fix(application-form-fill): 清理填表说明并收窄按钮话术

review_agent/application_form_fill/services/field_extract.py

@@ -28,6 +28,15 @@ FIELD_ALIASES = {
     "storage_condition_and_validity": ["产品储存条件及有效期", "储存条件及有效期", "储存条件", "有效期"],
 }
 
+STATIC_STOP_LABELS = [
+    "申请人",
+    "国家药品监督管理局",
+    "填表说明",
+    "注",
+    "保证书",
+    "应附资料",
+]
+
 
 def collect_document_texts(summary_batch: FileSummaryBatch) -> dict[str, str]:
     texts: dict[str, str] = {}
@@ -180,7 +189,7 @@ def _field_aliases(field: dict[str, str]) -> list[str]:
 
 
 def _all_field_labels(fields: list[dict[str, str]]) -> list[str]:
-    labels: list[str] = []
+    labels: list[str] = list(STATIC_STOP_LABELS)
     for field in fields:
         for label in _field_aliases(field):
             if label not in labels:
@@ -194,7 +203,7 @@ def _extract_label_value(text: str, label: str, labels: list[str]) -> tuple[str,
 
 def _extract_colon_label_value(text: str, label: str, labels: list[str]) -> tuple[str, str]:
     escaped_labels = "|".join(re.escape(item) for item in labels if item != label)
-    stop_pattern = rf"(?=\n\s*(?:{escaped_labels})\s*[:：])" if escaped_labels else r"(?=\Z)"
+    stop_pattern = rf"(?=\n\s*(?:{escaped_labels})(?:\s*[:：]|\s*$))" if escaped_labels else r"(?=\Z)"
     pattern = re.compile(rf"{re.escape(label)}\s*[:：]\s*(.+?)(?:{stop_pattern}|\Z)", re.S)
     match = pattern.search(text or "")
     if not match:

review_agent/application_form_fill/services/word_fill.py

@@ -22,6 +22,7 @@ def fill_template(
     conflicts: list[dict] | None = None,
 ) -> Path:
     document = Document(str(template_path))
+    remove_fill_instructions(document)
     conflict_keys = {item.get("field_key") for item in conflicts or []}
     for field_config in spec.fields:
         target = field_config.get("target") or {}
@@ -43,6 +44,25 @@ def fill_template(
     return output
 
 
+def remove_fill_instructions(document: Document) -> None:
+    removing = False
+    for paragraph in list(document.paragraphs):
+        text = _normalize_label(paragraph.text)
+        if text == "填表说明":
+            removing = True
+        if removing:
+            _remove_paragraph(paragraph)
+            continue
+        if text.startswith("注填表前") and "填表说明" in text:
+            _remove_paragraph(paragraph)
+
+    for table in document.tables:
+        for row in list(table.rows):
+            row_text = _normalize_label("".join(cell.text for cell in row.cells))
+            if row_text == "填表说明" or row_text.startswith("注填表前"):
+                _remove_row(row)
+
+
 def fill_table_row(document: Document, row_label: str, value: str, *, conflict: bool = False) -> bool:
     normalized_label = _normalize_label(row_label)
     for table in document.tables:
@@ -71,6 +91,15 @@ def apply_cell_shading(cell, fill: str) -> None:
     shading.set(qn("w:fill"), fill)
 
 
+def _remove_paragraph(paragraph) -> None:
+    element = paragraph._element
+    element.getparent().remove(element)
+
+
+def _remove_row(row) -> None:
+    row._tr.getparent().remove(row._tr)
+
+
 def create_word_export(
     batch: ApplicationFormFillBatch,
     spec: TemplateSpec,