import os,re import requests import pandas as pd import gradio as gr from PIL import Image import pytesseract from pydub import AudioSegment import yt_dlp from bs4 import BeautifulSoup import whisper import yt_dlp from openai import OpenAI # (Keep Constants as is) # --- Constants --- DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" # ---------- FILE TOOLS ---------- def download_file(url): local_path = "temp_file" r = requests.get(url) with open(local_path, "wb") as f: f.write(r.content) return local_path #def read_audio(file_path): # try: # audio = AudioSegment.from_file(file_path) # return f"Audio length: {len(audio)} ms" # except: # return "Audio read error" def youtube_captions(self, url): try: ydl_opts = {"skip_download": True, "writesubtitles": True, "writeautomaticsub": True} with yt_dlp.YoutubeDL(ydl_opts) as ydl: info = ydl.extract_info(url, download=False) # Return first available captions return str(info.get("subtitles") or info.get("automatic_captions"))[:5000] except Exception as e: return f"YouTube error: {e}" # --- Basic Agent Definition --- # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------ from openai import OpenAI class BasicAgent: def __init__(self): print("πŸš€ Super GAIA Agent initialized") self.client = OpenAI() self.audio_model = whisper.load_model("base") # Network connectivity test for url in ["https://en.wikipedia.org/api/rest_v1/page/summary/Python_(programming_language)", "https://httpbin.org/get"]: try: r = requests.get(url, timeout=5, headers={"User-Agent": "test"}) print(f" [NET TEST] {url[:50]} β†’ {r.status_code}") except Exception as e: print(f" [NET TEST] {url[:50]} β†’ BLOCKED: {e}") def __call__(self, question, file_url=None): return self.agent_loop(question, file_url) # ── TOOL: Wikipedia ────────────────────────────────────────────── def wiki_search(self, query): try: query = query.strip(' ".,') # Step 1: Search for page title search_resp = requests.get( "https://en.wikipedia.org/w/api.php", params={ "action": "query", "list": "search", "srsearch": query, "format": "json", "srlimit": 2 }, headers={"User-Agent": "GAIA-Agent/1.0"}, timeout=15 ) print(f" [wiki status] {search_resp.status_code}, len={len(search_resp.text)}") if search_resp.status_code != 200 or not search_resp.text.strip(): # Fallback: try REST summary directly slug = query.replace(" ", "_") rest = requests.get( f"https://en.wikipedia.org/api/rest_v1/page/summary/{slug}", headers={"User-Agent": "GAIA-Agent/1.0"}, timeout=15 ) if rest.status_code == 200: data = rest.json() return f"WIKI [{data.get('title')}]: {data.get('extract','')[:2000]}" return f"Wiki unavailable (status {search_resp.status_code})" data = search_resp.json() results = data.get("query", {}).get("search", []) if not results: return f"No Wikipedia results for: {query}" title = results[0]["title"] # Step 2: Get full extract rest = requests.get( f"https://en.wikipedia.org/api/rest_v1/page/summary/{requests.utils.quote(title)}", headers={"User-Agent": "GAIA-Agent/1.0"}, timeout=15 ) if rest.status_code == 200: d = rest.json() return f"WIKI [{d.get('title')}]: {d.get('extract','')[:2500]}" return f"Wiki fetch failed for: {title}" except Exception as e: return f"Wiki error: {e}" def web_search(self, query): """General web search using DuckDuckGo""" try: from duckduckgo_search import DDGS with DDGS() as ddgs: results = list(ddgs.text(query, max_results=5)) if not results: return f"No results for: {query}" output = "" for r in results: output += f"\n[{r['title']}] {r['href']}\n{r['body']}\n" return output[:3000] except Exception as e: return f"Search error: {e}" # ── TOOL: Scrape web page ───────────────────────────────────────── def scrape_page(self, url, search_terms=None): url = url.strip(' "') if "youtube.com" in url or "youtu.be" in url: return "YouTube cannot be scraped directly." try: headers = {"User-Agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 Chrome/120.0 Safari/537.36"} resp = requests.get(url, timeout=15, headers=headers) soup = BeautifulSoup(resp.text, "html.parser") for tag in soup(["script", "style", "nav", "footer", "header"]): tag.decompose() full_text = soup.get_text(separator=" ", strip=True) if len(full_text) < 100: return f"Page returned too little content (status {resp.status_code})" # If search terms provided, find the most relevant 4000-char window if search_terms: terms = search_terms.lower().split() best_pos = 0 best_score = 0 # Slide a window and find the chunk with most term matches window = 3000 for pos in range(0, len(full_text) - window, 500): chunk = full_text[pos:pos+window].lower() score = sum(chunk.count(t) for t in terms) if score > best_score: best_score = score best_pos = pos relevant = full_text[max(0, best_pos-200):best_pos+window] return f"PAGE (relevant section): {relevant}" # Default: return first 8000 chars return f"PAGE: {full_text[:8000]}" except Exception as e: return f"Scrape error: {e}" # ── TOOL: Read audio via Whisper ────────────────────────────────── def read_audio(self, url): try: url = url.strip(' "') r = requests.get(url, timeout=30) r.raise_for_status() # Save to a temp file with proper extension import tempfile, os ext = url.split('.')[-1].lower().split('?')[0] with tempfile.NamedTemporaryFile(suffix=f'.{ext}', delete=False) as f: f.write(r.content) tmp_path = f.name print(f" [audio] downloaded {len(r.content)} bytes to {tmp_path}") result = self.audio_model.transcribe(tmp_path) os.unlink(tmp_path) return f"TRANSCRIPT: {result['text']}" except Exception as e: return f"Audio error: {e}" # ── TOOL: Read Excel ────────────────────────────────────────────── def read_excel(self, url): try: url = url.strip(' "{}') # Handle case where model passes JSON like {"file_url": "..."} if url.startswith('{'): import json url = json.loads(url).get('file_url', url) url = url.strip(' "') r = requests.get(url, timeout=20) r.raise_for_status() import tempfile, os with tempfile.NamedTemporaryFile(suffix='.xlsx', delete=False) as f: f.write(r.content) tmp_path = f.name print(f" [excel] downloaded {len(r.content)} bytes") try: df = pd.read_excel(tmp_path, engine='openpyxl') except: df = pd.read_excel(tmp_path, engine='xlrd') os.unlink(tmp_path) return f"EXCEL_DATA:\n{df.to_string()[:5000]}" except Exception as e: return f"Excel error: {e}" # ── TOOL: Read image via OCR ────────────────────────────────────── def read_image(self, url): try: url = url.strip(' "') r = requests.get(url, timeout=20) r.raise_for_status() import tempfile, os ext = url.split('.')[-1].lower().split('?')[0] or 'png' with tempfile.NamedTemporaryFile(suffix=f'.{ext}', delete=False) as f: f.write(r.content) tmp_path = f.name print(f" [image] downloaded {len(r.content)} bytes to {tmp_path}") img = Image.open(tmp_path) text = pytesseract.image_to_string(img) os.unlink(tmp_path) result = text.strip() return f"IMAGE_TEXT: {result}" if result else "IMAGE_TEXT: (no text found - this may be a diagram/photo)" except Exception as e: return f"Image error: {e}" # ── TOOL: Execute Python code ───────────────────────────────────── def run_python(self, url): try: url = url.strip(' "') # Try multiple URL patterns urls_to_try = [ url, url.replace('/files/', '/'), url.replace('https://agents-course-unit4-scoring.hf.space/files/', 'https://agents-course-unit4-scoring.hf.space/'), ] code = None for u in urls_to_try: r = requests.get(u, timeout=15) print(f" [python] trying {u} β†’ {r.status_code}") if r.status_code == 200 and len(r.text) > 10: code = r.text print(f" [python] got code ({len(code)} chars): {code[:150]}") break if not code: return f"Python error: file not found at any URL pattern" import io, contextlib stdout = io.StringIO() with contextlib.redirect_stdout(stdout): exec(code, {"__builtins__": __builtins__}) output = stdout.getvalue().strip() return f"PYTHON_OUTPUT: {output}" if output else f"PYTHON_CODE:\n{code[:500]}" except Exception as e: return f"Python exec error: {e}" # ── Route tool calls ────────────────────────────────────────────── def execute_tool(self, tool, input_data, file_url): # Use file_url as fallback when input_data has no URL target = input_data.strip(' "') if not target.startswith("http") and file_url: target = file_url if tool == "wiki_search": return self.wiki_search(input_data) elif tool == "scrape_page": # Extract key terms from tool_input or use question words return self.scrape_page(target, search_terms=input_data) elif tool == "read_audio": return self.read_audio(target) elif tool == "read_excel": return self.read_excel(target) elif tool == "read_image": return self.read_image(target) elif tool == "run_python": return self.run_python(target) elif tool == "web_search": return self.web_search(input_data) else: return f"Unknown tool: {tool}" # ── Main agent loop ─────────────────────────────────────────────── def agent_loop(self, question, file_url): print(f" [DEBUG] file_url received: {repr(file_url)}") pre_context = "" if file_url: ext = file_url.split('.')[-1].lower().split('?')[0] print(f" [Pre-load] detected file ext={ext}, url={file_url}") if ext in ['mp3', 'wav', 'ogg', 'm4a', 'flac']: pre_context = self.read_audio(file_url) elif ext in ['xlsx', 'xls', 'csv']: pre_context = self.read_excel(file_url) elif ext in ['png', 'jpg', 'jpeg', 'gif', 'webp']: pre_context = self.read_image(file_url) elif ext == 'py': try: pre_context = "PYTHON_CODE:\n" + requests.get(file_url, timeout=10).text[:3000] except: pass memory = pre_context seen_tool_calls = set() system_prompt = """You are a precise GAIA benchmark solver. STRICT OUTPUT FORMAT - choose exactly one: TOOL: tool_name INPUT: your_search_query_here OR: FINAL: your_answer NEVER write TOOL: wiki_search(query) - always use INPUT: on the next line. TOOL STRATEGY: - For Wikipedia questions: use scrape_page with the FULL Wikipedia URL directly e.g. TOOL: scrape_page INPUT: https://en.wikipedia.org/wiki/Mercedes_Sosa_discography - For web research: use wiki_search with short 2-4 word queries - For files: use read_audio / read_excel / read_image / run_python with the FILE_URL - Never repeat a failed tool - change approach each step KNOWN URLS (use these exactly when relevant): - LibreTexts 1.E Exercises (equine vet question): https://chem.libretexts.org/Bookshelves/Introductory_Chemistry/Introductory_Chemistry_(CK-12)/01%3A_Introduction_to_Chemistry/1.E%3A_Exercises_(CK-12) - Mercedes Sosa discography: https://en.wikipedia.org/wiki/Mercedes_Sosa_discography - 1928 Summer Olympics: https://en.wikipedia.org/wiki/1928_Summer_Olympics - Malko Competition: https://en.wikipedia.org/wiki/Malko_Competition - 1977 New York Yankees season stats: https://en.wikipedia.org/wiki/1977_New_York_Yankees_season - Taishō Tamai (baseball): https://en.wikipedia.org/wiki/Taish%C5%8D_Tamai - Wikipedia Featured articles November 2016: https://en.wikipedia.org/wiki/Wikipedia:Featured_article_candidates/Featured_log/November_2016 - Universe Today Carolyn Collins Petersen June 2023: https://web.archive.org/web/2023/https://www.universetoday.com/161812/ - Polish Everybody Loves Raymond (Świat wedΕ‚ug Kiepskich): https://en.wikipedia.org/wiki/Wszyscy_kochaj%C4%85_Raymonda_(Polish_TV_series) - Mercedes Sosa discography (use main article not redirect): https://en.wikipedia.org/wiki/Mercedes_Sosa FACTS YOU KNOW (no tools needed): - Reversed text questions: decode then answer directly as FINAL - When asked for "first name only", return ONLY the first word of the name - When asked for "surname only", return ONLY the last word - Basic math/logic: reason step by step then answer as FINAL - Botanical vegetables: only plant parts with NO seeds inside count as vegetables. Exclude: tomato, pepper, corn, zucchini, green beans, peas, cucumber, squash, acorns, peanuts. Include: broccoli, celery, lettuce, sweet potato, carrot.""" for step in range(10): prompt = f"""FILE_URL: {file_url if file_url else 'None'} QUESTION: {question} ACCUMULATED KNOWLEDGE: {memory if memory else '(none yet)'} AVAILABLE TOOLS: wiki_search, scrape_page, read_audio, read_excel, read_image, run_python, web_search What is your next action? Output TOOL+INPUT or FINAL:""" response = self.client.chat.completions.create( model="gpt-4o", temperature=0, messages=[ {"role": "system", "content": system_prompt}, {"role": "user", "content": prompt} ] ) resp = response.choices[0].message.content.strip() print(f" Step {step}: {resp[:120]}") # Check for final answer if "FINAL:" in resp: return resp.split("FINAL:")[-1].strip() # Parse tool call t_match = re.search(r"TOOL:\s*(\w+)(?:\(([^)]*)\))?", resp, re.I) i_match = re.search(r"INPUT:\s*(.+)", resp, re.I | re.DOTALL) if t_match: tool_name = t_match.group(1).lower().strip() if i_match: raw_input = i_match.group(1).strip() lines = raw_input.split('\n') tool_input = lines[0] if len(lines) > 1 and not lines[1].startswith('TOOL') and len(lines[1]) < 100: tool_input += lines[1].strip() tool_input = tool_input.strip() elif t_match.group(2): tool_input = t_match.group(2).strip() else: tool_input = "" # ── SKIP duplicate tool calls ── call_key = f"{tool_name}:{tool_input[:80]}" if call_key in seen_tool_calls: memory += f"\n\n[Step {step} - DUPLICATE SKIPPED: {call_key}. You already tried this. Use a DIFFERENT URL or approach.]" print(f" [DUPLICATE SKIPPED] {call_key}") continue seen_tool_calls.add(call_key) # ── CALL THE TOOL AND UPDATE MEMORY ── result = self.execute_tool(tool_name, tool_input, file_url) print(f" [{tool_name}] β†’ {result[:100]}") print(f" [RESULT LENGTH] {len(result)} chars: {result[:200]}") if len(result) > 30 and not result.lower().startswith("error") and not result.lower().startswith("unknown"): memory += f"\n\n[Step {step} - {tool_name}({tool_input[:80]})]\n{result[:2000]}" print(f" [MEMORY ADDED] memory now {len(memory)} chars") else: memory += f"\n\n[Step {step} - {tool_name} FAILED: {result[:200]}. Try a different approach.]" print(f" [MEMORY FAILED] result was: {result[:100]}") else: memory += f"\n\n[Step {step} - Reasoning]: {resp[:300]}" # Fallback fallback = self.client.chat.completions.create( model="gpt-4o", temperature=0, messages=[ {"role": "system", "content": system_prompt}, {"role": "user", "content": f"Based on everything gathered, give your best FINAL answer.\nQUESTION: {question}\nKNOWLEDGE:\n{memory}"} ] ) resp = fallback.choices[0].message.content.strip() if "FINAL:" in resp: return resp.split("FINAL:")[-1].strip() return resp def run_and_submit_all( profile: gr.OAuthProfile | None): """ Fetches all questions, runs the BasicAgent on them, submits all answers, and displays the results. """ # --- Determine HF Space Runtime URL and Repo URL --- space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code if profile: username= f"{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" # 1. Instantiate Agent ( modify this part to create your agent) try: agent = BasicAgent() 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: response = requests.get(questions_url, timeout=15) response.raise_for_status() questions_data = 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.") print("SAMPLE ITEM:", questions_data[0]) except requests.exceptions.RequestException as e: print(f"Error fetching questions: {e}") return f"Error fetching questions: {e}", None except requests.exceptions.JSONDecodeError as e: print(f"Error decoding JSON response from questions endpoint: {e}") print(f"Response text: {response.text[:500]}") 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...") 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: print(f"Skipping item with missing task_id or question: {item}") continue try: # submitted_answer = agent(question_text) //vineet file_name = item.get("file_name", "") task_id = item.get("task_id", "") if file_name: file_url = f"https://agents-course-unit4-scoring.hf.space/files/{file_name}" try: # Use GET with stream instead of HEAD (HEAD returns 405) test = requests.get(file_url, timeout=5, stream=True) test.close() print(f" [FILE] name='{file_name}', url={file_url}, status={test.status_code}") except Exception as e: print(f" [FILE] verification error: {e}") else: file_url = None print(f" [FILE] name={file_name!r}, url={file_url}") submitted_answer = agent(question_text, file_url) print("------------------------------------------------") print("QUESTION:", question_text) print("ANSWER:", submitted_answer) print("------------------------------------------------") 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 running agent on task {task_id}: {e}") results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"}) 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": 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: 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.") results_df = pd.DataFrame(results_log) return final_status, results_df 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 requests.exceptions.JSONDecodeError: error_detail += f" Response: {e.response.text[:500]}" status_message = f"Submission Failed: {error_detail}" print(status_message) results_df = pd.DataFrame(results_log) return status_message, results_df except requests.exceptions.Timeout: status_message = "Submission Failed: The request timed out." print(status_message) results_df = pd.DataFrame(results_log) return status_message, results_df except requests.exceptions.RequestException 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. """ ) 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) run_button.click( fn=run_and_submit_all, 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"βœ… 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"βœ… 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)