PDF Extraction Quick Start Guide

Get pricing data from supplier catalogs automatically using Claude AI.

🚀 Quick Setup (5 minutes)

1. Get Anthropic API Key

# Visit https://console.anthropic.com/
# Sign up for free tier (includes $5 credit)
# Navigate to: Settings → API Keys → Create Key

2. Add to Environment

Create .env file in project root:

VITE_ANTHROPIC_API_KEY=sk-ant-api03-xxxxxxxxxxxxxxxxxxxxxxxxxxxxx

3. Restart Dev Server

npm run dev

That's it! PDF extraction is now active.

📤 How to Use (Shepherd View)

Step 1: Navigate to Service Chat

Log in as Shepherd
Click any service card (e.g., "Tires")
Chat interface opens

Step 2: Upload Supplier's PDF Catalog

Click 📎 button
Select supplier's PDF pricing catalog
Watch the magic happen:

You: 📎 Uploaded: tire_pricing_2025.pdf
AI: 🤖 Analyzing PDF catalog and extracting pricing data...
AI: ✅ Extraction Complete!

    Found 12 services from Cyber Tire & Wheel
    Price range: $120-$1800
    Confidence: high

    Cost: $0.0234

Step 3: View Extracted Data

Extracted data stored in conversation
Available for quote comparison
Can re-upload if extraction fails

📥 How to Use (Supplier View)

Step 1: Navigate to RFQ

Log in as Supplier
View open RFQs
Click RFQ to open chat

Step 2: Upload Your Catalog

Click 📎 button
Select your pricing PDF
AI extracts your services:

You: 📎 Uploaded: my_catalog_2025.pdf
AI: 🤖 Analyzing your PDF catalog and extracting pricing data...
AI: ✅ Extraction Complete!

    Found 8 services from your catalog
    Price range: $800-$1200
    Confidence: high

    💡 Ready to generate competitive quote for this RFQ!

Step 3: Send Quote

Use extracted data to respond to RFQ
AI can auto-generate quote message
Pricing data saved for future RFQs

📄 Supported File Formats

PDFs (Recommended)

✅ Text-based PDFs (best results)
✅ Scanned PDFs with OCR (good results)
✅ Multi-page catalogs
✅ Tables and structured pricing
⚠️ Image-heavy PDFs (variable results)
❌ Password-protected PDFs

CSVs (Alternative)

✅ Simple pricing spreadsheets
✅ Headers: name, price, description
✅ Faster processing than PDF
⚠️ Less metadata extracted

Size Limits

Max file size: 10MB
Max pages: ~50 pages (varies by complexity)
Average processing time: 5-10 seconds

💰 Cost Tracking

Pricing

Input tokens: $3 per million
Output tokens: $15 per million

Typical Costs

Document Type

Input Tokens

Output Tokens

Cost

1-page PDF

500

200

$0.005

5-page PDF

2000

500

$0.014

10-page PDF

4000

800

$0.024

20-page catalog

8000

1000

$0.039

Free Tier

Anthropic gives $5 credit for new accounts
That's ~200-500 PDF extractions
Perfect for testing and MVP

Monitoring Costs

Each extraction logs cost to browser console
Check Anthropic dashboard for usage
Set billing alerts at $10, $50, $100

🎯 What Gets Extracted

Service Information

{
  "name": "Tesla Model Y Tire Rotation",
  "description": "Rotation for 20\" Induction wheels",
  "pricePerUnit": 120.00,
  "unitType": "per service"
}

Optional Features

features: ["24/7 availability", "Mobile service"]
includesTPMS: true (tire pressure monitoring)
warrantyMonths: 12
responseTimeMinutes: 60
coverageZones: ["San Francisco", "Oakland"]

Volume Discounts

{
  "minVehicles": 50,
  "discountPercent": 15,
  "description": "15% off for fleets over 50 vehicles"
}

Contact Info

{
  "phone": "+1-555-1234",
  "email": "sales@supplier.com",
  "address": "123 Main St, San Francisco"
}

🔧 Troubleshooting

"Extraction failed: API key not found"

Fix: Add VITE_ANTHROPIC_API_KEY to .env file

"Extraction failed: Invalid PDF format"

Fix:

Ensure PDF is not password-protected
Try re-saving PDF from original source
Try converting to CSV instead

"Confidence: low" results

Reasons:

PDF is image-only (no text layer)
Unusual formatting/layout
Missing pricing information
Handwritten pricing

Fix:

Ask supplier for text-based PDF
Convert to CSV manually
Edit PDF to add text layer with OCR tool

High token costs

Reasons:

Very large multi-page catalogs
Repeated extractions of same file
Image-heavy PDFs

Fix:

Extract pricing pages only
Cache extraction results
Use CSV for simple price lists

🎓 Best Practices

For Shepherds

Ask suppliers for digital catalogs (not scanned)
Request CSV exports for simpler price lists
Upload once per supplier (results cached in conversation)
Verify extracted prices before finalizing contracts

For Suppliers

Provide clean, text-based PDFs
Include all pricing tiers in one document
Clearly label volume discounts
Update catalog regularly (upload new version when prices change)

General

Always review confidence score
Cross-check critical prices manually
Use extraction as starting point, not final source of truth
Report extraction errors to improve prompt engineering

🚨 Security Notes

Development Mode

API key in .env file (local only)
dangerouslyAllowBrowser: true enabled
⚠️ DO NOT COMMIT .env TO GIT

Production Mode (TODO)

Move API calls to backend
Create /api/extract-pdf endpoint
API key stored server-side only
Rate limiting per user
Cost monitoring and alerts

📊 Example Extraction

Input PDF

CYBER TIRE & WHEEL
2025 Fleet Pricing Catalog

SERVICES:
1. Tesla Model Y Tire Rotation - $120
   Includes: Rotation, pressure check, TPMS reset
   Warranty: 30 days

2. Full Tire Replacement (Michelin Pilot Sport EV) - $1,800/set
   Includes: 4 tires, mounting, balancing, disposal
   Warranty: 50,000 miles

3. Wheel Alignment - $250
   Laser alignment for autonomous vehicle spec
   Warranty: 6 months

VOLUME DISCOUNTS:
- 10-25 vehicles: 10% off
- 25-50 vehicles: 15% off
- 50+ vehicles: 20% off

Contact: sales@cybertire.com | (415) 555-TIRE

Output JSON

{
  "supplierName": "Cyber Tire & Wheel",
  "serviceCategory": "TIRES",
  "services": [
    {
      "name": "Tesla Model Y Tire Rotation",
      "description": "Includes rotation, pressure check, TPMS reset",
      "pricePerUnit": 120,
      "unitType": "per service",
      "features": ["Rotation", "Pressure check", "TPMS reset"],
      "includesTPMS": true,
      "warrantyMonths": 1
    },
    {
      "name": "Full Tire Replacement (Michelin Pilot Sport EV)",
      "description": "Includes 4 tires, mounting, balancing, disposal",
      "pricePerUnit": 1800,
      "unitType": "per set",
      "features": ["4 tires", "Mounting", "Balancing", "Disposal"],
      "warrantyMonths": null
    },
    {
      "name": "Wheel Alignment",
      "description": "Laser alignment for autonomous vehicle spec",
      "pricePerUnit": 250,
      "unitType": "per service",
      "features": ["Laser alignment", "Autonomous vehicle spec"],
      "warrantyMonths": 6
    }
  ],
  "volumeDiscounts": [
    { "minVehicles": 10, "discountPercent": 10 },
    { "minVehicles": 25, "discountPercent": 15 },
    { "minVehicles": 50, "discountPercent": 20 }
  ],
  "contactInfo": {
    "email": "sales@cybertire.com",
    "phone": "(415) 555-TIRE"
  },
  "extractionConfidence": "high"
}

Token Usage

Input: 1,234 tokens
Output: 456 tokens
Cost: $0.0105

🎉 Success Stories (Expected)

Scenario 1: Tire Services

Before: Shepherd calls 10 tire shops, manually records pricing After: 10 suppliers upload PDFs, AI extracts all pricing in 2 minutes Time saved: 3 hours → 2 minutes

Scenario 2: Fleet Parking

Before: Receive 5 email quotes, copy prices to spreadsheet After: Ask suppliers to upload catalogs in chat, instant comparison Time saved: 30 minutes → instant

Scenario 3: Multi-Service Comparison

Before: Manually build comparison table across 8 categories After: Suppliers upload master catalogs once, all services extracted Time saved: 8 hours → 5 minutes

📞 Support

Questions?

Check PHASE_2C_PDF_EXTRACTION_COMPLETE.md for technical details
Review services/PDFExtractionService.ts for implementation

Issues?

Low confidence extractions → Contact Claude to improve prompt
API errors → Check Anthropic status page
Cost concerns → Monitor usage in dashboard

Ready to extract pricing? Upload your first PDF and watch the magic! 🪄

PreviousBackend API Setup Guide NextT420 Agentic Workflow - Quick Reference Card

Last updated 8 days ago

Was this helpful?

hashtag🚀 Quick Setup (5 minutes)

hashtag1. Get Anthropic API Key

hashtag2. Add to Environment

hashtag3. Restart Dev Server

hashtag📤 How to Use (Shepherd View)

hashtagStep 1: Navigate to Service Chat

hashtagStep 2: Upload Supplier's PDF Catalog

hashtagStep 3: View Extracted Data

hashtag📥 How to Use (Supplier View)

hashtagStep 1: Navigate to RFQ

hashtagStep 2: Upload Your Catalog

hashtagStep 3: Send Quote

hashtag📄 Supported File Formats

hashtagPDFs (Recommended)

hashtagCSVs (Alternative)

hashtagSize Limits

hashtag💰 Cost Tracking

hashtagPricing

hashtagTypical Costs

hashtagFree Tier

hashtagMonitoring Costs

hashtag🎯 What Gets Extracted

hashtagService Information

hashtagOptional Features

hashtagVolume Discounts

hashtagContact Info

hashtag🔧 Troubleshooting

hashtag"Extraction failed: API key not found"

hashtag"Extraction failed: Invalid PDF format"

hashtag"Confidence: low" results

hashtagHigh token costs

hashtag🎓 Best Practices

hashtagFor Shepherds

hashtagFor Suppliers

hashtagGeneral

hashtag🚨 Security Notes

hashtagDevelopment Mode

hashtagProduction Mode (TODO)

hashtag📊 Example Extraction

hashtagInput PDF

hashtagOutput JSON

hashtagToken Usage

hashtag🎉 Success Stories (Expected)

hashtagScenario 1: Tire Services

hashtagScenario 2: Fleet Parking

hashtagScenario 3: Multi-Service Comparison

hashtag📞 Support

hashtagQuestions?

hashtagIssues?

🚀 Quick Setup (5 minutes)

1. Get Anthropic API Key

2. Add to Environment

3. Restart Dev Server

📤 How to Use (Shepherd View)

Step 1: Navigate to Service Chat

Step 2: Upload Supplier's PDF Catalog

Step 3: View Extracted Data

📥 How to Use (Supplier View)

Step 1: Navigate to RFQ

Step 2: Upload Your Catalog

Step 3: Send Quote

📄 Supported File Formats

PDFs (Recommended)

CSVs (Alternative)

Size Limits

💰 Cost Tracking

Pricing

Typical Costs

Free Tier

Monitoring Costs

🎯 What Gets Extracted

Service Information

Optional Features

Volume Discounts

Contact Info

🔧 Troubleshooting

"Extraction failed: API key not found"

"Extraction failed: Invalid PDF format"

"Confidence: low" results

High token costs

🎓 Best Practices

For Shepherds

For Suppliers

General

🚨 Security Notes

Development Mode

Production Mode (TODO)

📊 Example Extraction

Input PDF

Output JSON

Token Usage

🎉 Success Stories (Expected)

Scenario 1: Tire Services

Scenario 2: Fleet Parking

Scenario 3: Multi-Service Comparison

📞 Support

Questions?

Issues?