthProfile object
            username = profile.username
        else:
            # It's a string or other type
            username = str(profile)
        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 = GeminiAgent()
    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:
        async with aiohttp.ClientSession() as session:
            async with session.get(questions_url, timeout=15) as response:
                response.raise_for_status()
                questions_data = await 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 aiohttp.ClientError as e:
        print(f"Error fetching questions: {e}")
        return f"Error fetching questions: {e}", None
    except ValueError as e:  # JSON decode error
        print(f"Error decoding JSON response from questions endpoint: {e}")
        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...")
    
    # Process questions with controlled concurrency
    semaphore = asyncio.Semaphore(2)  # Process 2 questions at a time
    
    async def process_question(item):
        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}")
            return None
        
        async with semaphore:
            try:
                print(f"Processing task {task_id}")
                submitted_answer = await agent(question_text)
                return {"task_id": task_id, "submitted_answer": submitted_answer, 
                        "log": {"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}")
                default_answer = "I don't have enough information to answer this question accurately."
                return {"task_id": task_id, "submitted_answer": default_answer, 
                        "log": {"Task ID": task_id, "Question": question_text, "Submitted Answer": default_answer}}
    
    # Create tasks for all questions
    tasks = [process_question(item) for item in questions_data]
    results = await asyncio.gather(*tasks)
    
    # Process results
    for result in results:
        if result is not None:
            answers_payload.append({"task_id": result["task_id"], "submitted_answer": result["submitted_answer"]})
            results_log.append(result["log"])

    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": str(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:
        async with aiohttp.ClientSession() as session:
            async with session.post(submit_url, json=submission_data, timeout=60) as response:
                response.raise_for_status()
                result_data = await 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 aiohttp.ClientResponseError as e:
        error_detail = f"Server responded with status {e.status}."
        try:
            error_text = await e.response.text()
            try:
                error_json = await e.response.json()
                error_detail += f" Detail: {error_json.get('detail', error_text)}"
            except ValueError:
                error_detail += f" Response: {error_text[:500]}"
        except:
            pass
        status_message = f"Submission Failed: {error_detail}"
        print(status_message)
        results_df = pd.DataFrame(results_log)
        return status_message, results_df
    except asyncio.TimeoutError:
        status_message = "Submission Failed: The request timed out."
        print(status_message)
        results_df = pd.DataFrame(results_log)
        return status_message, results_df
    except aiohttp.ClientError 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.
        """
    )

    login_button = 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)
    
    def sync_wrapper(profile):
        # This wrapper ensures we have access to the profile
        if not profile:
            print("No profile available in sync_wrapper")
            return "Please Login to Hugging Face with the button.", None
        print(f"Profile type in wrapper: {type(profile)}")
        try:
            return asyncio.run(run_and_submit_all(profile))
        except Exception as e:
            print(f"Error in sync_wrapper: {e}")
            return f"Error processing request: {e}", None
    
    run_button.click(
        fn=sync_wrapper,
        inputs=login_button,
        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)