Fix date range processing: extract first range only, add JSON format instructions, increase max_tokens

src/pipelines/saas/prompts/prompt_calls.py

@@ -899,7 +899,7 @@ def prompt_split_reimb_dates(date_range: str, filename: str) -> dict:
         prompt,
         model_id="sonnet_latest",
         filename=filename,
-        max_tokens=200,
+        max_tokens=500,
         cache=True,
         instruction=prompt_templates.SPLIT_REIMB_DATES_INSTRUCTION(),
         usage_label="SPLIT_REIMB_DATES",

src/pipelines/shared/extraction/one_to_n_funcs.py

@@ -679,6 +679,13 @@ def split_reimb_dates(one_to_n_results: list, filename: str) -> list:
             llm_answer_final = prompt_calls.prompt_split_reimb_dates(
                 date_range, filename
             )
+            # Debug: print raw parsed LLM output
+            print(f"DEBUG split_reimb_dates - Index {index}:")
+            print(f"  Input date_range: {date_range}")
+            print(f"  Parsed LLM output: {llm_answer_final} (type: {type(llm_answer_final)})")
+            if isinstance(llm_answer_final, dict):
+                print(f"  start_date: {llm_answer_final.get('start_date')}")
+                print(f"  end_date: {llm_answer_final.get('end_date')}")
             # Update the specific record with parsed dates
             if llm_answer_final and isinstance(llm_answer_final, dict):
                 if "start_date" in llm_answer_final:

src/prompts/prompt_templates.py

@@ -2010,7 +2010,7 @@ def EFFECTIVE_DATE_FIX_PROMPT() -> Tuple[str, Callable[[str], dict]]:
 
 def SPLIT_REIMB_DATES_INSTRUCTION() -> str:
     """Static instruction for SPLIT_REIMB_DATES prompt caching."""
-    return """[OBJECTIVE]
+    return f"""[OBJECTIVE]
 Extract start and end dates from a date range string and return them in YYYY/MM/DD format.
 
 [RULES]
@@ -2034,17 +2034,23 @@ Extract start and end dates from a date range string and return them in YYYY/MM/
 5. If only one date is present:
    - Assume it is the start_date and set end_date as "N/A"
 
-6. If no extractable dates are found in the text:
+6. If multiple date ranges are present (separated by commas, e.g., "Effective 5/1/2018 - 4/30/2019, Effective 5/1/2019 - 4/30/2020"):
+   - Extract the FIRST (earliest) date range only
+   - Use the start_date from the first range and end_date from the first range
+   - Ignore all subsequent date ranges
+
+7. If no extractable dates are found in the text:
    - Return "N/A" for both start_date and end_date
 
 [OUTPUT FORMAT]
-Return the dates in the following JSON format:
-{
+Return ONLY the JSON dictionary with no explanatory text. Use the following format:
+{{
     "start_date": "YYYY/MM/DD",
     "end_date": "YYYY/MM/DD"
-}
+}}
 
-Note: Use "N/A" for dates that cannot be determined or extracted from the input text."""
+Note: Use "N/A" for dates that cannot be determined or extracted from the input text.
+{JSON_DICT_FORMAT_INSTRUCTIONS}"""
 
 
 def SPLIT_REIMB_DATES(date_range: str) -> Tuple[str, Callable[[str], dict]]: