Spaces:
Sleeping
Sleeping
File size: 42,721 Bytes
d0490ec 6efe12c b86e7fd d0490ec 575233a e93cb13 e946b39 4697ce0 ae4a304 b86e7fd eae276f a12a1d9 6efe12c d0490ec 6efe12c d0490ec d62ee7c ae4a304 6efe12c d0490ec 6efe12c d0490ec 0b83ba2 d0490ec a12a1d9 d0490ec 8e91397 d0490ec a12a1d9 d0490ec a12a1d9 8e91397 a12a1d9 8e91397 a12a1d9 8e91397 a12a1d9 8e91397 d0490ec 8e91397 0b83ba2 d0490ec 6efe12c d0490ec a12a1d9 eae276f a12a1d9 eae276f a12a1d9 d0490ec a12a1d9 d0490ec a12a1d9 d0490ec a12a1d9 d0490ec da2d730 a12a1d9 da2d730 b86e7fd da2d730 b86e7fd da2d730 b86e7fd da2d730 b86e7fd da2d730 b86e7fd 4697ce0 a12a1d9 4697ce0 a12a1d9 4697ce0 a12a1d9 4697ce0 eae276f a12a1d9 4697ce0 da2d730 c803551 a12a1d9 c803551 a12a1d9 c803551 a12a1d9 c803551 4cad7a4 c803551 8e63348 eae276f 8e63348 b86e7fd 8e63348 872fb01 b86e7fd 872fb01 b86e7fd 0b83ba2 b86e7fd 872fb01 b86e7fd eae276f 8e63348 b86e7fd 8e63348 b86e7fd 8e63348 a12a1d9 8e63348 b86e7fd 8e63348 b86e7fd 8e63348 eae276f 8e63348 eae276f b86e7fd eae276f b86e7fd eae276f b86e7fd c803551 8e63348 eae276f a12a1d9 eae276f a12a1d9 eae276f a12a1d9 eae276f b86e7fd eae276f a12a1d9 eae276f a12a1d9 eae276f a12a1d9 eae276f a12a1d9 2af4dae 2cb6927 2af4dae a12a1d9 d0490ec f3b56d3 932aa92 a12a1d9 b86e7fd 0b83ba2 a12a1d9 b86e7fd a12a1d9 b86e7fd a12a1d9 324000b 0b83ba2 b86e7fd f3b56d3 d0490ec 75c9b07 d0490ec 75c9b07 d0490ec b86e7fd 2af4dae a12a1d9 7db1ce5 b57c693 2af4dae a12a1d9 51b4921 b86e7fd 2af4dae a12a1d9 0b83ba2 91279e0 b57c693 91279e0 2af4dae b57c693 0b83ba2 b86e7fd 75c9b07 b86e7fd 0b83ba2 92c77a0 0b83ba2 d0490ec 75c9b07 4697ce0 75c9b07 d0490ec 75c9b07 d0490ec 75c9b07 d0490ec 75c9b07 d0490ec 75c9b07 d0490ec 75c9b07 4697ce0 75c9b07 d0490ec 75c9b07 d0490ec ae4a304 e93cb13 47d092c 4697ce0 d0490ec 2948a14 ead2a02 6efe12c e946b39 da2d730 44e8e4f ae4a304 bb4d19c da2d730 f80a1f8 0a04e0d da2d730 ae4a304 44e8e4f e946b39 556e783 44e8e4f ae4a304 c0b248d 51cc59e 8e63348 7db1ce5 c0b248d e946b39 d0490ec a12a1d9 d0490ec ae4a304 2af4dae d766d61 f7728a8 c0b248d f7728a8 c0b248d d766d61 2af4dae d766d61 d0490ec a12a1d9 d0490ec 095cb52 d0490ec 6efe12c d0490ec 095cb52 02bdc0b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 |
import os
import gradio as gr
import requests, tempfile, base64, json, datetime, re, subprocess, mimetypes, fitz
import pandas as pd
from langchain.tools import tool
from langchain_huggingface import HuggingFaceEndpoint, ChatHuggingFace
from langchain.agents import initialize_agent, AgentType
from bs4 import BeautifulSoup
from langchain_openai import ChatOpenAI
from langchain_community.utilities import ArxivAPIWrapper
from youtube_transcript_api import YouTubeTranscriptApi
import yt_dlp
from PIL import Image
from transformers import pipeline
## # Load environment variables from .env file
# --- Constants ---
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
# Load the environment variables
HF_ACCESS_KEY = os.getenv('HF_ACCESS_KEY')
WEATHER_API_KEY = os.getenv('WEATHER_API_KEY')
OPENAI_KEY = os.getenv('OPENAI_KEY')
OPENAI_MODEL = os.getenv ('OPENAI_MODEL')
########## ----- DEFINING TOOLS -----##########
# --- TOOL 1: Web Search Tool (DuckDuckGo) ---
@tool
def current_events_news_search_tool(query: str) -> str:
"""
General web search tool for current events, news, or trending topics not yet on Wikipedia.
Returns relevant context and source URL if available.
"""
url = f"https://api.duckduckgo.com/?q={query}&format=json&no_html=1"
try:
resp = requests.get(url, timeout=30)
resp.raise_for_status()
data = resp.json()
# Check main answer fields
for key in ["AbstractText", "Answer", "Definition"]:
if data.get(key):
answer = data[key].strip()
break
else:
answer = None
# Try to extract more from RelatedTopics
if not answer:
related = data.get("RelatedTopics")
if related and isinstance(related, list):
for topic in related:
if isinstance(topic, dict) and topic.get("Text"):
answer = topic["Text"].strip()
# Optionally, add the URL
if topic.get("FirstURL"):
answer += f"\nSource: {topic['FirstURL']}"
break
# Try to extract from Results
if not answer:
results = data.get("Results")
if results and isinstance(results, list):
for result in results:
if isinstance(result, dict) and result.get("Text"):
answer = result["Text"].strip()
if result.get("FirstURL"):
answer += f"\nSource: {result['FirstURL']}"
break
# Fallback: return "no_answer"
if answer:
return answer
return "no_answer"
except Exception as e:
return f"error: {e}"
# when you use the @tool decorator from langchain.tools, the tool.name and tool.description are automatically extracted from your function
# tool.name is set to the function name (e.g., `search_tool`), and
# tool.description is set to the docstring of the function (the triple-quoted string right under def ...) (e.g., "Answer general knowledge or current events queries using DuckDuckGo.").
# --- TOOL 3: Calculator Tool ---
@tool
def calculator(expression: str) -> str:
"""Evaluate math expressions."""
try:
allowed = "0123456789+-*/(). "
if not all(c in allowed for c in expression):
return "error"
result = eval(expression, {"__builtins__": None}, {})
return str(result)
except Exception:
return "error"
# --- TOOL 6: Wikipedia Summary Tool ---
@tool
def wikipedia_and_generalknowledge_search(query: str) -> str:
"""
Answer questions related to general knowledge, world information, facts, sports, olympics, history, etc. from Wikipedia by scraping the text and returns text as context for LLM to use.
"""
# Step 1: Search Wikipedia for the most relevant page
search_url = "https://en.wikipedia.org/w/api.php"
params = {
"action": "query",
"list": "search",
"srsearch": query,
"format": "json"
}
try:
resp = requests.get(search_url, params=params, timeout=150)
resp.raise_for_status()
results = resp.json().get("query", {}).get("search", [])
if not results:
return "no_answer"
page_title = results[0]["title"]
page_url = f"https://en.wikipedia.org/wiki/{page_title.replace(' ', '_')}"
except Exception:
return "error: Could not search Wikipedia"
# Step 2: Fetch the Wikipedia page and extract main text
try:
page_resp = requests.get(page_url, timeout=120)
page_resp.raise_for_status()
soup = BeautifulSoup(page_resp.text, "html.parser")
output = f"Source: {page_url}\n"
# Extract main text from all paragraphs
paragraphs = soup.find_all("p")
text = " ".join(p.get_text(separator=" ", strip=True) for p in paragraphs)
# Limit to first 3000 characters for brevity
output += text[:3000] if text else "No textual content found."
return output
except Exception as e:
return f"error: {e}"
# --- TOOL 9: Image Captioning Tool ---
@tool
def image_caption(image_url: str) -> str:
"""Generate a descriptive caption for an image given its URL."""
api_url = "https://api-inference.huggingface.co/models/Salesforce/blip-image-captioning-base"
headers = {"Authorization": f"Bearer {HF_ACCESS_KEY}"}
payload = {"inputs": image_url}
try:
resp = requests.post(api_url, headers=headers, json=payload, timeout=120)
resp.raise_for_status()
data = resp.json()
return data[0]["generated_text"] if isinstance(data, list) else data.get("generated_text", "no_caption")
except Exception:
return "error"
# --- TOOL 10: Optical Character Recognition (OCR) Tool ---
@tool
def ocr_image(image_url: str) -> str:
"""
Extracts all readable text from an image using HuggingFace TrOCR (microsoft/trocr-base-stage1).
Input: URL to an image (e.g., PNG or JPG).
Output: Recognized text string.
"""
api_url = "https://api-inference.huggingface.co/models/microsoft/trocr-base-stage1"
headers = {
"Authorization": f"Bearer {HF_ACCESS_KEY}",
"Content-Type": "application/json"
}
payload = {"inputs": image_url}
try:
resp = requests.post(api_url, headers=headers, json=payload, timeout=60)
resp.raise_for_status()
data = resp.json()
return data[0]["generated_text"]
except Exception as e:
return f"OCR error: {e}"
# --- TOOL 11: Image Classification Tool ---
@tool
def clasify_describe_image(image_url: str) -> str:
"""
Generates a caption describing the contents of an image using HuggingFace (ViT-GPT2).
Use this tool to identify the main subject of an image so that an LLM can use it to answer further.
Input: image URL
Output: caption like 'A golden retriever lying on a couch.'
"""
api_url = "https://api-inference.huggingface.co/models/nlpconnect/vit-gpt2-image-captioning"
headers = {"Authorization": f"Bearer {HF_ACCESS_KEY}"}
try:
img_resp = requests.get(image_url, timeout=120)
img_resp.raise_for_status()
image_bytes = img_resp.content
response = requests.post(api_url, headers=headers, data=image_bytes, timeout=60)
response.raise_for_status()
result = response.json()
return result[0]["generated_text"] if isinstance(result, list) else "no_caption"
except Exception as e:
return f"caption error: {e}"
# --- TOOL 12: Web Scraping Tool ---
@tool
def URL_scrape_tool(url: str) -> str:
"""
Scrape the main textual content from a given website URL and returns the text - to be used as context by model.
"""
try:
headers = {
"User-Agent": "Mozilla/5.0 (compatible; WebScrapeTool/1.0)"
}
resp = requests.get(url, headers=headers, timeout=120)
resp.raise_for_status()
soup = BeautifulSoup(resp.text, "html.parser")
# Try to extract main content from common tags
paragraphs = soup.find_all("p")
text = " ".join(p.get_text() for p in paragraphs)
# Limit to first 2000 characters for brevity
return text[:4000] if text else "No textual content found."
except Exception as e:
return f"error: {e}"
# --- TOOL 13: Audio to Text Transcription Tool ---
@tool
def audio_url_to_text(audio_url: str) -> str:
"""
Transcribe speech from an audio file URL to text using Hugging Face's Whisper model.
Input: A direct link to an audio file (e.g., .mp3, .wav).
Output: The transcribed text.
"""
api_url = "https://api-inference.huggingface.co/models/openai/whisper-large-v3"
headers = {"Authorization": f"Bearer {HF_ACCESS_KEY}"}
try:
# Download the audio file
audio_resp = requests.get(audio_url, timeout=120)
audio_resp.raise_for_status()
audio_bytes = audio_resp.content
# Encode audio as base64 for API
audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
payload = {
"inputs": audio_b64,
"parameters": {"return_timestamps": False}
}
resp = requests.post(api_url, headers=headers, json=payload, timeout=120)
resp.raise_for_status()
data = resp.json()
return data.get("text", "no_answer")
except Exception as e:
return f"error: {e}"
# --- TOOL 14: Python Code Executor Tool ---
@tool
def python_executor(code: str) -> str:
"""
Safely execute simple Python code and return the result if the code is in the question. If the question has .py file attached, use 'python_excel_audio_video_attached_file_tool' tool first.
Only supports expressions and basic statements (no imports, file I/O, or system access).
"""
try:
# Restrict built-ins for safety
allowed_builtins = {"abs": abs, "min": min, "max": max, "sum": sum, "len": len, "range": range}
# Only allow expressions, not statements
result = eval(code, {"__builtins__": allowed_builtins}, {})
return str(result)
except Exception as e:
return f"error: {e}"
# --- TOOL 15: Attachment Processing Tool ---
@tool
def python_excel_audio_video_attached_file_tool(input_str: str) -> str:
"""
Accepts a JSON string with one of:
• 'file_bytes' : base-64–encoded bytes (existing behaviour)
• 'file_path' : local absolute/relative path to a file
• 'file_url' : downloadable URL (e.g. Hugging Face dataset link)
Keys (at least one bytes / path / url required):
• filename (str) – original name with extension
• file_bytes (str, base-64) – optional
• file_path (str) – optional
• file_url (str) – optional
Returns: textual summary / preview ready for the LLM.
"""
# ---------- 1. Parse JSON ------------------------------------------------
try:
# Robustly pull out the first {...} block even if extra tokens are around it
match = re.search(r'(\{.*\})', input_str, re.DOTALL)
payload = json.loads(match.group(1) if match else input_str)
except Exception as e:
return f"error: Could not parse JSON → {e}"
filename = payload.get("filename")
b64_data = payload.get("file_bytes")
file_path = payload.get("file_path")
file_url = payload.get("file_url")
if not filename:
return "error: 'filename' is required."
# ---------- 2. Acquire raw bytes ----------------------------------------
try:
if b64_data: # inline bytes
file_bytes = base64.b64decode(b64_data)
elif file_path and os.path.exists(file_path): # local path
with open(file_path, "rb") as f:
file_bytes = f.read()
elif file_url: # remote URL
# stream to avoid loading huge files into memory at once
r = requests.get(file_url, timeout=60, stream=True)
r.raise_for_status()
file_bytes = r.content
else:
return "error: Provide 'file_bytes', 'file_path', or 'file_url'."
except Exception as e:
return f"error: Could not load file → {e}"
# Detect file type
mime_type, _ = mimetypes.guess_type(filename)
# fallback for common extensions if guess_type fails
if not mime_type:
ext = filename.lower()
mime_type = (
"text/x-python" if ext.endswith(".py") else
"text/csv" if ext.endswith(".csv") else
"application/vnd.ms-excel" if ext.endswith((".xls", ".xlsx")) else
None
)
if not mime_type:
return "error: Could not determine file type. Skip the file."
# Handle audio files
if mime_type.startswith("audio"):
api_url = "https://api-inference.huggingface.co/models/openai/whisper-large-v3"
headers = {"Authorization": f"Bearer {HF_ACCESS_KEY}"}
files = {"file": (filename, file_bytes)}
try:
resp = requests.post(api_url, headers=headers, files=files, timeout=120)
resp.raise_for_status()
data = resp.json()
transcript = data.get("text", "")
if transcript:
return f"Transcript of the audio: {transcript}"
else:
return "error: No transcript returned."
except Exception as e:
return f"error: {e}"
# Handle image files
elif mime_type.startswith("image"):
# image_b64 = base64.b64encode(file_bytes).decode()
api_url = "https://api-inference.huggingface.co/models/nlpconnect/vit-gpt2-image-captioning"
headers = {"Authorization": f"Bearer {os.getenv('HF_ACCESS_KEY', '')}"}
try:
resp = requests.post(api_url, headers=headers, data=file_bytes, timeout=60)
resp.raise_for_status()
result = resp.json()
if isinstance(result, list) and result and "generated_text" in result[0]:
caption = result[0]["generated_text"]
else:
caption = "no_caption"
# Optionally also include base-64 so the LLM can refer to the raw image
b64 = base64.b64encode(file_bytes).decode()
return f"Image caption: {caption}\nAttached image (base64): {b64}"
except Exception as e:
return f"caption error: {e}"
return f"Attached image (base64): {image_b64}"
# Handle video files (extract audio, then transcribe)
elif mime_type.startswith("video"):
try:
with tempfile.NamedTemporaryFile(delete=False, suffix=filename.split('.')[-1]) as tmp_video:
tmp_video.write(file_bytes)
tmp_video.flush()
video_path = tmp_video.name
audio_path = video_path + ".wav"
# import subprocess
subprocess.run([
"ffmpeg", "-i", video_path, "-vn", "-acodec", "pcm_s16le", "-ar", "16000", "-ac", "1", audio_path
], check=True)
with open(audio_path, "rb") as f:
audio_bytes = f.read()
api_url = "https://api-inference.huggingface.co/models/openai/whisper-large-v3"
headers = {"Authorization": f"Bearer {HF_ACCESS_KEY}"}
files = {"file": ("audio.wav", audio_bytes)}
resp = requests.post(api_url, headers=headers, files=files, timeout=120)
resp.raise_for_status()
data = resp.json()
transcript = data.get("text", "")
if transcript:
return f"Transcript of the video audio: {transcript}"
else:
return "error: No transcript returned from video audio."
except Exception as e:
return f"error: {e}"
# Handle Excel files (.xls, .xlsx, .csv)
elif mime_type in ["application/vnd.ms-excel", "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet", "text/csv"]:
try:
with tempfile.NamedTemporaryFile(delete=False, suffix=filename.split('.')[-1]) as tmp_excel:
tmp_excel.write(file_bytes)
tmp_excel.flush()
excel_path = tmp_excel.name
if filename.lower().endswith(".csv"):
df = pd.read_csv(excel_path)
preview = df.head(500).to_csv(index=False)
return f"CSV file preview (first 5 rows):\n{preview}"
else:
xl = pd.ExcelFile(excel_path)
sheet_names = xl.sheet_names
preview = ""
for sheet in sheet_names:
df = xl.parse(sheet)
preview += f"\nSheet: {sheet}\n{df.head(500).to_csv(index=False)}"
return f"Excel file sheets: {sheet_names}\nPreview (first 3 rows per sheet):{preview}"
except Exception as e:
return f"error: {e}"
# Handle Python files (.py)
elif mime_type == "text/x-python" or filename.lower().endswith(".py"):
try:
code = file_bytes.decode("utf-8", errors="replace")
lines = code.splitlines()
preview = "\n".join(lines[:40])
return f"Python file preview (first 40 lines):\n{preview}"
except Exception as e:
return f"error: {e}"
else:
return "error: Unsupported file type. Please skip the file usage."
# --- TOOL 16: Research Paper Info Extraction Tool ---
@tool
def research_paper_search(query: str) -> str:
"""
Search arXiv for journals/research/technical papers matching a query.
Returns top results including title, authors, abstract, and PDF link.
"""
wrapper = ArxivAPIWrapper(
top_k_results=2, # how many papers to return
doc_content_chars_max=2000 # max chars of abstract to show
)
results_text = wrapper.run(query)
return results_text
# --- TOOL 17:Tool for sports, awards, competitions etc. ---
@tool
def sports_awards_historicalfacts_tool(query: str) -> str:
"""
For questions about sports, awards, competitions, historical facts, or generic wikipedia available data, this tool fetches relevant context from Wikipedia.
"""
# Step 1: Search Wikipedia for the most relevant page
search_url = "https://en.wikipedia.org/w/api.php"
params = {
"action": "query",
"list": "search",
"srsearch": query,
"format": "json"
}
try:
resp = requests.get(search_url, params=params, timeout=150)
resp.raise_for_status()
results = resp.json().get("query", {}).get("search", [])
if not results:
return "no_answer"
page_title = results[0]["title"]
page_url = f"https://en.wikipedia.org/wiki/{page_title.replace(' ', '_')}"
except Exception:
return "error: Could not search Wikipedia"
# Step 2: Fetch the Wikipedia page and extract tables and lists
try:
page_resp = requests.get(page_url, timeout=150)
page_resp.raise_for_status()
soup = BeautifulSoup(page_resp.text, "html.parser")
output = f"Source: {page_url}\n"
# Extract all tables with relevant columns
tables = soup.find_all("table", {"class": ["wikitable", "sortable"]})
found_table = False
for table in tables:
table_str = str(table)
if any(word in table_str.lower() for word in ["winner", "name", "year", "nationality", "country"]):
try:
df = pd.read_html(table_str)[0]
output += "\n--- Extracted Table ---\n"
output += df.to_csv(index=False)
found_table = True
except Exception:
continue
# If no relevant table, extract lists (e.g., <ul> or <ol> with <li>)
if not found_table:
lists = soup.find_all(['ul', 'ol'])
for lst in lists:
items = lst.find_all('li')
if len(items) > 2: # Only consider lists with more than 2 items
output += "\n--- Extracted List ---\n"
for item in items:
text = item.get_text(separator=" ", strip=True)
output += f"{text}\n"
break # Only include the first relevant list
# Fallback: return the first paragraph if nothing else
if not found_table and "--- Extracted List ---" not in output:
first_p = soup.find("p")
output += first_p.get_text(strip=True)[:500] if first_p else "no_answer"
# Limit output length for LLM context
return output[:3500]
except Exception as e:
return f"error: {e}"
# --- TOOL 17: YouTube Transcript Tool ---
@tool
def youtube_transcript_tool(video_url: str) -> str:
"""
Get transcript (if available) for a YouTube video without downloading audio.
Works only if subtitles or auto-captions exist.
"""
try:
# Extract video ID
match = re.search(r"(?:v=|youtu\.be/)([a-zA-Z0-9_-]{11})", video_url)
if not match:
return "Invalid YouTube URL."
video_id = match.group(1)
transcript = YouTubeTranscriptApi.get_transcript(video_id)
full_text = " ".join([chunk['text'] for chunk in transcript])
return full_text[:5000] # truncate to keep LLM input manageable
except Exception as e:
return f"Transcript error: {e}"
# --- TOOL 18: YouTube Transcript Tool ---
@tool
def video_url_to_transcript_tool(media_url: str) -> str:
"""
Given a URL to a video or audio file (YouTube, direct .mp4/.mp3/.wav, etc.), download the audio and return a transcript.
"""
api_url = "https://api-inference.huggingface.co/models/openai/whisper-large-v3"
headers = {"Authorization": f"Bearer {HF_ACCESS_KEY}"}
try:
with tempfile.TemporaryDirectory() as tmpdir:
audio_path = None
# Check if it's a YouTube URL
if "youtube.com" in media_url or "youtu.be" in media_url:
ydl_opts = {
'format': 'bestaudio/best',
'outtmpl': f'{tmpdir}/audio.%(ext)s',
'quiet': True,
'noplaylist': True,
'extractaudio': True,
'audioformat': 'wav',
'postprocessors': [{
'key': 'FFmpegExtractAudio',
'preferredcodec': 'wav',
'preferredquality': '192',
}],
}
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
info = ydl.extract_info(media_url, download=True)
audio_path = ydl.prepare_filename(info).rsplit('.', 1)[0] + '.wav'
else:
# Download direct media file
resp = requests.get(media_url, timeout=120)
resp.raise_for_status()
# Guess extension
ext = media_url.split('?')[0].split('.')[-1].lower()
if ext not in ["mp3", "wav", "m4a", "mp4"]:
ext = "mp3"
file_path = os.path.join(tmpdir, f"audio.{ext}")
with open(file_path, "wb") as f:
f.write(resp.content)
# If video, extract audio using ffmpeg
if ext in ["mp4", "mkv", "webm"]:
audio_path = os.path.join(tmpdir, "audio.wav")
import subprocess
subprocess.run([
"ffmpeg", "-i", file_path, "-vn", "-acodec", "pcm_s16le", "-ar", "16000", "-ac", "1", audio_path
], check=True)
else:
audio_path = file_path
# Read audio bytes
with open(audio_path, "rb") as f:
audio_bytes = f.read()
# Encode audio as base64 for API
audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
payload = {
"inputs": audio_b64,
"parameters": {"return_timestamps": False}
}
resp = requests.post(api_url, headers=headers, json=payload, timeout=120)
resp.raise_for_status()
data = resp.json()
return data.get("text", "no_answer")
except Exception as e:
return f"error: {e}"
# --- TOOL 19: Audio to Text Transcription Tool ---
@tool
def max_object_in_video(video_url: str, object_label: str = "bird") -> str:
"""
Given a video URL and an object label, extracts frames and uses an object detection model to count the specified object in each frame.
Returns the maximum number of objects detected in any single frame.
Example: max_object_in_video("https://...", "car") -> "Maximum car count in a frame: 4"
"""
# Download video
try:
resp = requests.get(video_url, timeout=120)
resp.raise_for_status()
with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as tmp_video:
tmp_video.write(resp.content)
tmp_video.flush()
video_path = tmp_video.name
except Exception as e:
return f"error: Could not download video: {e}"
# Extract frames every 2 seconds (adjust as needed)
frames_dir = tempfile.mkdtemp()
frame_pattern = os.path.join(frames_dir, "frame_%04d.jpg")
try:
subprocess.run([
"ffmpeg", "-i", video_path, "-vf", "fps=0.5", frame_pattern
], check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
except Exception as e:
return f"error: Could not extract frames: {e}"
# Load object detection pipeline
try:
detector = pipeline("object-detection", model="facebook/detr-resnet-50")
except Exception as e:
return f"error: Could not load detection model: {e}"
max_count = 0
for fname in sorted(os.listdir(frames_dir)):
fpath = os.path.join(frames_dir, fname)
try:
image = Image.open(fpath)
results = detector(image)
count = sum(1 for obj in results if obj['label'].lower() == object_label.lower() and obj['score'] > 0.5)
if count > max_count:
max_count = count
except Exception:
continue
# Clean up
try:
os.remove(video_path)
for fname in os.listdir(frames_dir):
os.remove(os.path.join(frames_dir, fname))
os.rmdir(frames_dir)
except Exception:
pass
return f"Maximum {object_label} count in a single frame: {max_count}"
'''
def extract_final_answer(output: str) -> str:
# Try to extract answer after [YOUR FINAL ANSWER] or Final Answer:
match = re.search(r"\[YOUR FINAL ANSWER\]\s*(.+)", output)
if match:
return match.group(1).strip()
match = re.search(r"Final Answer:\s*(.+)", output)
if match:
return match.group(1).strip()
# Fallback: return the whole output if no match
return output.strip()
'''
##-- Tool Discovery ---
# Use @tool for each function.
# Use get_all_tools() to auto-discover all decorated tools.
# tools_list = get_all_tools()
tools_list = [
python_excel_audio_video_attached_file_tool,
wikipedia_and_generalknowledge_search,
# sports_awards_historicalfacts_tool,
research_paper_search,
python_executor,
# get_weather,
# calculator,
# convert_units,
# get_time,
# get_date,
# dictionary_lookup,
# currency_convert,
# image_caption,
# ocr_image,
# classify_image,
current_events_news_search_tool,
ocr_image,
clasify_describe_image,
URL_scrape_tool,
# audio_url_to_text,
# sports_awards_historicalfacts_tool,
youtube_transcript_tool,
# video_url_to_transcript_tool,
max_object_in_video,
]
tool_descriptions = "\n".join(f"- {tool.name}: {tool.description}" for tool in tools_list)
## --
# --- System Prompt for the Agent ---
system_prompt = f"""
You are a general AI assistant, who can answer about general knowledge, historical facts, and also can analyze audios, images, and videos. You should think through the input question step-by-step and use tools if needed.
Use this reasoning format repeatedly:
Thought: (what you think is happening or what you want to do next)
Action: (the tool to use, if needed)
Action Input: (input to the tool)
Observation: (result of the tool call)
Repeat this process as needed. ONLY AFTER finishing your reasoning and/or tool use, provide YOUR FINAL ANSWER
Your output should be just a number, string, or comma-separated list. Don't give your Thoughts, Actions, Observations or any other descriptions.
You also have access to a set of tools, which you can use to answer the question. The available tools are:
{tool_descriptions}
If the question is related to sports, awards, historical facts or similar topic that can be answered from wikipedia, you should use the 'wikipedia_and_generalknowledge_search'.
If the question is about current events or news or similar current affairs category, you can utilize the tool 'current_events_news_search_tool' to fetch relevant page information and answer from it.
If the tool returns a long text, table, or list, extract only the most relevant information/paragraphs or data from which you can derive the answer, and return that as your final answer.
You must not use multiple tools in a single call. Don't hallucinate.
**Examples:**
Q: Which country had the least number of athletes at the 1928 Summer Olympics?
Your Output: Luxembourg
Q: What are the top 3 programming languages?
Your Output: Python, JavaScript, Java
If even after 12 iterations, a tool usage is not useful then try to answer directly based on your knowledge without any hallucination. If you cannot answer then just say "no_answer" as YOUR FINAL ANSWER.
"""
# If your final answer is something like 'there were 5 studio albums published between 2000 and 2009' then modify YOUR FINAL ANSWER as: '5'
# If your final answer is something like 'b, e' then YOUR FINAL ANSWER be: 'b, e'
# For each question, follow this format:
# Question: the input question you must answer
# Thought: your reasoning about what to do next
# Action: the action to take, must be one of the tools. If no relevant tools, answer the question directly.
# Action Input: the input to the action
# Observation: the result of the action
# ... (repeat Thought/Action/Action Input/Observation as needed)
# Final Answer: the answer to the original question, as concise as possible (number, short string, or comma-separated list, no extra explanation).
# system_prompt = f"""
# You are an intelligent assistant with access to the following tools:
# {tool_descriptions}
# For every question, you must do your internal reasoning using the Thought → Action → Observation → Answer process, but your output to the user should be ONLY the final answer as a single value (number, string, or comma-separated list), with no extra explanation, thoughts, actions, or observations.
# **If a tool returns a long text or description (such as from a web scraping tool), you must carefully read and process that output, and extract or identify ONLY the most relevant, concise answer to the user's question, and provide a single string as output. Do not return the full text or irrelevant details.**
# **Your output must be only the answer. Do not include any reasoning, tool calls, or explanations.**
# Examples:
# Q: What is 7 * (3 + 2)?
# Your Output: 35
# Q: What’s the weather in Tokyo?
# Your Output: 22
# Q: What is the capital of France?
# Your Output: Paris
# Q: Which year was python 3.0 released as per the website https://en.wikipedia.org/wiki/Python_(programming_language)?
# (Tool returns a long description about Python.)
# Your Output: 2008
# Q: Convert 10 meters to feet.
# Your Output: 32.81
# Instructions:
# - Always do your internal reasoning (Thought → Action → Observation → Answer) before producing the answer, but DO NOT show this reasoning to the user.
# - Use a tool only if necessary, and don't use multiple tools in a call. Don't use a tool if you can answer directly.
# - Your output must be a single value (number, string, or comma-separated list) with no extra explanation or formatting.
# - If you cannot answer the question or if you couldn't process the input question just answer as "no_answer".
# - Be concise and accurate.
# """
## --- Initialize Hugging Face Model ---
# Generate the chat interface, including the tools
'''
llm = HuggingFaceEndpoint(
repo_id="meta-llama/Llama-3.3-70B-Instruct",
# repo_id="Qwen/Qwen2.5-32B-Instruct",
huggingfacehub_api_token=HF_ACCESS_KEY,
# model_kwargs={'prompt': system_prompt}
# system_prompt=system_prompt,
)
chat_llm = ChatHuggingFace(llm=llm)
'''
# Initialize the OpenAI chat model
chat_llm = ChatOpenAI(
openai_api_key=OPENAI_KEY,
model_name=OPENAI_MODEL,
temperature=0.05,
# max_tokens=10
)
# Initialize the agent with the tools and system prompt
agent = initialize_agent(
tools=tools_list,
# llm=llm,
llm=chat_llm,
agent=AgentType.OPENAI_FUNCTIONS,#AgentType.ZERO_SHOT_REACT_DESCRIPTION,
agent_kwargs={"system_message": system_prompt},
verbose=True,
max_iterations=15, # Increase as needed
max_execution_time=4000, # Increase as needed
early_stopping_method="generate",
handle_parsing_errors=True,
# return_intermediate_steps=False
)
## --
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=120)
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.")
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:
# full_prompt = f"{system_prompt}\n Input Question: {question_text}"
# submitted_answer = agent.run(full_prompt)
# submitted_answer_raw = agent.run(question_text)
submitted_answer = agent.run(question_text)
'''
if "YOUR FINAL ANSWER:" in submitted_answer:
match = re.search(r"YOUR FINAL ANSWER:\s*(.+)", submitted_answer, re.IGNORECASE | re.DOTALL)
scraped_answer = match.group(1).strip()
else:
scraped_answer = submitted_answer.strip()
'''
# submitted_answer = extract_final_answer(submitted_answer_raw)
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=120)
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()
# login_btn = gr.LoginButton()
# login_btn.activate()
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"✅ 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"✅ 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...")
# Launch the Gradio app
demo.launch(debug=True, share=True) #share=True |