Extractly 🕸️

Turn Any Webpage Into a Spreadsheet Using Plain English

No code. No browser extensions. No local installs. Just paste a URL, describe what you want, and download your CSV.

---

Try It In 10 Seconds

Paste any of these into the app to see it in action:

URL	Prompt
https://news.ycombinator.com	Extract the title and link of each news item
https://books.toscrape.com	Extract the titles and prices of all books
https://quotes.toscrape.com	Extract the quote text, author, and tags

---

Who Is This For?

• Researchers who need data from a webpage without writing Python
• Journalists who want to turn a table or list into a spreadsheet
• Product managers who want competitor pricing or feature lists
• Data enthusiasts who know what they want but not how to get it
• QA engineers exploring NLP-driven browser automation

---

How It Works

You describe it → AI identifies the structure → Playwright extracts it → You download the CSV

1. Paste a URL — any publicly accessible webpage
2. Describe what you want in plain English — "Extract the title and price of each product" — "Get the author, quote, and tags from each entry"
3. Extractly navigates the page server-side using a headless Chromium browser — no extension needed
4. An LLM analyzes the page structure and identifies the right CSS selectors automatically
5. Your data appears as a clean preview table with a one-click CSV download

---

Under the Hood: Advanced Extraction Engineering

Extractly isn't just a simple raw HTML scraper. It employs several advanced runtime engineering mechanisms to ensure selector stability and high extraction accuracy:

• DFS Accessibility Tree Serialization: Performs a Depth-First Search (DFS) traversal of the Chrome Accessibility Tree via direct Chrome DevTools Protocol (CDP) sessions. This preserves semantic reading order and document hierarchy, preventing token truncation and layout-ordering issues.
• Semantic Text Hoisting: Automatically hoists text content from inner StaticText and InlineTextBox child nodes to their parent semantic elements, providing full contextual layout info to the LLM.
• Self-Correcting Selector Guards:
  • Absolute Selector Stripping: Sanitizes selectors to prevent the LLM from generating brittle absolute selector paths.
  • Parent Self-Match Auto-Correction: Detects if a relative sub-field selector matches the card container itself. Automatically falls back to query narrower sub-elements (e.g. .author, small, .tags, .price) to prevent retrieving the entire card block.
  • List & Tag Formatting: Detects fields representing multiple inline tags or categories. Automatically cleanses label text (removing prefixes like "by " or "(about)"), strips formatting whitespace, and joins items into a single-line comma-separated string.
  • Main Link & Utility Filtering: Discriminates utility links (like category tags, upvotes, flags, or comments) from the primary destination URL to guarantee correct outbound links.

---

What Makes This Different

Most scraping tools require either coding knowledge or expensive subscriptions. Extractly fills the gap:

	Extractly	Code (Playwright/BS4)	Paid Platforms
Setup required	None	Node/Python install	Account + subscription
Technical knowledge	None	Required	Some
Cost	Free	Free	$$$
Plain English input	✅	❌	Partial
Runs in browser	✅	❌	✅

---

Privacy & Compliance

• robots.txt viewer built in — check a site's scraping guidelines before you extract
• SSRF protection — private IPs and localhost are blocked server-side
• No data retention — extracted data lives only in your session and is never stored
• Users are responsible for ensuring their use complies with target site terms of service

---

Tech Stack

Layer	Technology
Backend	Python 3.11, FastAPI, Uvicorn
Headless Browser	Playwright (Chromium, server-side via CDP)
AI / NLP	Groq Cloud — llama-3.1-8b-instant
Data	pandas
Frontend	Jinja2 + Vanilla CSS (Glassmorphic Dark Theme with One-Click Quick-Try Chips)
Deployment	HuggingFace Spaces, Docker

---

Local Development

Prerequisites

• Python 3.11
• Groq API key — free at console.groq.com

Setup

git clone https://github.com/joshchamo/extractly.git
cd extractly
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
playwright install chromium

Run

export GROQ_API_KEY="your-key-here"
uvicorn app.main:app --host 0.0.0.0 --port 7860

Open http://localhost:7860

---

Limitations

• Single page only — pagination is not yet supported
• JavaScript-heavy SPAs may require additional load time
• Results depend on page structure — highly dynamic or obfuscated sites may not extract cleanly
• Respects a 20 second navigation timeout per page

---

Built with FastAPI + Playwright + Groq. Part of an ongoing series of AI automation tools.
Developer: Josh Chamo [joshchamo@gmail.com]