allowing multiple summaries per bill text

pipelines/jobs/extract_bill_topics.py

@@ -19,6 +19,7 @@ from pipelines.orm.common import get_postgres_engine
 from pipelines.orm.data_science_dev.congress import (
     Bill,
     BillText,
+    BillTextSummary,
     BillTopic,
     BillTopicPosition,
     SubjectType,
@@ -72,11 +73,19 @@ class ExtractedBillTopic:
 def _select_bill_text_for_topic_extraction(bill: Bill) -> BillText | None:
     """Pick one summarized bill_text row from the already-loaded relationship."""
     for bill_text in bill.bill_texts:
-        if bill_text.summary and bill_text.summary.strip():
+        summary_row = bill_text.default_summary()
+        if summary_row and summary_row.summary.strip():
             return bill_text
     return None
 
 
+def _bill_text_has_summary_clause() -> ColumnElement[bool]:
+    """Return a correlated EXISTS clause for bill texts with at least one summary."""
+    return exists(
+        select(BillTextSummary.id).where(BillTextSummary.bill_text_id == BillText.id)
+    )
+
+
 def normalize_topic_label(value: str) -> str:
     """Normalize a topic label for storage, comparison, and de-duping."""
     normalized = value.strip().strip("\"'")
@@ -323,11 +332,7 @@ def create_select_bills_for_topic_extraction(
     limit: int | None = None,
 ) -> Select[tuple[Bill]]:
     """Select bill rows that have summarized bill_text rows for topic extraction."""
-    has_summary = (BillText.summary.is_not(None), BillText.summary != "")
-    summarized_text_filters: list[ColumnElement[bool]] = [
-        BillText.bill_id == Bill.id,
-        *has_summary,
-    ]
+    summarized_text_filters: list[ColumnElement[bool]] = [_bill_text_has_summary_clause()]
     if with_votes_only:
         summarized_text_filters.append(
             exists(
@@ -347,11 +352,17 @@ def create_select_bills_for_topic_extraction(
                 )
             )
         )
-    summarized_text_exists = exists(select(BillText.id).where(*summarized_text_filters))
+    summarized_text_exists = exists(
+        select(BillText.id).where(BillText.bill_id == Bill.id, *summarized_text_filters)
+    )
+    bill_text_loader = selectinload(Bill.bill_texts.and_(*summarized_text_filters))
     stmt = (
         select(Bill)
         .where(summarized_text_exists)
-        .options(selectinload(Bill.bill_texts.and_(*summarized_text_filters[1:])))
+        .options(
+            bill_text_loader.selectinload(BillText.summaries),
+            bill_text_loader.selectinload(BillText.primary_summary),
+        )
         .order_by(Bill.id)
     )
     if congress is not None:
@@ -363,7 +374,7 @@ def create_select_bills_for_topic_extraction(
             select(BillText.id).where(
                 BillText.bill_id == Bill.id,
                 BillText.id.in_(bill_text_ids),
-                *summarized_text_filters[1:],
+                *summarized_text_filters,
             )
         )
         stmt = stmt.where(selected_text_exists)
@@ -416,8 +427,7 @@ def collect_topic_extraction_diagnostics(
             )
         )
 
-    has_summary = (BillText.summary.is_not(None), BillText.summary != "")
-    summary_filters = [*bill_text_filters, *has_summary]
+    summary_filters = [*bill_text_filters, _bill_text_has_summary_clause()]
 
     bills_with_summaries = session.scalar(
         select(func.count(func.distinct(Bill.id)))
@@ -607,7 +617,11 @@ def main(
             if bill_text is None:
                 logger.warning("Skipping bill id=%s: no usable summary", bill.id)
                 continue
-            summary = bill_text.summary.strip()
+            summary_row = bill_text.default_summary()
+            if summary_row is None:
+                logger.warning("Skipping bill id=%s: no default summary", bill.id)
+                continue
+            summary = summary_row.summary.strip()
 
             try:
                 extracted_topics = extract_topics_for_bill_text(