app.py

import os
import gradio as gr
import requests
import pandas as pd
import torch
from transformers import BartForConditionalGeneration, BartTokenizer

from audio_transcriber import AudioTranscriptionTool
from image_analyzer import ImageAnalysisTool
from wikipedia_searcher import WikipediaSearcher
from smolagents import ToolCallingAgent

DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"

SYSTEM_PROMPT = (
    "You are an agent solving the GAIA benchmark and must provide exact answers.\n"
    "Rules:\n"
    "1. Return only the exact requested answer: no explanation.\n"
    "2. For yes/no, return 'Yes' or 'No'.\n"
    "3. For dates, use the exact requested format.\n"
    "4. For numbers, use only the number.\n"
    "5. For names, use the exact name from sources.\n"
    "6. If the question has a file, download it using the task ID.\n"
    "Examples:\n"
    "- '42'\n"
    "- 'Arturo Nunez'\n"
    "- 'Yes'\n"
    "- 'October 5, 2001'\n"
    "- 'Buenos Aires'\n"
    "Never say 'the answer is...'. Only return the answer.\n"
)

# Local wrapper for facebook/bart-base that exposes generate()
class LocalBartModel:
    def __init__(self, model_name="facebook/bart-base"):
        self.tokenizer = BartTokenizer.from_pretrained(model_name)
        self.model = BartForConditionalGeneration.from_pretrained(model_name)
        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        self.model.to(self.device)

    def generate(self, input_ids, **generate_kwargs):
        return self.model.generate(input_ids.to(self.device), **generate_kwargs)

    def __call__(self, prompt: str) -> str:
        inputs = self.tokenizer(prompt, return_tensors="pt").to(self.device)
        output_ids = self.generate(
            inputs.input_ids,
            max_length=100,
            num_beams=5,
            early_stopping=True
        )
        output_text = self.tokenizer.decode(output_ids[0], skip_special_tokens=True)
        return output_text.strip()

class GaiaAgent:
    def __init__(self):
        print("Gaia Agent Initialized")
        self.model = LocalBartModel()
        self.tools = [
            AudioTranscriptionTool(),
            ImageAnalysisTool(),
            WikipediaSearcher()
        ]
        self.agent = ToolCallingAgent(
            tools=self.tools,
            model=self.model
        )

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

        full_prompt = f"{SYSTEM_PROMPT}\nQUESTION:\n{question}"

        try:
            result = self.agent.run(full_prompt)
            print(f"Raw result from agent: {result}")

            if isinstance(result, dict) and "answer" in result:
                return str(result["answer"]).strip()
            elif isinstance(result, str):
                return result.strip()
            elif isinstance(result, list):
                for item in reversed(result):
                    if isinstance(item, dict) and item.get("role") == "assistant" and "content" in item:
                        return item["content"].strip()
                return "ERROR: Unexpected list format"
            else:
                return "ERROR: Unexpected result type"
        except Exception as e:
            print(f"Exception during agent run: {e}")
            return f"AGENT ERROR: {e}"

def run_and_submit_all(profile: gr.OAuthProfile | None):
    space_id = os.getenv("SPACE_ID")

    if profile:
        username = 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"

    try:
        agent = GaiaAgent()
    except Exception as e:
        print(f"Error initializing agent: {e}")
        return f"Error initializing agent: {e}", None

    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
    print(f"Agent code URL: {agent_code}")

    try:
        response = requests.get(questions_url, timeout=15)
        response.raise_for_status()
        questions_data = response.json()
        if not questions_data:
            return "Fetched questions list is empty or invalid format.", None
        print(f"Fetched {len(questions_data)} questions.")
    except Exception as e:
        return f"Error fetching questions: {e}", None

    results_log = []
    answers_payload = []

    for item in questions_data:
        task_id = item.get("task_id")
        if not task_id:
            continue
        try:
            submitted_answer = agent(item.get("question", ""))
            answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
            results_log.append({
                "Task ID": task_id,
                "Question": item.get("question", ""),
                "Submitted Answer": submitted_answer
            })
        except Exception as e:
            error_msg = f"AGENT ERROR: {e}"
            results_log.append({
                "Task ID": task_id,
                "Question": item.get("question", ""),
                "Submi