Update app.py

app.py

@@ -97,8 +97,8 @@ def extract_pdf_text(file_data):
 # Function to summarize text using OpenAI GPT
 def extract_invoice_data(file_data, content_type, json_schema):
     """
-    Extracts data from a PDF (converted to images) or an image.
-    Only PDFs with 1 or 2 pages are allowed.
+    Extracts invoice data from PDFs (text-based) and images using OpenAI's GPT-4o-mini model.
+    Ensures accurate JSON schema binding.
     """
     system_prompt = """You are an expert in invoice data extraction. 
 Your task is to extract key fields from an invoice image. Ensure accurate extraction and return the data in JSON format.
@@ -111,21 +111,22 @@ Extract the following fields:
 2. Tax Amount (if available)
 3. Vendor GST (if available)
 4. Vendor Name
-5. Invoice Date (format: "DD-MMM-YYYY", e.g., "15-Sep-2023")
+5. Invoice Date (format: "DD-MMM-YYYY")
 6. Total Amount (numeric)
-7. Invoice Number (e.g., "INV-2023-001")
+7. Invoice Number (alpha-numeric)
 8. Vendor Address
-9. Invoice Currency (e.g., "USD", "EUR")
+9. Invoice Currency
 
 Ensure that:
 - All extracted fields match the invoice.
 - If any field is missing, return null instead of hallucinating data.
 - Do not generate synthetic values—only extract real information from the image.
 """
+    
     base64_images = []
     base64DataResp = []
     extracted_text = ""
-    
+
     if content_type == "application/pdf":
         try:
             extracted_text = extract_pdf_text(file_data)
@@ -151,14 +152,15 @@ Ensure that:
 
     elif content_type.startswith("image/"):
         # Handle direct image files
-        base64_pdf = base64.b64encode(file_data).decode('utf-8')
-        base64DataResp.append(f"data:{content_type};base64,{base64_pdf}")
-        
+        base64_img = base64.b64encode(file_data).decode('utf-8')
+        base64DataResp.append(f"data:{content_type};base64,{base64_img}")
+        base64_images.append(f"data:{content_type};base64,{base64_img}")
+
     else:
         return {"error": f"Unsupported file type: {content_type}"}
 
     if extracted_text:
-            return {"extracted_text": extracted_text}, base64DataResp
+        return {"extracted_text": extracted_text}, base64DataResp
     
     # Prepare OpenAI request
     openai_content = [{"type": "image_url", "image_url": {"url": img_base64}} for img_base64 in base64_images]