Build a Receipt Expense Pipeline

Turn a folder of receipts into a deduped expense report with merchant, date, and total.

OCR every receipt to JSON with ARVA OCR
Parse merchant, date, and total with Python patterns
Dedupe repeat scans by file hash

Time to complete: ~15 minutes

Prerequisites

A SAGEA_API_KEY exported as an environment variable
Receipt images in PDF, PNG, JPG, TIFF, or WEBP format, max 100MB, ideally 300 DPI
Python 3.9+ with the requests package installed
Set up document processing
ARVA OCR model card

Step 1: OCR each receipt to JSON

Loop over a folder and post every file to the OCR endpoint.

curl -X POST https://api.sagea.space/v1/ocr/process \
  -H "Authorization: Bearer $SAGEA_API_KEY" \
  -F model="arva-ocr" \
  -F [email protected] \
  -F output_format="json"

Each response contains text, pages, confidence, blocks, and model_used. Accuracy is 99.2% on clean scans across 150+ languages.

Step 2: Parse merchant, date, and total

Extract the three expense fields from OCR text with regular expressions.

curl -X POST https://api.sagea.space/v1/ocr/process \
  -H "Authorization: Bearer $SAGEA_API_KEY" \
  -F model="arva-ocr" \
  -F [email protected] \
  -F output_format="text"

Step 3: Dedupe by hash and export CSV

Hash file bytes to drop duplicate scans, then write the expense report.

import csv
import hashlib
 
seen, rows = set(), []
for path in receipts:
    digest = hashlib.sha256(open(path, "rb").read()).hexdigest()
    if digest in seen:
        print(f"skip duplicate: {path}")
        continue
    seen.add(digest)
    data = ocr_receipt(path)
    parsed = parse_receipt(data["text"])
    rows.append({
        "file": path,
        "hash": digest[:12],
        "confidence": data["confidence"],
        **parsed,
    })
 
with open("expenses.csv", "w", newline="") as f:
    writer = csv.DictWriter(f, fieldnames=["file", "hash", "merchant", "date", "total", "confidence"])
    writer.writeheader()
    writer.writerows(rows)
 
print(f"wrote {len(rows)} unique expenses to expenses.csv")

Standard pages cost $1.00 per 1000 pages, so a thousand receipts cost about one dollar.

Verify

Error	Cause	Fix
`401 Unauthorized`	Missing or invalid API key	Export a valid key as `SAGEA_API_KEY` and retry
`400 unsupported_format`	File type not in PDF, PNG, JPG, TIFF, WEBP	Convert the receipt to JPG or PDF and retry
`413 Payload Too Large`	File is over 100MB	Compress the image and retry
`429 rate_limit_exceeded`	Too many requests at once	Add a short sleep between receipts and retry