You know the pain of expense reporting — not for you, for your users. The receipts piling up, the manual entry, the typos that slip through. With 4D AIKit, that grind is gone. A simple upload turns a paper receipt or invoice into clean, structured JSON, ready for your database.
No more wasted time on totals, dates, or vendor names. Vision AI reads it, language models structure it, and 4D ties it straight into your app. From paper to database in seconds — and your users never feel the friction.
With just a few lines of code, you can turn a photo of a paper receipt into a structured, ready-to-store JSON object — no manual entry, no human error.
Example: Extracting Structured Data from a Receipt
Imagine your user uploading a receipt through your application. With AIKit, the system can automatically detect:
- Document type (receipt, invoice, or purchase order)
- Date
- Vendor name
- Invoice number if available
- Currency
- Totals (including and excluding tax)
- Tax amounts
All this information is parsed into a clean JSON object, ready to be stored in your database.
Code Example:
Here’s what that looks like in practice:
var $client := cs.AIKit.OpenAI.new("YourAPIKey")
// CONVERT THE IMAGE TO BASE64 IF IT'S NOT A URL
var $blob := $pictureEntity.picture
var $base64Encoded : Text
BASE64 ENCODE($blob; $base64Encoded)
var $picture := "data:image/jpeg;base64," + $base64Encoded
var $prompt := "You are an expert assistant in processing accounting documents. Analyze the image of the attached document (receipt, invoice, or purchase order). Return only a JSON object, without any explanation, without markdown tags, without commentary, and without any line break before or after. The JSON must be clean and follow this exact structure: { \"document_type\": \"receipt\" ' \"invoice\" ' \"purchase_order\", \"date\": \"MM-DD-YYYY\", \"vendor\": \"Merchant or company name\", \"invoice_number\": \"text or null\", \"currency\": \"EUR\" ' \"USD\" ' \"other\", \"total_incl_tax\": float or null, \"total_excl_tax\": float or null, \"tax\": float or null, \"items\": [ { \"name\": \"string\", \"quantity\": number or null, \"unit_price\": float or null, \"total_price\": float or null } ] } Remember: no sentences, tags, or text should surround or precede the JSON. Just the pure JSON object, that’s all."
var $result := $client.chat.vision.create($picture).prompt($prompt).choice.message.content
var $parsedJSON : Object
$parsedJSON := JSON Parse($result)
return $parsedJSONBenefits: From Paper to Database in Seconds
This solution is ideal for:
- Employees submitting expense reports quickly and easily
- Finance teams aiming to reduce manual entry and errors
- Mobile apps offering document scanning and automation features
No more typing totals, dates, or vendor names — AIKit does the heavy lifting.
Why This Matters
This feature combines the best of three worlds:
- Vision AI: to read and interpret printed documents.
- Language Models: to intelligently extract structured data.
- 4D Integration: to seamlessly connect it all with your application logic and database.
Try It Yourself
All you need is:
- An image input (camera, scanner, or upload).
- A valid OpenAI API key.
- The 4D AIKit component in your project.
Then let the AI handle the rest : clean, structured, and production-ready.
Coming Next…
And this is just the beginning. We’ll be sharing more business-focused use cases with AIKit in upcoming posts. If you already have a scenario in mind, we’d love to hear it — it might even be featured in a future post.