import os import gradio as gr import requests import pandas as pd from transformers import pipeline import re # --- Constants --- DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" # ----------------------------------------- # BASIC AGENT # ----------------------------------------- class BasicAgent: def __init__(self): print("Loading lightweight GAIA agent model...") # Lightweight model for HF CPU Spaces (stable) self.generator = pipeline( "text-generation", model="TinyLlama/TinyLlama-1.1B-Chat-v1.0", max_new_tokens=32, do_sample=False, temperature=0.0, ) print("Model loaded successfully.") # ------------------------- # TOOL 1: Reverse text # ------------------------- def try_reverse(self, question: str): q = question.strip() # Only reverse if clearly reversed (starts with dot) if q.startswith("."): return q[::-1] return None # ------------------------- # TOOL 2: Safe arithmetic # ------------------------- def try_math(self, question: str): try: pattern = r"\d+\.?\d*\s*[\+\-\*\/]\s*\d+\.?\d*" match = re.search(pattern, question) if match: expression = match.group() result = eval(expression) if float(result).is_integer(): return str(int(result)) return str(result) except: pass return None # ------------------------- # STRICT CLEANING (Exact Match) # ------------------------- def clean_answer(self, text: str) -> str: text = text.strip() if "Answer:" in text: text = text.split("Answer:")[-1] text = text.split("\n")[0].strip() # Remove quotes and trailing punctuation text = text.strip('"').strip("'") text = re.sub(r"[\.]$", "", text) return text.strip() # ------------------------- # MODEL CALL # ------------------------- def ask_model(self, question: str): prompt = f"""You are answering a benchmark question. Return ONLY the exact final answer. No explanation. No extra words. If number → return number only. If word → return word only. Question: {question} Answer:""" output = self.generator(prompt)[0]["generated_text"] answer = output.replace(prompt, "") return self.clean_answer(answer) # ------------------------- # MAIN LOGIC # ------------------------- def __call__(self, question: str) -> str: print(f"Processing: {question[:60]}...") # 1️⃣ Reverse tool reversed_q = self.try_reverse(question) if reversed_q: print("Used reverse tool.") return self.ask_model(reversed_q) # 2️⃣ Math tool math_result = self.try_math(question) if math_result: print("Used math tool.") return math_result # 3️⃣ LLM reasoning answer = self.ask_model(question) # Retry once if output too long if len(answer.split()) > 5: print("Retrying for shorter answer...") answer = self.ask_model(question) print(f"Final Answer: {answer}") return answer # ----------------------------------------- # RUN + SUBMIT FUNCTION # ----------------------------------------- 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: 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" # Instantiate agent try: agent = BasicAgent() except Exception as e: return f"Error initializing agent: {e}", None agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" # Fetch Questions try: response = requests.get(questions_url, timeout=15) response.raise_for_status() questions_data = response.json() 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") question_text = item.get("question") if not task_id or question_text is None: 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: results_log.append( { "Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}", } ) if not answers_payload: return "Agent did not produce any answers.", pd.DataFrame(results_log) submission_data = { "username": username.strip(), "agent_code": agent_code, "answers": answers_payload, } # Submit 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', '?')}/" f"{result_data.get('total_attempted', '?')} correct)\n" f"Message: {result_data.get('message', '')}" ) return final_status, pd.DataFrame(results_log) except Exception as e: return f"Submission Failed: {e}", pd.DataFrame(results_log) with gr.Blocks() as demo: gr.Markdown("# Basic Agent Evaluation Runner") gr.LoginButton() run_button = gr.Button("Run Evaluation & Submit All Answers") status_output = gr.Textbox( label="Run Status / Submission Result", lines=5, interactive=False ) 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__": demo.launch(debug=True, share=False)