Fix CLAIM_TYPE_CD being converted to list in get_crosswalk_fields and add debug prints for troubleshooting

src/pipelines/saas/file_processing.py

@@ -522,6 +522,11 @@ def run_one_to_n_prompts(
     )
 
     ################## CONVERT TO DF ##################
+    # Debug: print CLAIM_TYPE_CD before DataFrame creation
+    if one_to_n_results:
+        for i, row in enumerate(one_to_n_results[:3]):  # Check first 3 rows
+            if "CLAIM_TYPE_CD" in row:
+                print(f"CLAIM_TYPE_CD before DataFrame (row {i}): {row['CLAIM_TYPE_CD']} (type: {type(row['CLAIM_TYPE_CD'])})")
     one_to_n_df = pd.DataFrame(one_to_n_results)
 
     return one_to_n_df, dynamic_one_to_one_fields, first_reimbursement_page

src/pipelines/saas/prompts/prompt_calls.py

@@ -37,21 +37,6 @@ def prompt_exhibit_level(
 
     llm_answer_final = _parser(llm_answer_raw)
 
-    # Normalize CLAIM_TYPE_CD if present: convert single-element lists to strings
-    # This field should be a single value even though it's a 1:N field
-    if "CLAIM_TYPE_CD" in llm_answer_final:
-        original_value = llm_answer_final["CLAIM_TYPE_CD"]
-        print(
-            f"CLAIM_TYPE_CD before normalization: {original_value} (type: {type(original_value)})"
-        )
-        normalized_value = string_utils.normalize_one_to_one_field_value(
-            "CLAIM_TYPE_CD", original_value
-        )
-        print(
-            f"CLAIM_TYPE_CD after normalization: {normalized_value} (type: {type(normalized_value)})"
-        )
-        llm_answer_final["CLAIM_TYPE_CD"] = normalized_value
-
     return llm_answer_final
 
 

src/pipelines/shared/extraction/one_to_n_funcs.py

@@ -769,18 +769,33 @@ def get_lob_relationship(answer_dicts, exhibit_text, filename):
 def one_to_n_cleaning(
     all_exhibit_rows: list[dict], exhibit_text: str, constants: Constants, filename: str
 ):
+    # Debug: print CLAIM_TYPE_CD at start of cleaning
+    if all_exhibit_rows and "CLAIM_TYPE_CD" in all_exhibit_rows[0]:
+        print(f"CLAIM_TYPE_CD at start of one_to_n_cleaning: {all_exhibit_rows[0]['CLAIM_TYPE_CD']} (type: {type(all_exhibit_rows[0]['CLAIM_TYPE_CD'])})")
 
     ################################ Crosswalk Fields ################################
     all_exhibit_rows = aarete_derived.get_crosswalk_fields(all_exhibit_rows, constants)
+    # Debug: print CLAIM_TYPE_CD after crosswalk
+    if all_exhibit_rows and "CLAIM_TYPE_CD" in all_exhibit_rows[0]:
+        print(f"CLAIM_TYPE_CD after get_crosswalk_fields: {all_exhibit_rows[0]['CLAIM_TYPE_CD']} (type: {type(all_exhibit_rows[0]['CLAIM_TYPE_CD'])})")
 
     ################################ Determine LOB Relationship ################################
     all_exhibit_rows = get_lob_relationship(all_exhibit_rows, exhibit_text, filename)
+    # Debug: print CLAIM_TYPE_CD after LOB relationship
+    if all_exhibit_rows and "CLAIM_TYPE_CD" in all_exhibit_rows[0]:
+        print(f"CLAIM_TYPE_CD after get_lob_relationship: {all_exhibit_rows[0]['CLAIM_TYPE_CD']} (type: {type(all_exhibit_rows[0]['CLAIM_TYPE_CD'])})")
 
     ################################ Fill NA Mapping ################################
     all_exhibit_rows = aarete_derived.fill_na_mapping(all_exhibit_rows)
+    # Debug: print CLAIM_TYPE_CD after fill_na_mapping
+    if all_exhibit_rows and "CLAIM_TYPE_CD" in all_exhibit_rows[0]:
+        print(f"CLAIM_TYPE_CD after fill_na_mapping: {all_exhibit_rows[0]['CLAIM_TYPE_CD']} (type: {type(all_exhibit_rows[0]['CLAIM_TYPE_CD'])})")
 
     ################################ Split REIMB_DATES ################################
     all_exhibit_rows = split_reimb_dates(all_exhibit_rows, filename)
+    # Debug: print CLAIM_TYPE_CD after split_reimb_dates
+    if all_exhibit_rows and "CLAIM_TYPE_CD" in all_exhibit_rows[0]:
+        print(f"CLAIM_TYPE_CD after split_reimb_dates: {all_exhibit_rows[0]['CLAIM_TYPE_CD']} (type: {type(all_exhibit_rows[0]['CLAIM_TYPE_CD'])})")
 
     ################################ Normalize Single-Value Fields to Strings ################################
     # REIMB_TERM, SERVICE_TERM, and LOB_PROGRAM_RELATIONSHIP should always be strings (not lists)
@@ -811,6 +826,10 @@ def one_to_n_cleaning(
             else:
                 answer_dict["LOB_PRODUCT_RELATIONSHIP"] = "Exclusive"
 
+    # Debug: print CLAIM_TYPE_CD at end of one_to_n_cleaning
+    if all_exhibit_rows and "CLAIM_TYPE_CD" in all_exhibit_rows[0]:
+        print(f"CLAIM_TYPE_CD at end of one_to_n_cleaning: {all_exhibit_rows[0]['CLAIM_TYPE_CD']} (type: {type(all_exhibit_rows[0]['CLAIM_TYPE_CD'])})")
+
     return all_exhibit_rows
 
 

src/pipelines/shared/extraction/row_funcs.py

@@ -29,6 +29,9 @@ def combine_one_to_n(
             if key not in combined_dict or string_utils.is_empty(
                 combined_dict.get(key)
             ):
+                # Debug: print CLAIM_TYPE_CD when it's being added from exhibit_level
+                if key == "CLAIM_TYPE_CD":
+                    print(f"CLAIM_TYPE_CD in combine_one_to_n (from exhibit_level): {value} (type: {type(value)})")
                 combined_dict[key] = value
         # Add exhibit text for contextual analysis (e.g., for bill type code determination)
         combined_dict["EXHIBIT_TEXT"] = exhibit_text

src/pipelines/shared/extraction/tin_npi_funcs.py

@@ -484,6 +484,11 @@ def get_provider_info(
     # Defensive programming for any json parsing errors
     try:
         providers = _parser(llm_answer_raw)
+        # Print raw LLM provider NAME values to debug format issues
+        print(f"Raw LLM provider info (parsed) on page {page_num}: {providers}")
+        for i, provider in enumerate(providers):
+            if "NAME" in provider:
+                print(f"  Provider {i} NAME: {provider['NAME']} (type: {type(provider['NAME'])})")
         logging.debug(f"Parsed provider info on page {page_num}: {providers}")
         # Ensure we have a list of providers (not a dict or other type)
         if isinstance(providers, dict):

src/pipelines/shared/postprocessing/aarete_derived.py

@@ -239,8 +239,9 @@ def get_crosswalk_fields(answer_dicts: list, constants: Constants):
                     else:
                         answer_dict[to_field_name] = to_field_answer_list
                     
-                    # update from field value to list if not already
-                    if not isinstance(from_field_value, list):
-                        answer_dict[from_field_name] = from_field_value_list
+                    # DO NOT update from_field_name - preserve its original format
+                    # The crosswalk should only create/update the target field (to_field_name),
+                    # not modify the source field (from_field_name)
+                    # If from_field_name needs to be a list, that should be handled elsewhere
 
     return answer_dicts