PO PDF Extraction (Cost Price Report)

Automated extraction of parts pricing data from PartsCheck Cost Price Report PDFs using Claude AI.

System: Invoice Extraction Server (Node.js + Claude API) Integration: Zoho Creator → Extraction Server → Zoho Creator Processing time: ~30-60 seconds per PDF

Overview

PartsCheck is a parts ordering system used in the Australian automotive industry. The extraction workflow processes Cost Price Report PDFs generated by PartsCheck, extracting quote details, vehicle information, supplier pricing, and line-item parts data.

What It Extracts

Header Fields:

Field	Description	Example
`quote_no`	Quote/order number	"16568"
`claim_no`	Insurance claim number	"CLM-2024-001"
`vehicle`	Make, Model, Year	"Honda, HR-V, 2019"
`estimator`	Name of estimator	"Izac"
`vin`	Vehicle Identification Number (17 chars)	"MRHRU5890LP060773"
`date`	Report date (dd/MM/yyyy)	"03/07/2026"
`registration_no`	Vehicle registration/plate	"EGC94K"
`requested_time`	Time requested	"9:30am"

Summary/Totals:

Field	Description
`subtotal`	Subtotal ex GST & freight
`freight`	Freight/shipping charge
`gst`	GST amount
`total`	Total including everything

Line Items (Parts Selected for Purchase):

Field	Description	Example
`part_description`	Part description (without part number)	"HINGE ASY FRT DR RH BG A"
`part_number`	Extracted part number	"F22800"
`supplier`	Supplier name	"Australian Automotive Group (City Ford)"
`part_type`	OEM, AFT, USED, etc.	"OEM"
`expiry`	Quote expiry date (dd/MM/yyyy)	"10/07/2026"
`comment`	Additional comments	"EXTRA"
`qty`	Quantity ordered	1
`cost_price`	Cost price ex GST	56.25
`sell_price`	Sell price ex GST	75.00

Process Flow

graph TD
    A[User uploads PartsCheck PDF] --> B[Zoho Creator webhook triggers]
    B --> C[Server: Download PDF]
    C --> D[Extract data with Claude AI]
    D --> E[Parse and validate JSON]
    E --> F[Clean registration number]
    F --> G[Send to Zoho API]
    G --> H[Zoho creates/updates record]
    H --> I[Complete]

Step-by-Step Process

1. File Upload & Trigger

Who: Parts team, Estimators Where: Zoho Creator → PartsCheck upload form What happens:

User uploads PartsCheck Cost Price Report PDF
Zoho Creator webhook fires with:
- Record ID (upload_id)
- File URL
- Environment (dev/production)

2. PDF Download

System: Invoice Extraction Server What happens:

Downloads PDF from Zoho Creator using OAuth token
Saves to temporary directory: temp/partscheck-{timestamp}/
Retry mechanism: 3 attempts with 3-second intervals

:::info Download Retry File downloads retry automatically up to 3 times with 3-second delays. Claude API and Zoho API calls do NOT retry automatically. :::

3. PDF Direct Extraction (No Image Conversion)

Model: Claude Opus 4 (Anthropic) Input: PDF file sent directly as base64-encoded document What happens:

PDF encoded to base64
Sent to Claude API using document type (not images)
Claude processes entire PDF as a single document
Returns structured JSON with all extracted fields

Advantages of direct PDF processing:

Faster than image conversion
Better text recognition
Preserves document structure
Handles multi-page PDFs seamlessly

4. Registration Number Cleaning

Special handling for registration field:

PartsCheck PDFs often include the file/claim number appended to the registration:

Registration No: EGC94K - 220265
                 ^^^^^^   ^^^^^^
                 Rego     File#

Cleaning logic:

If registration contains a dash followed by numbers, extract ONLY the part before the dash
Remove all trailing spaces and special characters
Example: "EGC94K - 220265" → "EGC94K"

:::caution Registration Number Format The extraction prompt specifically instructs Claude to extract only the vehicle registration, excluding any file/claim numbers. This was added after discovering registration numbers were being extracted with file numbers appended. :::

5. Part Number Extraction

Challenge: Part numbers are embedded in descriptions

PartsCheck descriptions include part numbers inline:

"HINGE ASY FRT DR RH BG F22800 A"
 └─ Description ──────┘ └─ Part# ─┘

Solution:

Claude extracts part number into separate part_number field
part_description contains ONLY the description text
Part numbers typically alphanumeric: F22800, BG72910-ABC

Examples:

Original	part_description	part_number
"HINGE ASY FRT DR RH BG F22800 A"	"HINGE ASY FRT DR RH BG A"	"F22800"
"BUMPER BAR REAR BG72910-ABC"	"BUMPER BAR REAR"	"BG72910-ABC"

6. Data Validation

Checks performed:

VIN exactly 17 characters
Date format: dd/MM/yyyy
Numeric values parsed correctly
Registration cleaned of file numbers
All line items extracted

7. Send to Zoho API

Endpoint: Custom Zoho Creator API Environment-aware:

Development: poc_extract_parts_details
Production: po_pdf_extract_details_prod

Payload:

{
  "upload_id": "1234567890",
  "status": "success",
  "extracted_data": "{...JSON string of all extracted fields...}"
}

No retry mechanism - single attempt

8. Response & Cleanup

What happens:

Zoho processes extracted data
Updates PartsCheck upload record
Server cleans up temporary PDF file
Returns success/failure status

Prompt Engineering

The PartsCheck extraction uses a specialized prompt that includes:

Date Handling

Always convert to dd/MM/yyyy format
Handle various date formats from PDF

Numeric Values

Remove currency symbols ($, AUD)
Parse as decimal numbers
Use 0 for missing numeric values

Registration Cleaning Rules

If registration includes dash + numbers:
  "EGC94K - 220265" → extract "EGC94K"

Remove trailing spaces and special characters

Part Number Extraction Rules

Part numbers are embedded in descriptions:
  "HINGE ASY FRT DR RH BG F22800 A"

Extract part number separately:
  part_description: "HINGE ASY FRT DR RH BG A"
  part_number: "F22800"

VIN Validation

Must be exactly 17 alphanumeric characters
Flag if incorrect length

Error Handling

Error Type	Handling
File download failure	Retry 3 times with 3s delay, then fail
Claude API error	Fail immediately, log error, no retry
JSON parsing error	Fail with parse error message
Zoho API error	Log error, no retry

Environment Configuration

Environment	Zoho API	API Key
Development	`poc_extract_parts_details`	`DK6ThON2OyGvTxFkg24phyQ8a`
Production	`po_pdf_extract_details_prod`	`1FCKNxWAhzN7yPB7gUnYZTe9O`

Common Data Issues

Issue: Registration with File Numbers

Problem:

Registration No: EGC94K - 220265

Solution: Prompt updated July 2026 to extract only EGC94K

Issue: Part Numbers in Description

Problem:

Description: "HINGE ASY FRT DR RH BG F22800 A"

Solution: Extraction splits into:

part_description: "HINGE ASY FRT DR RH BG A"
part_number: "F22800"

Issue: Missing Freight

Problem: Not all PartsCheck reports include freight charges

Solution: Use null for freight field if not present

Technical Details

Processing method: Direct PDF extraction (not image-based) Model: Claude Opus 4 (Anthropic) Max tokens: 16,000 Timeout: 30 seconds per PDF

Dependencies:

@anthropic-ai/sdk - Claude AI API
fs - File system operations
axios - HTTP client for Zoho API

Best Practices

For Parts Team

Use latest PartsCheck format - Extraction optimized for current format
Ensure registration visible - Check PDF preview before upload
Multi-page OK - System handles multi-page reports
Verify extractions - Review extracted data in Zoho

For IT/Developers

Monitor Claude API usage - Track token consumption
Check registration cleaning - Verify file numbers removed
Validate part number extraction - Ensure descriptions split correctly
Test both environments - Dev and production have different endpoints

Troubleshooting

Issue	Cause	Solution
Registration includes file number	Prompt not followed	Prompt updated July 2026, should be fixed
Part numbers in description	No splitting logic	Part number extraction implemented
VIN wrong length	OCR error, typo in PDF	Manual correction needed
Missing line items	Poor PDF quality	Request clearer export from PartsCheck
Freight always null	Not in PDF	Expected if PartsCheck doesn't show freight

Invoice Extraction - Bulk invoice processing
Parts Assessment Extraction - Parts & Sublet List
Reference: Order Parts Form - Where extracted parts go

For technical support or issues, contact the development team.

Overview​

What It Extracts​

Process Flow​

Step-by-Step Process​

1. File Upload & Trigger​

2. PDF Download​

3. PDF Direct Extraction (No Image Conversion)​

4. Registration Number Cleaning​

5. Part Number Extraction​

6. Data Validation​

7. Send to Zoho API​

8. Response & Cleanup​

Prompt Engineering​

Date Handling​

Numeric Values​

Registration Cleaning Rules​

Part Number Extraction Rules​

VIN Validation​

Error Handling​

Environment Configuration​

Common Data Issues​

Issue: Registration with File Numbers​

Issue: Part Numbers in Description​

Issue: Missing Freight​

Technical Details​

Best Practices​

For Parts Team​

For IT/Developers​

Troubleshooting​

Related Documentation​