PDF Intelligence - AI-Powered PDF Analysis, OCR & RAG Preparation

Extract text, tables, and AI insights from any PDF in seconds.

Transform PDFs into structured, actionable data with AI-powered analysis. Extract text with 95%+ accuracy, automatically OCR scanned documents, detect tables with AI precision, and prepare content for RAG workflows.

---

Quick Start

Get results in 30 seconds:

1. Click Start - the default example PDF runs automatically
2. View results in the Output tab
3. Switch to AI Analysis view for intelligent insights

No configuration needed for basic extraction!

---

What This Actor Does

Core Features Text Extraction - Clean text from any PDF AI-Powered OCR - Convert scanned PDFs to text Table Detection - Extract structured table data RAG Chunking - Split for vector databases AI Analysis - Summary, entities, classification

Output Includes Executive summary of document Document type classification Named entity extraction (people, orgs, dates) Key topics and themes Action items and recommendations Quality score and confidence level

---

Pricing

Transparent pay-per-use pricing. Only pay for what you process.

Base Processing

Event	Price	Description
Page Processed	$0.002	Per PDF page extracted
Document Analyzed	$0.01	Metadata extraction
RAG Chunking	$0.02	Chunk preparation

AI Features (Require API Key)

Event	Price	Description
OCR Page	$0.03	AI Vision OCR per page
AI Table Extraction	$0.015	Intelligent table detection
AI Document Analysis	$0.04	Full AI analysis

Pricing Examples

Use Case	What You Get	Cost
10-page PDF text extraction	Text + metadata	~$0.03
50-page PDF with AI analysis	Text + AI insights	~$0.14
RAG preparation (20 pages)	Chunks ready for vectors	~$0.06
Scanned PDF OCR (5 pages)	OCR text + analysis	~$0.19

---

How to Use

Option 1: Apify Console (Easiest)

1. Enter your PDF URL in the PDF URL field
2. Select an action (Extract Text, Extract Tables, etc.)
3. Click Start
4. View results in the Output tab

Option 2: Apify API

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });

const run = await client.actor('cvs/pdf-intelligence').call({
    pdfUrl: 'https://example.com/document.pdf',
    action: 'full_analysis'
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items[0]);

Option 3: Direct HTTP API

curl -X POST "https://api.apify.com/v2/acts/cvs~pdf-intelligence/runs?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "pdfUrl": "https://example.com/document.pdf",
    "action": "extract_text"
  }'

Option 4: Claude Desktop (MCP)

Add to your Claude Desktop config:

{
  "mcpServers": {
    "pdf-intelligence": {
      "url": "https://cvs--pdf-intelligence.apify.actor/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_APIFY_TOKEN"
      }
    }
  }
}

---

Input Parameters

Basic Settings

Parameter	Type	Default	Description
pdfUrl	string	Example PDF	URL of PDF to process
pdfContent	string	-	Base64-encoded PDF (alternative to URL)
action	string	extract_text	What to extract (see below)
maxPages	integer	0 (all)	Limit pages to process

Actions Available

Action	Description
extract_text	Get all text content with page markers
extract_tables	Extract tabular data as JSON/CSV/Markdown
get_metadata	Document properties (title, author, dates)
chunk_for_rag	Split into chunks for vector databases
full_analysis	All of the above combined

AI Configuration

Parameter	Type	Description
googleApiKey	string	Google API key for Gemini (recommended)
openaiApiKey	string	OpenAI key for GPT-4 Vision OCR
anthropicApiKey	string	Anthropic key for Claude Vision
preferredAiProvider	string	"auto", "gemini", "openai", or "anthropic"

---

Output Format

Example Output

{
  "success": true,
  "overview": {
    "summary": "Technical report on web accessibility guidelines...",
    "documentType": "technical",
    "keyFindings": ["Contains accessibility standards", "Includes implementation examples"],
    "confidence": "high"
  },
  "stats": {
    "pageCount": 12,
    "wordCount": 3450,
    "tableCount": 3,
    "chunkCount": 15,
    "processingTimeMs": 2340
  },
  "quality": {
    "score": 92,
    "issues": [],
    "recommendations": []
  },
  "content": {
    "text": "Full extracted text...",
    "tables": [...],
    "metadata": {...}
  }
}

Output Views in Console

View	What It Shows
Summary	AI-generated executive summary
AI Analysis	Entities, topics, action items
Quality Report	Score, confidence, recommendations
Metadata	Title, author, dates, page count
Content	Extracted text and tables
RAG Chunks	Prepared chunks for vector DBs
Full Output	Complete raw JSON

---

Use Cases

📄 Invoice Processing

Extract line items, totals, and vendor information automatically.

{
  "pdfUrl": "https://example.com/invoice.pdf",
  "action": "extract_tables",
  "googleApiKey": "your-key"
}

📋 Contract Analysis

Extract key clauses, parties, dates, and obligations from legal documents.

{
  "pdfUrl": "https://example.com/contract.pdf",
  "action": "full_analysis",
  "googleApiKey": "your-key"
}

📚 Research Paper RAG

Chunk academic papers with semantic awareness for better retrieval.

{
  "pdfUrl": "https://example.com/paper.pdf",
  "action": "chunk_for_rag",
  "chunkSize": 500,
  "semanticChunking": true,
  "googleApiKey": "your-key"
}

🔍 Scanned Document OCR

Convert scanned PDFs to searchable text.

{
  "pdfUrl": "https://example.com/scanned.pdf",
  "action": "extract_text",
  "enableOcr": true,
  "googleApiKey": "your-key"
}

---

FAQ

---

Limitations

Limitation	Details
Max file size	50MB
Output truncation	Text: 100k chars, Chunks: 50 items (full data in dataset)
OCR requirement	Requires AI API key and embedded images in PDF
Rate limit	100 requests/minute per client
Memory	4GB recommended, up to 16GB for large documents

---

Error Codes

Code	Description	Solution
VALIDATION_ERROR	Invalid input	Check parameter types and values
INVALID_PDF	Corrupted PDF	Ensure PDF is valid and not encrypted
PROCESSING_ERROR	Runtime error	Retry the request
RESOURCE_LIMIT	File too large	Use smaller file or increase memory
RATE_LIMIT_EXCEEDED	Too many requests	Wait and retry

---

Technical Details

• Runtime: Node.js 22
• Memory: 4GB default, 16GB max
• PDF Libraries: pdf-parse, pdf-lib
• AI Models: Gemini 2.5 Flash, GPT-4V, Claude Vision
• Protocols: MCP (Model Context Protocol), REST API

---

Changelog

v3.0.0

• AI Document Analysis with executive summary, entities, and classification
• 7 specialized output views in Apify Console
• Memory-efficient streaming for 100+ page documents
• Gemini 2.5 Flash as default AI provider

v2.1.0

• AI-powered OCR with Vision APIs
• Semantic chunking with AI boundary detection
• Multi-provider AI support (OpenAI, Anthropic, Gemini)

v2.0.0

• Dual operation modes (One Click and BYOK)
• HTTP REST API for external clients
• Pay-per-event pricing model

---

Support

• Issues: Report bugs on GitHub
• Questions: Contact via Apify Console
• Documentation: This README and input schema tooltips

---

Built with ❤️ using Apify SDK