Spaces:
Sleeping
Sleeping
| import os | |
| import time | |
| import gradio as gr | |
| import requests | |
| import pandas as pd | |
| from smolagents import CodeAgent, OpenAIServerModel, PythonInterpreterTool, Tool | |
| from smolagents import FinalAnswerTool | |
| # --- Constants --- | |
| DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" | |
| # --- Custom Tools --- | |
| class WebSearchTool(Tool): | |
| name = "web_search" | |
| description = "Search the web for information. Use for any factual question." | |
| inputs = {"query": {"type": "string", "description": "The search query"}} | |
| output_type = "string" | |
| def forward(self, query: str) -> str: | |
| try: | |
| from ddgs import DDGS | |
| with DDGS() as ddgs: | |
| results = list(ddgs.text(query, max_results=5)) | |
| if not results: | |
| return "No results found." | |
| output = "" | |
| for r in results: | |
| output += f"Title: {r.get('title', '')}\n" | |
| output += f"URL: {r.get('href', '')}\n" | |
| output += f"Summary: {r.get('body', '')}\n\n" | |
| return output[:3000] | |
| except Exception as e: | |
| return f"Search error: {e}" | |
| class WikipediaTool(Tool): | |
| name = "wikipedia_search" | |
| description = "Search Wikipedia directly. Use when the question mentions Wikipedia or needs encyclopedic facts like discographies, biographies, lists." | |
| inputs = {"query": {"type": "string", "description": "The Wikipedia article title or topic to search"}} | |
| output_type = "string" | |
| def forward(self, query: str) -> str: | |
| try: | |
| # First search for the right article | |
| search_url = ( | |
| "https://en.wikipedia.org/w/api.php" | |
| f"?action=query&list=search&srsearch={requests.utils.quote(query)}" | |
| "&format=json&srlimit=1" | |
| ) | |
| r = requests.get(search_url, timeout=10) | |
| results = r.json()["query"]["search"] | |
| if not results: | |
| return "No Wikipedia article found." | |
| title = results[0]["title"] | |
| # Then fetch full article text | |
| content_url = ( | |
| "https://en.wikipedia.org/w/api.php" | |
| f"?action=query&titles={requests.utils.quote(title)}" | |
| "&prop=extracts&explaintext=true&format=json" | |
| ) | |
| r2 = requests.get(content_url, timeout=10) | |
| pages = r2.json()["query"]["pages"] | |
| page = next(iter(pages.values())) | |
| text = page.get("extract", "No content found") | |
| return f"Article: {title}\n\n{text[:5000]}" | |
| except Exception as e: | |
| return f"Wikipedia error: {e}" | |
| class YouTubeTranscriptTool(Tool): | |
| name = "youtube_transcript" | |
| description = "Gets the transcript/captions of a YouTube video. Use when the question contains a YouTube URL." | |
| inputs = {"url": {"type": "string", "description": "YouTube video URL or video ID"}} | |
| output_type = "string" | |
| def forward(self, url: str) -> str: | |
| try: | |
| from youtube_transcript_api import YouTubeTranscriptApi | |
| if "v=" in url: | |
| video_id = url.split("v=")[1].split("&")[0] | |
| elif "youtu.be/" in url: | |
| video_id = url.split("youtu.be/")[1].split("?")[0] | |
| else: | |
| video_id = url.strip() | |
| ytt = YouTubeTranscriptApi() | |
| transcript = ytt.fetch(video_id) | |
| return " ".join([t.text for t in transcript])[:5000] | |
| except Exception as e: | |
| return f"Transcript error: {e}" | |
| class FileDownloadTool(Tool): | |
| name = "download_file" | |
| description = "Downloads a file attached to a GAIA question using its task_id. Use when the question references an attached file, image, CSV, or PDF." | |
| inputs = {"task_id": {"type": "string", "description": "The task_id of the current question"}} | |
| output_type = "string" | |
| def forward(self, task_id: str) -> str: | |
| try: | |
| url = f"https://agents-course-unit4-scoring.hf.space/files/{task_id}" | |
| r = requests.get(url, timeout=15) | |
| if r.status_code == 200: | |
| return r.text[:5000] | |
| return f"No file found for task_id {task_id}" | |
| except Exception as e: | |
| return f"File download error: {e}" | |
| class VisitWebpageTool(Tool): | |
| name = "visit_webpage" | |
| description = "Fetches the full content of a webpage given its URL. Use when you have a specific URL to read." | |
| inputs = {"url": {"type": "string", "description": "The URL of the webpage to visit"}} | |
| output_type = "string" | |
| def forward(self, url: str) -> str: | |
| try: | |
| headers = {"User-Agent": "Mozilla/5.0"} | |
| r = requests.get(url, timeout=10, headers=headers) | |
| # strip html tags roughly | |
| import re | |
| text = re.sub(r'<[^>]+>', ' ', r.text) | |
| text = re.sub(r'\s+', ' ', text).strip() | |
| return text[:5000] | |
| except Exception as e: | |
| return f"Webpage error: {e}" | |
| # --- Agent --- | |
| class BasicAgent: | |
| def __init__(self): | |
| model = OpenAIServerModel( | |
| model_id="meta-llama/llama-4-scout-17b-16e-instruct", | |
| api_base="https://api.groq.com/openai/v1", | |
| api_key=os.getenv("GROQ_API_KEY") | |
| ) | |
| self.agent = CodeAgent( # <-- back to CodeAgent | |
| model=model, | |
| tools=[ | |
| WebSearchTool(), | |
| WikipediaTool(), | |
| YouTubeTranscriptTool(), | |
| FileDownloadTool(), | |
| VisitWebpageTool(), | |
| PythonInterpreterTool(), | |
| ], | |
| max_steps=6, | |
| ) | |
| def __call__(self, question: str, task_id: str = "") -> str: | |
| try: | |
| prompt = f"""Answer the following question accurately. | |
| Return ONLY the final answer with no explanation, no punctuation, no extra words. | |
| - If the answer is a number, return just the number. | |
| - If the answer is a name, return just the name. | |
| - If the answer is a list, return comma separated values in alphabetical order. | |
| - If the question asks about a YouTube video, use the youtube_transcript tool. | |
| - If the question mentions Wikipedia, use the wikipedia_search tool. | |
| - If the question references an attached file, use download_file with the task_id below. | |
| Task ID: {task_id} | |
| Question: {question}""" | |
| result = self.agent.run(prompt) | |
| if isinstance(result, list): | |
| for block in result: | |
| if isinstance(block, dict) and block.get('type') == 'text': | |
| return block['text'].strip() | |
| return str(result).strip() | |
| except Exception as e: | |
| print(f"Agent error: {e}") | |
| return "I don't know" | |
| # --- Main Evaluation Function --- | |
| def run_and_submit_all(profile: gr.OAuthProfile | None): | |
| space_id = os.getenv("SPACE_ID") | |
| if profile: | |
| username = f"{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" | |
| 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" | |
| print(agent_code) | |
| 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: | |
| 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 = [] | |
| print(f"Running agent on {len(questions_data)} questions...") | |
| for i, item in enumerate(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 | |
| print(f"\n[{i+1}/{len(questions_data)}] Task: {task_id}") | |
| print(f"Question: {question_text[:120]}...") | |
| try: | |
| submitted_answer = agent(question_text, task_id) | |
| print(f"Answer: {submitted_answer}") | |
| 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 on task {task_id}: {e}") | |
| results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"}) | |
| if i < len(questions_data) - 1: | |
| print("Waiting 15s for rate limits...") | |
| time.sleep(15) | |
| if not answers_payload: | |
| return "Agent did not produce any answers to submit.", pd.DataFrame(results_log) | |
| submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload} | |
| print(f"\nSubmitting {len(answers_payload)} answers...") | |
| 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.") | |
| return final_status, pd.DataFrame(results_log) | |
| 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 Exception: | |
| error_detail += f" Response: {e.response.text[:500]}" | |
| return f"Submission Failed: {error_detail}", pd.DataFrame(results_log) | |
| except Exception as e: | |
| return f"Submission error: {e}", pd.DataFrame(results_log) | |
| # --- Gradio UI --- | |
| with gr.Blocks() as demo: | |
| gr.Markdown("# GAIA Agent Evaluation Runner") | |
| gr.Markdown( | |
| """ | |
| **Instructions:** | |
| 1. Log in to your Hugging Face account using the button below. | |
| 2. Click 'Run Evaluation & Submit All Answers' to start. | |
| 3. Takes ~6 minutes for all 20 questions due to rate limits. | |
| """ | |
| ) | |
| 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__": | |
| print("\n" + "-"*30 + " App Starting " + "-"*30) | |
| space_host_startup = os.getenv("SPACE_HOST") | |
| space_id_startup = os.getenv("SPACE_ID") | |
| if space_host_startup: | |
| print(f"✅ SPACE_HOST found: {space_host_startup}") | |
| else: | |
| print("ℹ️ SPACE_HOST not found (running locally).") | |
| if space_id_startup: | |
| print(f"✅ SPACE_ID found: {space_id_startup}") | |
| else: | |
| print("ℹ️ SPACE_ID not found (running locally).") | |
| print("-"*(60 + len(" App Starting ")) + "\n") | |
| print("Launching Gradio Interface...") | |
| demo.launch(debug=True, share=False) |