data-science-document-ai 2.1.0__tar.gz → 2.2.0__tar.gz

{data_science_document_ai-2.1.0 → data_science_document_ai-2.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: data-science-document-ai
-Version: 2.1.0
+Version: 2.2.0
 Summary: "Document AI repo for data science"
 Author: Naomi Nguyen
 Author-email: naomi.nguyen@forto.com

{data_science_document_ai-2.1.0 → data_science_document_ai-2.2.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "data-science-document-ai"
-version = "2.1.0"
+version = "2.2.0"
 description = "\"Document AI repo for data science\""
 authors = ["Naomi Nguyen <naomi.nguyen@forto.com>", "Kumar Rajendrababu <kumar.rajendrababu@forto.com>", "Igor Tonko <igor.tonko@forto.com>", "Osman Demirel <osman.demirel@forto.com>"]
 packages = [

{data_science_document_ai-2.1.0 → data_science_document_ai-2.2.0}/src/pdf_processing.py

@@ -261,6 +261,7 @@ async def process_file_w_llm(params, file_content, input_doc_type, llm_client):
         if use_chunking
         else [file_content]
     ):
+        logger.info(f"chunking the document into {number_of_pages} pages....")
         tasks.append(
             process_chunk_with_retry(
                 chunk, prompt, response_schema, llm_client, input_doc_type, retries=3

{data_science_document_ai-2.1.0 → data_science_document_ai-2.2.0}/src/postprocessing/common.py

@@ -727,6 +727,9 @@ async def format_all_entities(result, document_type_code, params, mime_type):
     # intact — the customs-pool filter there depends on it.
     if document_type_code in ["partnerInvoice", "customsInvoice", "bundeskasse"]:
         await process_partner_invoice(params, aggregated_data, document_type_code)
+        # Resolve the vendor to its legal-entity code so document-capture
+        # receives the code directly instead of a (possibly ambiguous) name.
+        await set_vendor_legal_entity_code(aggregated_data)
 
     if document_type_code in ["bookingConfirmation"]:
         aggregated_data["legalEntity"] = await get_legal_entity(
@@ -769,6 +772,44 @@ async def get_legal_entity(name, address=None):
     }
 
 
+async def set_vendor_legal_entity_code(aggregated_data):
+    """Overwrite the invoice vendor's ``formattedValue`` with its legal-entity code.
+
+    Invoices identify the vendor by a legal-entity code (e.g. ``LE33894``) so two
+    partners that share a name stay distinct downstream. The human-readable name
+    is preserved in ``documentValue`` (kept searchable in document-capture); only
+    ``formattedValue`` is replaced, and only when a code is resolved — otherwise
+    the name is left untouched so the field degrades gracefully.
+
+    Only the name is sent to the lookup: adding the address reduces mapping
+    accuracy.
+
+    Args:
+        aggregated_data (dict): The formatted extraction result. Mutated in place.
+    """
+    vendor = aggregated_data.get("vendorName")
+    if not isinstance(vendor, dict):
+        return
+
+    # formattedValue holds the canonical name post-formatting; fall back to the
+    # raw documentValue when it is missing.
+    formatted_name = vendor.get("formattedValue")
+    document_name = vendor.get("documentValue")
+    name = formatted_name or document_name
+    if not name:
+        return
+
+    legal_entity = await get_legal_entity(name)
+    code = legal_entity.get("formattedValue")
+    if code:
+        # Retain the human-readable name in documentValue before formattedValue
+        # is replaced by the code, so the original name is never lost (e.g. when
+        # only formattedValue was populated).
+        if not document_name:
+            vendor["documentValue"] = name
+        vendor["formattedValue"] = code
+
+
 def process_location_nodes(obj):
     """Add terminal and depot codes to the extracted data."""