📄 PDF Text Extractor & Chunker

Extract clean, ordered text from any PDF on the web — page-by-page or split into LLM-ready chunks with controllable size and overlap. Point it at one URL or thousands; results stream into your dataset section by section, live.

Perfect for building RAG pipelines, question-answering systems, document search, and any workflow that needs PDF content as plain text. 🚀

🌟 Why Choose This Actor?

• ⚡ Live results — every page/chunk is saved the moment it's ready. A long run never leaves you staring at an empty output table.
• 🧩 LLM-friendly chunking — character-based chunking with overlap, so context isn't cut mid-sentence.
• 📦 Bulk input — drop in a whole list of PDF URLs at once.
• 🛡️ Smart anti-rate-limit ladder — starts with a direct connection and automatically falls back to datacenter, then residential proxies if a host blocks you.
• 🎉 Engaging real-time logs — watch exactly what's happening, page by page.

✨ Key Features

• Extract text from PDFs provided as URLs.
• Toggle between page mode (one record per page) and chunk mode.
• Configure chunkSize and chunkOverlap for perfect LLM context windows.
• Resilient downloads with proxy fallback and retries.
• Output ready for JSON / CSV / XLSX export.

📥 Input

Field	Type	Description
urls	array	🔗 Direct URLs of the PDF files (bulk supported).
performChunking	boolean	✂️ true → split into chunks. false → one record per page.
chunkSize	integer	📏 Max characters per chunk (chunk mode). Default 1000.
chunkOverlap	integer	🔁 Characters shared between adjacent chunks. Default 0.
proxyConfiguration	object	🛡️ Apify proxy used to power the automatic fallbacks.

Example input

{
  "urls": ["https://arxiv.org/pdf/2307.12856"],
  "performChunking": true,
  "chunkSize": 1000,
  "chunkOverlap": 0,
  "proxyConfiguration": { "useApifyProxy": true }
}

📤 Output

Each record is one text section:

{
  "url": "https://arxiv.org/pdf/2307.12856",
  "index": 0,
  "text": "A Real-World WebAgent with Planning, Long Context Understanding…"
}

Field	Description
url	🔗 Source PDF URL.
index	🔢 Position of the section (chunk number, or page number in page mode).
text	📝 Extracted text for that section.

🛡️ How the connection ladder works

1. 🌐 Direct — no proxy; the request goes straight to the PDF host.
2. 🛰️ Datacenter proxy — engaged automatically if the host blocks or rate-limits the direct request.
3. 🏠 Residential proxy — the final fallback, retried up to 3 times. Once residential is engaged, the run sticks with it for every remaining PDF.

Every switch is logged clearly so you always know which path delivered your data.

🚀 How to Use (Apify Console)

1. Log in at Apify Console → Actors.
2. Open PDF Text Extractor & Chunker.
3. Paste your PDF URLs, set chunking options, pick a proxy.
4. Click Start and watch the sections roll in live. 📡
5. Open the Output tab and export to JSON / CSV / XLSX.

🤖 Use via API

curl -X POST "https://api.apify.com/v2/acts/<ACTOR_ID>/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
     -H "Content-Type: application/json" \
     -d '{"urls":["https://arxiv.org/pdf/2307.12856"],"performChunking":true,"chunkSize":1000,"chunkOverlap":0}'

💡 Best Use Cases

• 📚 Build RAG / knowledge bases from PDF libraries.
• 🤖 Feed document text into LLMs (chunk mode).
• 🔍 Full-text search across PDF collections.
• 🧾 Convert reports, papers, and manuals to plain text.

❓ FAQ

Does it work on scanned/image-only PDFs? It extracts the text layer of a PDF. Image-only scans without an embedded text layer will return little or no text (OCR is not performed).

Can I pass many URLs? Yes — urls accepts a bulk list, processed one after another with results saved live.

What if a host rate-limits me? The Actor automatically falls back through datacenter and residential proxies and retries, then sticks with residential.

🛟 Support & Feedback

Found a bug or have a feature request? Open an issue on the Actor's Issues tab in the Apify Console.

---

⚖️ Use responsibly. Only extract content from PDFs you are authorized to access. You are responsible for compliance with applicable laws and the source site's terms.