File size: 21,866 Bytes
e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 316fb4c e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 316fb4c 228aaa0 316fb4c 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 316fb4c 228aaa0 316fb4c e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 316fb4c e7c9c17 228aaa0 e7c9c17 316fb4c e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 228aaa0 e7c9c17 |
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 |
import asyncio
import os
import re
import json
import time
import zipfile
from urllib.parse import urljoin, urlparse
from typing import List, Dict, Any, Optional, Tuple, Set
import requests
import pandas as pd
from bs4 import BeautifulSoup
import gradio as gr
# =========================
# Config
# =========================
MAX_CONCURRENCY = 4 # concurrent pages to scrape
PLAYWRIGHT_WAIT_MS = 1500 # wait a bit for JS
FETCH_RETRIES = 2 # playwright retries per URL
SEARCH_PAGES = 2 # DDG result pages per query
RESULTS_PER_QUERY = 10 # target results per query
USER_AGENT = (
"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
"(KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36"
)
# =========================
# Optional LLM (OpenAI)
# =========================
def openai_extract_json(html: str, url: str, fields: List[str], api_key: Optional[str]) -> Optional[List[Dict[str, Any]]]:
if not api_key:
return None
try:
from openai import OpenAI
client = OpenAI(api_key=api_key)
field_hint = ", ".join(fields) if fields else "title, price, image, rating, url"
system = (
"You are a robust web extractor. Given raw HTML and the page URL, "
"return an array of JSON objects with fields you can infer (and the requested fields if present). "
"Always output strictly valid JSON with double-quoted keys/strings. Include absolute image URLs if possible."
)
user = (
f"URL: {url}\n\n"
f"Required fields to attempt: [{field_hint}]\n\n"
"Return JSON array only. Do not include any commentary.\n\n"
f"HTML:\n{html[:180000]}"
)
resp = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "system", "content": system}, {"role": "user", "content": user}],
temperature=0,
)
content = resp.choices[0].message.content.strip()
content = re.sub(r"^```(?:json)?|```$", "", content).strip()
data = json.loads(content)
if isinstance(data, dict):
data = [data]
if isinstance(data, list):
return data
return None
except Exception as e:
print("OpenAI extraction failed:", e)
return None
# =========================
# Playwright page loader (with retries)
# =========================
async def _fetch_dom_once(url: str, wait_ms: int) -> str:
from playwright.async_api import async_playwright
async with async_playwright() as p:
browser = await p.chromium.launch(headless=True)
page = await browser.new_page(user_agent=USER_AGENT)
await page.goto(url, wait_until="domcontentloaded", timeout=30000)
try:
await page.wait_for_load_state("networkidle", timeout=8000)
except Exception:
pass
if wait_ms > 0:
await asyncio.sleep(wait_ms / 1000)
html = await page.content()
await browser.close()
return html
async def fetch_dom(url: str, wait_ms: int = PLAYWRIGHT_WAIT_MS, retries: int = FETCH_RETRIES) -> str:
last_err = None
for attempt in range(1, retries + 2):
try:
return await _fetch_dom_once(url, wait_ms)
except Exception as e:
last_err = e
await asyncio.sleep(0.6 * attempt)
raise last_err
# =========================
# Heuristic extraction
# =========================
def extract_images_and_items(html: str, base_url: str, card_selector: Optional[str] = None) -> Tuple[List[Dict[str, Any]], List[str]]:
soup = BeautifulSoup(html, "html.parser")
# Collect all images on page
images = []
for img in soup.find_all("img"):
src = img.get("src") or img.get("data-src") or img.get("data-original")
if not src:
continue
abs_src = urljoin(base_url, src)
images.append(abs_src)
# Find likely product/article cards
items = []
if card_selector:
candidates = soup.select(card_selector)
else:
candidates = soup.select(
"div.product, li.product, div.card, article, div.product-item, "
"div.s-result-item, div._1AtVbE, div._4ddWXP, div.MuiCard-root, "
"section, li.grid-item"
)
if not candidates:
candidates = [a.parent for a in soup.select("a img") if a.parent]
for c in candidates:
try:
title = None
for sel in ["h1", "h2", "h3", ".title", ".product-title", "._4rR01T", ".s1Q9rs", "a[title]"]:
n = c.select_one(sel)
if n and n.get_text(strip=True):
title = n.get_text(strip=True)
break
if not title:
img = c.find("img")
if img and img.get("alt"):
title = img.get("alt").strip()
price = None
price_text = c.get_text(" ", strip=True)
m = re.search(r"(?:βΉ|Rs\.?|INR|\$|β¬|Β£)\s?\d[\d,]*(?:\.\d+)?", price_text)
if m:
price = m.group(0)
link = c.find("a")
href = urljoin(base_url, link.get("href")) if link and link.get("href") else base_url
img = c.find("img")
img_src = None
if img:
img_src = img.get("src") or img.get("data-src") or img.get("data-original")
if img_src:
img_src = urljoin(base_url, img_src)
if any([title, price, img_src]):
items.append({"title": title, "price": price, "url": href, "image": img_src})
except Exception:
continue
# De-duplicate images
seen = set()
unique_images = []
for u in images:
if u not in seen:
seen.add(u)
unique_images.append(u)
return items, unique_images
# =========================
# Image download & optional captioning
# =========================
def download_images(image_urls: List[str], out_dir: str) -> List[str]:
os.makedirs(out_dir, exist_ok=True)
saved = []
s = requests.Session()
s.headers.update({"User-Agent": USER_AGENT})
for u in image_urls:
try:
name = os.path.basename(urlparse(u).path) or f"img_{len(saved)+1}.jpg"
if not os.path.splitext(name)[1]:
name += ".jpg"
path = os.path.join(out_dir, name)
r = s.get(u, timeout=20)
if r.status_code == 200 and r.content:
with open(path, "wb") as f:
f.write(r.content)
saved.append(path)
except Exception as e:
print("Image download failed:", u, e)
return saved
def caption_images(paths: List[str]) -> Dict[str, str]:
try:
from transformers import BlipProcessor, BlipForConditionalGeneration
from PIL import Image
import torch
device = "cuda" if torch.cuda.is_available() else "cpu"
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base").to(device)
captions = {}
for p in paths:
try:
im = Image.open(p).convert("RGB")
inputs = processor(im, return_tensors="pt").to(device)
out = model.generate(**inputs, max_new_tokens=40)
text = processor.decode(out[0], skip_special_tokens=True)
captions[p] = text
except Exception as e:
captions[p] = f"(caption failed: {e})"
return captions
except Exception as e:
print("Captioning unavailable:", e)
return {}
# =========================
# ZIP helper
# =========================
def zip_paths(paths: List[str], zip_path: str) -> str:
with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zf:
for p in paths:
if os.path.isfile(p):
zf.write(p, arcname=os.path.basename(p))
return zip_path
# =========================
# Search helpers (Prompt β Queries β Links)
# =========================
ADS_PRESETS = [
# public/archival ad sources (safer than scraping walled platforms)
"site:adsoftheworld.com",
"site:theinspiration.com",
"site:ads-of-the-world.s3", # mirrors sometimes
"site:behance.net ad campaign",
"site:dribbble.com case study ad",
]
NEWS_SIGNAL = [
"site:news.ycombinator.com", "site:techcrunch.com", "site:theverge.com",
"site:adage.com", "site:campaignlive.com"
]
def build_queries_from_prompt(prompt: str, include_ads_sources: bool) -> List[str]:
# very lightweight keyword clean
base = re.sub(r"[^a-zA-Z0-9\s:+\-_/\.]", " ", prompt).strip()
base = re.sub(r"\s+", " ", base)
core_variants = [
base,
f'{base} best examples',
f'{base} recent campaigns',
f'{base} case study',
f'{base} images',
]
queries = []
for v in core_variants:
queries.append(v)
# tilt towards news relevance
for ns in NEWS_SIGNAL[:2]:
queries.append(f"{v} {ns}")
if include_ads_sources:
for v in core_variants:
for siteq in ADS_PRESETS:
queries.append(f"{v} {siteq}")
# de-dup while keeping order
seen = set()
uniq = []
for q in queries:
if q not in seen:
seen.add(q)
uniq.append(q)
return uniq[:12] # cap
def ddg_search(query: str, pages: int = 1) -> List[Tuple[str, str]]:
"""
Returns list of (title, url) from DuckDuckGo HTML results, across pages.
"""
results = []
session = requests.Session()
session.headers.update({"User-Agent": USER_AGENT})
for page in range(pages):
params = {"q": query}
if page > 0:
params["s"] = str(page * 50) # pagination hint
r = session.get("https://duckduckgo.com/html/", params=params, timeout=20)
soup = BeautifulSoup(r.text, "html.parser")
for res in soup.select(".result"):
a = res.select_one(".result__a")
if not a:
continue
title = a.get_text(strip=True)
href = a.get("href")
if not href:
continue
results.append((title, href))
return results
def pick_best_links(all_results: List[Tuple[str, str]], want: int = 10) -> List[str]:
"""
Simple pragmatic ranking:
- de-duplicate by URL & domain
- prefer diverse domains
"""
picked = []
seen_urls: Set[str] = set()
seen_domains: Set[str] = set()
for _, url in all_results:
u = url.strip()
if not u or u in seen_urls:
continue
dom = urlparse(u).netloc.lower()
if dom.startswith("www."):
dom = dom[4:]
# skip obvious DDG redirectors or trackers if any
if dom in {"duckduckgo.com"}:
continue
if dom in seen_domains and len(picked) < want // 2:
# allow later, but early phase enforce domain diversity
continue
seen_urls.add(u)
seen_domains.add(dom)
picked.append(u)
if len(picked) >= want:
break
return picked
def search_links_from_prompt(prompt: str, include_ads_sources: bool, per_query: int, pages: int) -> List[str]:
queries = build_queries_from_prompt(prompt, include_ads_sources)
all_results: List[Tuple[str, str]] = []
for q in queries:
try:
res = ddg_search(q, pages=pages)
# take top-k per query
all_results.extend(res[:per_query])
except Exception as e:
print("Search failed for query:", q, e)
continue
# global pick
best = pick_best_links(all_results, want=max(5, per_query * 2))
return best
# =========================
# Main scrape orchestrator (async with semaphore)
# =========================
async def scrape_one(url: str, fields: List[str], use_llm: bool, api_key: Optional[str],
card_selector: Optional[str], log: List[str], sem: asyncio.Semaphore) -> Dict[str, Any]:
async with sem:
try:
html = await fetch_dom(url)
except Exception as e:
log.append(f"[ERROR] Failed to load: {url} -> {e}")
return {"url": url, "html": "", "items": [], "images": [], "llm_rows": []}
items, images = [], []
try:
items, images = extract_images_and_items(html, url, card_selector)
except Exception as e:
log.append(f"[WARN] Parse issue on: {url} -> {e}")
llm_rows = []
if use_llm:
try:
llm_rows = openai_extract_json(html, url, fields, api_key) or []
except Exception as e:
log.append(f"[WARN] LLM extraction failed: {url} -> {e}")
return {"url": url, "html": html, "items": items, "images": images, "llm_rows": llm_rows}
def to_dataframe(rows: List[Dict[str, Any]]) -> pd.DataFrame:
if not rows:
return pd.DataFrame()
all_keys = set()
for r in rows:
all_keys.update(r.keys())
ordered = []
for r in rows:
d = {k: r.get(k) for k in all_keys}
ordered.append(d)
df = pd.DataFrame(ordered)
preferred = [k for k in ["title", "name", "price", "rating", "image", "url"] if k in df.columns]
others = [c for c in df.columns if c not in preferred]
df = df[preferred + others]
return df
# =========================
# Gradio wrapper
# =========================
def run_scrape(input_mode: str,
prompt_or_urls: str,
fields_text: str,
card_selector: str,
include_ads_sources: bool,
per_query_results: int,
search_pages: int,
use_llm: bool,
api_key: str,
download_imgs: bool,
do_caption: bool):
start = time.time()
log: List[str] = []
# Resolve URLs
if input_mode == "Prompt":
if not prompt_or_urls.strip():
return pd.DataFrame(), [], None, None, None, "Enter a prompt.", "No prompt given."
log.append(f"[INFO] Building queries from prompt: {prompt_or_urls!r}")
urls = search_links_from_prompt(
prompt_or_urls.strip(),
include_ads_sources=include_ads_sources,
per_query=per_query_results,
pages=max(1, search_pages)
)
if not urls:
return pd.DataFrame(), [], None, None, None, "No links found.", "\n".join(log)
log.append(f"[INFO] Selected {len(urls)} links from search.")
else:
urls = [u.strip() for u in prompt_or_urls.splitlines() if u.strip()]
if not urls:
return pd.DataFrame(), [], None, None, None, "Enter at least one URL.", "No URLs supplied."
log.append(f"[INFO] Using {len(urls)} direct URL(s).")
fields = [f.strip() for f in fields_text.split(',')] if fields_text.strip() else []
out_dir = os.path.abspath("scrape_output")
os.makedirs(out_dir, exist_ok=True)
# Async scrape with semaphore
sem = asyncio.Semaphore(MAX_CONCURRENCY)
async def gather_all():
tasks = [
scrape_one(u, fields, use_llm, api_key if use_llm else None, card_selector or None, log, sem)
for u in urls
]
return await asyncio.gather(*tasks)
try:
scraped = asyncio.run(gather_all())
except RuntimeError:
scraped = asyncio.get_event_loop().run_until_complete(gather_all())
except Exception as e:
log.append(f"[FATAL] Async run failed: {e}")
return pd.DataFrame(), [], None, None, None, "Run failed.", "\n".join(log)
heuristic_rows: List[Dict[str, Any]] = []
llm_rows: List[Dict[str, Any]] = []
all_images: List[str] = []
for s in scraped:
if not isinstance(s, dict):
continue
heuristic_rows.extend(s.get("items", []))
llm_rows.extend(s.get("llm_rows", []))
all_images.extend(s.get("images", []))
# prefer LLM rows if available
rows = llm_rows if use_llm and llm_rows else heuristic_rows
df = to_dataframe(rows)
ts = int(time.time())
json_path = os.path.join(out_dir, f"scrape_{ts}.json")
csv_path = os.path.join(out_dir, f"scrape_{ts}.csv")
try:
df.to_csv(csv_path, index=False)
with open(json_path, "w", encoding="utf-8") as f:
json.dump(rows, f, ensure_ascii=False, indent=2)
except Exception as e:
log.append(f"[WARN] Failed to save CSV/JSON: {e}")
json_path = None
csv_path = None
gallery_paths, zip_path = [], None
if download_imgs and all_images:
try:
img_dir = os.path.join(out_dir, f"images_{ts}")
saved = download_images(all_images, img_dir)
gallery_paths = saved[:120]
if do_caption and saved:
try:
captions_map = caption_images(saved)
if not df.empty:
img_col = None
for c in df.columns:
if c.lower() in ("image", "image_url", "img", "imageurl"):
img_col = c
break
if img_col:
def _map_caption(u):
if not u:
return ""
fname = os.path.basename(urlparse(str(u)).path)
return captions_map.get(os.path.join(img_dir, fname), "")
df["caption"] = df[img_col].map(_map_caption)
df.to_csv(csv_path, index=False)
with open(json_path, "w", encoding="utf-8") as f:
json.dump(json.loads(df.to_json(orient="records")), f, ensure_ascii=False, indent=2)
except Exception as e:
log.append(f"[WARN] Captioning failed: {e}")
zip_path = os.path.join(out_dir, f"images_{ts}.zip")
try:
zip_paths(saved, zip_path)
except Exception as e:
log.append(f"[WARN] ZIP failed: {e}")
zip_path = None
except Exception as e:
log.append(f"[WARN] Image pipeline failed: {e}")
elapsed = round(time.time() - start, 2)
gallery_data = [(p, os.path.basename(p)) for p in gallery_paths]
status = f"Scraped {len(urls)} URL(s) β’ Rows: {len(df)} β’ Images found: {len(all_images)} β’ Time: {elapsed}s"
return df, gallery_data, (json_path if json_path and os.path.isfile(json_path) else None), \
(csv_path if csv_path and os.path.isfile(csv_path) else None), \
(zip_path if zip_path and os.path.isfile(zip_path) else None), \
status, "\n".join(log) if log else "OK"
# =========================
# Gradio UI
# =========================
with gr.Blocks(title="AI Scraper β Prompt β Best Links β Text+Images", css=".gradio-container {max-width: 1200px !important}") as demo:
gr.Markdown("""
# π·οΈ AI-Powered Prompt Scraper (2025)
- Give a **prompt** (e.g., "Gen Z pink organic skincare ad campaign in India 2024")
β we search smartly, pick strong links (optionally ad archives), and scrape **text + images**
- Or switch to **Direct URLs** mode and paste URLs.
- Optional **LLM semantic parsing** to structured JSON.
""")
with gr.Row():
input_mode = gr.Radio(choices=["Prompt", "Direct URLs"], value="Prompt", label="Input Mode")
with gr.Row():
prompt_or_urls = gr.Textbox(
label="Prompt (or URLs if in Direct mode)",
placeholder="e.g., gen z pink skincare ad campaign india 2024"
)
with gr.Row():
fields = gr.Textbox(label="Fields to extract (comma-separated)", placeholder="title, price, image, rating, url")
card_selector = gr.Textbox(label="Optional CSS selector for item cards", placeholder="div.product, article, .card")
with gr.Row():
include_ads_sources = gr.Checkbox(label="Bias search towards ad archives/sources", value=True)
per_query_results = gr.Slider(1, 15, value=6, step=1, label="Top results to keep per query")
search_pages = gr.Slider(1, 3, value=2, step=1, label="Search pages per query (DDG)")
with gr.Row():
use_llm = gr.Checkbox(label="Use OpenAI for semantic extraction", value=False)
api_key = gr.Textbox(label="OpenAI API Key (if using LLM)", type="password")
with gr.Row():
download_imgs = gr.Checkbox(label="Download images", value=True)
do_caption = gr.Checkbox(label="Caption images (slow)", value=False)
run_btn = gr.Button("π Run Scraper", variant="primary")
with gr.Row():
table = gr.Dataframe(label="Extracted Data (preview)", interactive=False)
gallery = gr.Gallery(label="Scraped Images (subset)", show_label=True, height=420, allow_preview=True)
with gr.Row():
json_file = gr.File(label="Download JSON")
csv_file = gr.File(label="Download CSV")
zip_file = gr.File(label="Download Images ZIP")
status = gr.Markdown("Ready.")
logs = gr.Textbox(label="Run Logs", lines=10)
run_btn.click(
fn=run_scrape,
inputs=[
input_mode, prompt_or_urls, fields, card_selector,
include_ads_sources, per_query_results, search_pages,
use_llm, api_key, download_imgs, do_caption
],
outputs=[table, gallery, json_file, csv_file, zip_file, status, logs]
)
if __name__ == "__main__":
demo.launch() |