awacke1/ComputerUseSeleniumPlaywrightDifflibSKLearnImagehash
1
1import streamlit as st2import requests3from bs4 import BeautifulSoup4from PIL import Image5import io6import base647from urllib.parse import urljoin, urlparse8import pandas as pd9import plotly.graph_objects as go10import numpy as np11from difflib import SequenceMatcher12from sklearn.feature_extraction.text import TfidfVectorizer13from sklearn.metrics.pairwise import cosine_similarity14import time15import asyncio16from playwright.sync_api import sync_playwright17import sys18import subprocess19 20def install_playwright_deps():21 try:22 from playwright.sync_api import sync_playwright23 # Install browsers if not already installed24 subprocess.run(['playwright', 'install'], check=True)25 except Exception as e:26 st.error(f"Error installing Playwright dependencies: {str(e)}")27 st.info("Try running 'pip install playwright' and 'playwright install' manually")28 29def initialize_session_state():30 if 'visited_urls' not in st.session_state:31 st.session_state.visited_urls = []32 if 'load_times' not in st.session_state:33 st.session_state.load_times = []34 if 'screenshots' not in st.session_state:35 st.session_state.screenshots = []36 if 'crawl_results' not in st.session_state:37 st.session_state.crawl_results = []38 39def setup_browser():40 """Initialize Playwright browser"""41 try:42 playwright = sync_playwright().start()43 browser = playwright.chromium.launch(headless=True)44 return playwright, browser45 except Exception as e:46 st.error(f"Error setting up browser: {str(e)}")47 return None, None48 49def capture_screenshot(page):50 """Capture screenshot using Playwright"""51 screenshot_bytes = page.screenshot()52 return Image.open(io.BytesIO(screenshot_bytes))53 54def calculate_similarity(text1, text2):55 # Basic similarity56 basic_ratio = SequenceMatcher(None, text1, text2).ratio()57 58 # Semantic similarity59 vectorizer = TfidfVectorizer()60 try:61 tfidf = vectorizer.fit_transform([text1, text2])62 semantic_ratio = cosine_similarity(tfidf[0:1], tfidf[1:2])[0][0]63 except:64 semantic_ratio = 065 66 return basic_ratio, semantic_ratio67 68async def crawl_website(url, max_pages=10, search_term=None):69 visited = set()70 to_visit = {url}71 results = []72 73 try:74 with sync_playwright() as p:75 browser = p.chromium.launch(headless=True)76 page = browser.new_page()77 78 while to_visit and len(visited) < max_pages:79 current_url = to_visit.pop()80 if current_url in visited:81 continue82 83 try:84 page.goto(current_url, wait_until="networkidle")85 visited.add(current_url)86 87 # Extract text content88 text_content = page.content()89 90 # If search term provided, check for matches91 match_found = search_term.lower() in text_content.lower() if search_term else True92 93 if match_found:94 results.append({95 'url': current_url,96 'title': page.title(),97 'content_preview': text_content[:200],98 'matches_search': match_found99 })100 101 # Find new links102 links = page.eval_on_selector_all('a[href]', 'elements => elements.map(el => el.href)')103 for href in links:104 absolute_url = urljoin(current_url, href)105 if urlparse(absolute_url).netloc == urlparse(url).netloc:106 to_visit.add(absolute_url)107 108 except Exception as e:109 st.error(f"Error crawling {current_url}: {str(e)}")110 111 browser.close()112 113 except Exception as e:114 st.error(f"Error in crawl process: {str(e)}")115 116 return results117 118def main():119 st.title("Web Testing and Crawling Suite")120 initialize_session_state()121 122 # Install dependencies if needed123 with st.spinner("Checking dependencies..."):124 install_playwright_deps()125 126 # Sidebar for tool selection127 tool = st.sidebar.radio(128 "Select Tool",129 ["WebTest", "Crawler", "AI Content Comparison"]130 )131 132 if tool == "WebTest":133 st.header("WebTest - Web Performance Testing")134 url = st.text_input("Enter URL to test")135 interval = st.slider("Time interval between requests (seconds)", 1, 30, 5)136 max_cycles = st.number_input("Number of test cycles", 1, 100, 1)137 138 if st.button("Start Testing"):139 playwright, browser = setup_browser()140 if playwright and browser:141 try:142 page = browser.new_page()143 144 for cycle in range(max_cycles):145 start_time = time.time()146 147 try:148 page.goto(url, wait_until="networkidle")149 load_time = time.time() - start_time150 st.session_state.load_times.append(load_time)151 152 # Capture screenshot153 screenshot = capture_screenshot(page)154 st.session_state.screenshots.append(screenshot)155 156 # Show results157 st.success(f"Cycle {cycle + 1} completed - Load time: {load_time:.2f}s")158 st.image(screenshot, caption=f"Screenshot - Cycle {cycle + 1}")159 160 # Plot load times161 fig = go.Figure(data=go.Scatter(162 x=list(range(1, len(st.session_state.load_times) + 1)),163 y=st.session_state.load_times,164 mode='lines+markers'165 ))166 fig.update_layout(title="Page Load Times", 167 xaxis_title="Cycle",168 yaxis_title="Load Time (s)")169 st.plotly_chart(fig)170 171 time.sleep(interval)172 173 except Exception as e:174 st.error(f"Error in cycle {cycle + 1}: {str(e)}")175 176 finally:177 browser.close()178 playwright.stop()179 180 elif tool == "Crawler":181 st.header("Web Crawler")182 base_url = st.text_input("Enter base URL to crawl")183 max_pages = st.number_input("Maximum pages to crawl", 1, 100, 10)184 search_term = st.text_input("Search term (optional)")185 186 if st.button("Start Crawling"):187 results = asyncio.run(crawl_website(base_url, max_pages, search_term))188 st.session_state.crawl_results = results189 190 # Display results191 df = pd.DataFrame(results)192 st.dataframe(df)193 194 # Export options195 if st.button("Export Results"):196 csv = df.to_csv(index=False)197 b64 = base64.b64encode(csv.encode()).decode()198 href = f'<a href="data:file/csv;base64,{b64}" download="crawl_results.csv">Download CSV</a>'199 st.markdown(href, unsafe_allow_html=True)200 201 else: # AI Content Comparison202 st.header("AI Content Comparison")203 url1 = st.text_input("Enter first URL (AI-generated content)")204 url2 = st.text_input("Enter second URL (Comparison content)")205 206 if st.button("Compare Content"):207 playwright, browser = setup_browser()208 if playwright and browser:209 try:210 page = browser.new_page()211 212 # Get content from first URL213 page.goto(url1, wait_until="networkidle")214 content1 = page.content()215 216 # Get content from second URL217 page.goto(url2, wait_until="networkidle")218 content2 = page.content()219 220 # Calculate similarities221 basic_ratio, semantic_ratio = calculate_similarity(content1, content2)222 223 # Display results224 st.subheader("Similarity Results")225 col1, col2 = st.columns(2)226 227 with col1:228 st.metric("Basic Similarity", f"{basic_ratio:.2%}")229 230 with col2:231 st.metric("Semantic Similarity", f"{semantic_ratio:.2%}")232 233 # Show content previews234 st.subheader("Content Previews")235 st.text_area("Content 1 (First 500 chars)", content1[:500])236 st.text_area("Content 2 (First 500 chars)", content2[:500])237 238 finally:239 browser.close()240 playwright.stop()241 242if __name__ == "__main__":243 main()244 245Goals="""246Computer Use2471. Browser based testing app2482. similar to apps I wrote years ago which would operate a browser then run tests against my web apps including being able to compare any image or text content together to search results from one of my ai programs to determine content overlap which is then used to evaluate the results and update my ai model context data to store anything that was found that adds to the original idea. When I looked at this problem before I found chrome driver for automatic testing, saucelabs which can kind of do it, and then some python testing libraries which could do it. Can you enlighten me on which python libraries and potenitally dev tools which would help me with this to automate my testing and evaluation of my ai generated content which resides at many different URLs on huggingface as running apps2493. Past apps per wayback from 2004: 250 - https://web.archive.org/web/20040520102150/http://www.evolvable.com/EStore/251 252WebTest 8.0253WebTest is a stress and load testing browser.254You can use WebTest to identify defects that occur when web sites incurr a large amount of traffic.255To use WebTest, simply visit the pages that you want to test and WebTest remembers your navigation history. You can save the history to a text file that you can open from other machines or other copies of WebTest running on your computer. To perform stress testing, set the time interval you would like WebTest to visit each Url, then set the cycle option to cycle through your site list.256 257Features Include:258Screen Captures259Graph of Page Load Times260Memory of Visited Pages261Tunable Delay and Maximum Wait Time262Hotkeys for Typical Browser Resolution Sizing263XML/XSL based Usability Reporting264 265Crawler 1.0266 267Have you ever wanted to analyze a web site to see what it contains yet didn't want to spend the next few days traversing all the links on the site?268 269With Crawler, you can enter a web address and Crawler visits every page, extracting a reusable list of links from the pages it reviews.270You can save the HTML source code from each page that Crawler encounters so you can analyze the site further, or automatically surf the site and take snapshots using Evolvable WebTest.271 272Crawler is an indespensable tool because it provides results in an easy to use format and also allows you to search for terms across different sites and save the list of links that match your search criteria.273 274Let me help you explore the Python libraries and tools that would be well-suited for automated browser testing and content comparison for AI-generated content.275The most popular and robust options for this type of testing are:276 277Selenium with Python278 279pythonCopyfrom selenium import webdriver280from selenium.webdriver.common.by import By281 282driver = webdriver.Chrome()283driver.get("https://huggingface.co/your-app")284content = driver.find_element(By.CLASS_NAME, "content").text285 286Playwright (more modern alternative to Selenium)287 288pythonCopyfrom playwright.sync_api import sync_playwright289 290with sync_playwright() as p:291 browser = p.chromium.launch()292 page = browser.new_page()293 page.goto("https://huggingface.co/your-app")294 content = page.inner_text(".content")295 296For content comparison and similarity analysis:297 298pythonCopyfrom difflib import SequenceMatcher299from sklearn.feature_extraction.text import TfidfVectorizer300from sklearn.metrics.pairwise import cosine_similarity301 302# Basic text similarity303def similarity_ratio(text1, text2):304 return SequenceMatcher(None, text1, text2).ratio()305 306# More advanced semantic similarity307def semantic_similarity(text1, text2):308 vectorizer = TfidfVectorizer()309 tfidf = vectorizer.fit_transform([text1, text2])310 return cosine_similarity(tfidf[0:1], tfidf[1:2])[0][0]311For image comparison:312pythonCopyfrom PIL import Image313import imagehash314import numpy as np315 316def compare_images(image1_path, image2_path):317 hash1 = imagehash.average_hash(Image.open(image1_path))318 hash2 = imagehash.average_hash(Image.open(image2_path))319 return hash1 - hash2 # Lower number means more similar320Would you like me to create a more comprehensive example that combines these elements into a complete testing framework? I could show you how to:321 322Automate browsing multiple HuggingFace URLs323Extract both text and image content324Compare content using different similarity metrics325Store results in a structured format for model context updates326 327"""328 329st.markdown(Goals)330