| import os |
| import re |
| import requests |
| import gradio as gr |
| from PIL import Image |
| from pypdf import PdfReader |
| from smolagents import CodeAgent, DuckDuckGoSearchTool, InferenceClientModel, tool |
|
|
| |
| HF_TOKEN = os.getenv("HF_TOKEN") |
| model = InferenceClientModel( |
| model_id="Qwen/Qwen2.5-Coder-1.5B-Instruct", |
| token=HF_TOKEN |
| ) |
|
|
| |
| @tool |
| def read_pdf_content(file_path: str) -> str: |
| """Reads and extracts text from a PDF file. |
| |
| Args: |
| file_path: Path to the PDF file. |
| """ |
| try: |
| reader = PdfReader(file_path) |
| text = "" |
| for p in reader.pages: |
| text += p.extract_text() or "" |
| return text[:3000] if text else "No text found." |
| except Exception as e: |
| return f"Error reading PDF: {e}" |
|
|
| @tool |
| def inspect_image(file_path: str) -> str: |
| """Inspects image dimensions and metadata. |
| |
| Args: |
| file_path: Path to the image. |
| """ |
| try: |
| with Image.open(file_path) as img: |
| return f"Format: {img.format}, Size: {img.size}, Mode: {img.mode}" |
| except Exception as e: |
| return f"Error reading image: {e}" |
|
|
| |
| gaia_agent = CodeAgent( |
| tools=[DuckDuckGoSearchTool(), read_pdf_content, inspect_image], |
| model=model, |
| max_steps=5, |
| additional_authorized_imports=["math", "statistics", "re", "json", "datetime"] |
| ) |
|
|
| DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" |
|
|
| def solve_single_task(question: str, file_path: str = None) -> str: |
| prompt = f"Question: {question}\n" |
| if file_path: |
| prompt += f"Local file available at: {file_path}\n" |
| prompt += "Provide ONLY the concise final answer without explanations, reasoning, or 'FINAL ANSWER' prefix." |
| try: |
| raw = str(gaia_agent.run(prompt)).strip() |
| return re.sub(r'^(FINAL ANSWER|Answer)[:\s]*', '', raw, flags=re.IGNORECASE).strip() |
| except Exception: |
| return "N/A" |
|
|
| def run_and_submit_all(profile: gr.OAuthProfile | None): |
| if not profile: |
| return "Please Login to Hugging Face with the button first.", None |
|
|
| username = profile.username |
| space_id = os.getenv("SPACE_ID", "Learnix-AI-Lab/Final_Assignment_Template") |
| agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" |
| |
| questions_url = f"{DEFAULT_API_URL}/questions" |
| submit_url = f"{DEFAULT_API_URL}/submit" |
|
|
| try: |
| resp = requests.get(questions_url, timeout=20) |
| resp.raise_for_status() |
| questions_data = resp.json() |
| except Exception as e: |
| return f"Error fetching questions: {e}", None |
|
|
| answers_payload = [] |
| display_rows = [] |
|
|
| for item in questions_data: |
| task_id = item.get("task_id") |
| q_text = item.get("question") |
| file_name = item.get("file_name") |
| local_path = None |
|
|
| if file_name: |
| file_dl_url = f"{DEFAULT_API_URL}/files/{task_id}" |
| dl_resp = requests.get(file_dl_url) |
| if dl_resp.status_code == 200: |
| local_path = f"tmp_{file_name}" |
| with open(local_path, "wb") as f: |
| f.write(dl_resp.content) |
|
|
| ans = solve_single_task(q_text, local_path) |
| answers_payload.append({"task_id": task_id, "submitted_answer": ans}) |
| display_rows.append([q_text, ans]) |
|
|
| if local_path and os.path.exists(local_path): |
| os.remove(local_path) |
|
|
| submit_data = { |
| "username": username, |
| "agent_code": agent_code, |
| "answers": answers_payload |
| } |
| |
| try: |
| sub_resp = requests.post(submit_url, json=submit_data, timeout=30) |
| sub_resp.raise_for_status() |
| result_json = sub_resp.json() |
| score = result_json.get("score", "N/A") |
| return f"Successfully submitted! Your Score: {score}", display_rows |
| except Exception as e: |
| return f"Submission error: {e}", display_rows |
|
|
| |
| with gr.Blocks() as demo: |
| gr.Markdown("# 🤖 GAIA Benchmark Evaluator") |
| gr.LoginButton() |
| run_btn = gr.Button("Run Evaluation & Submit All Answers", variant="primary") |
| status_box = gr.Textbox(label="Status / Result", lines=3) |
| results_table = gr.DataFrame(headers=["Question", "Agent Answer"], label="Answers") |
|
|
| run_btn.click( |
| fn=run_and_submit_all, |
| outputs=[status_box, results_table] |
| ) |
|
|
| if __name__ == "__main__": |
| demo.launch() |
