Hamza/musadeq
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

🚀 Multi-invoice extraction and detailed validation feedback

Browse Source
This commit is contained in:
Hamza-Ayed
2026-04-19 16:13:09 +03:00
parent 6b9ce6e95b
commit 2e2d76c0a8
5 changed files with 121 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

73
backend/src/modules/invoices/gemini-extractor.service.ts
View File

@@ -28,7 +28,7 @@ export class GeminiExtractorService {
}
/**
* استخراج البيانات من صورة الفاتورة
* استخراج البيانات من صورة الفاتورة (يدعم فواتير متعددة في ملف واحد)
*/
async extractInvoiceData(filePath: string, storageRoot: string): Promise<any> {
try {
@@ -36,40 +36,50 @@ export class GeminiExtractorService {
const fileData = fs.readFileSync(fullPath);
const prompt = `
You are a Jordanian tax expert. Extract all details from this invoice image for the Jordan National Invoicing System (JoFotara / JoInvoice).
The output MUST be a strict JSON object following this schema:
You are a Jordanian tax expert. Extract all details from this file (image or PDF).
The file may contain ONE or MULTIPLE distinct invoices.
Extract EACH invoice separately.
The output MUST be a strict JSON object with this schema:
{
"invoice_number": "string",
"invoice_date": "YYYY-MM-DD",
"invoice_type": "cash" | "credit",
"supplier_name": "string",
"supplier_tin": "string (10 digits)",
"buyer_name": "string (optional)",
"buyer_tin": "string (optional)",
"subtotal": number (before discount and tax),
"discount_total": number (total discount),
"tax_amount": number (total tax),
"grand_total": number (final amount),
"currency_code": "JOD",
"lines": [
"invoices": [
{
"line_number": number,
"description": "string",
"quantity": number,
"unit_price": number,
"discount": number,
"tax_rate": number (e.g. 0.16 for 16%),
"line_total": number (quantity * unit_price - discount)
"invoice_number": "string",
"invoice_date": "YYYY-MM-DD",
"invoice_type": "cash" | "credit",
"invoice_category": "standard" | "simplified",
"supplier_name": "string",
"supplier_tin": "string (10 digits)",
"buyer_name": "string (optional)",
"buyer_tin": "string (optional)",
"subtotal": number (before discount and tax),
"discount_total": number (total discount),
"tax_amount": number (total tax),
"grand_total": number (final amount),
"currency_code": "JOD",
"lines": [
{
"line_number": number,
"description": "string",
"quantity": number,
"unit_price": number,
"discount": number,
"tax_rate": number (e.g. 0.16 for 16%),
"line_total": number (quantity * unit_price - discount)
}
]
}
]
}
Pay close attention to Jordanian Tax Rules:
- Standard Rate: 0.16 (16%)
- Reduced Rate: 0.04 (4%)
- Zero Rate: 0.00 (0%)
- Exempt: null or 0 with explanation.
JoFotara Specific Rules:
- invoice_category: "simplified" if the buyer is a regular person (no TIN), "standard" if B2B (buyer has TIN).
- Prices in Jordan (JOD) often have 3 decimal places (e.g. 2.800).
- Standard VAT Rate: 0.16 (16%)
- The formula MUST hold: Subtotal - Discount + Tax = Grand Total.
Return ONLY the JSON. No markdown formatting. If a field is missing, return null.
- If multiple invoices are in the PDF, ensure the "invoices" array contains all of them.
Return ONLY the JSON. No markdown formatting.
`;
// Detect MIME type based on extension
@@ -93,9 +103,10 @@ export class GeminiExtractorService {
// Clean up markdown if any
const cleanedJson = responseText.replace(/```json|```/g, '').trim();
return JSON.parse(cleanedJson);
const data = JSON.parse(cleanedJson);
return data.invoices || [];
} catch (error: any) {
this.logger.error(`AI Extraction failed: ${error.message}`);
this.logger.error(\`AI Extraction failed: \${error.message}\`);
throw new InternalServerErrorException('AI Extraction failed');
}
}