Fix official policy selection and persistent attachment gaps

Read the OFAC action body and use focused macro policy phrases, including national budgets, appropriations and reserve policies. Keep unknown image text as an explicit gap instead of treating it as complete or irrelevant.\n\nPreserve meeting attachment retries and derive pending/exhausted document counts from saved source cursors. Cover real selection negatives, positive macro boundaries and attachment failures with isolated regressions.

backend/src/alphapulse/crawler/spiders/us_official.py

@@ -316,7 +316,7 @@ class USOfficialSpider(NewsSpider):
                 for result in self.complete(lane):
                     yield result
                 return
-            parsed = {'title': data['title'], 'content_html': '<pre>' + escape(json.dumps(data, ensure_ascii=False, indent=2)) + '</pre>', 'published_at': at, 'url': response.url, 'external_id': identity(response.url), 'raw_response_id': response.meta['raw_response_id'],
+            parsed = {'title': data['title'], 'content_html': '<pre>' + escape(json.dumps(data, ensure_ascii=False, indent=2)) + '</pre>', 'published_at': at, 'url': response.url, 'external_id': identity(response.url), 'raw_response_id': response.meta['raw_response_id'], 'retry_document': doc,
                       'extra': {'content_scope': 'article', 'published_at_precision': precision, 'published_at_basis': 'official_update', 'published_date': at.date().isoformat(), 'official_meeting': data, 'rule_version': RULE_VERSION, 'attachments': [{'url': official_url(self.name, d['url']), 'title': d.get('name') or d.get('documentType', ''), 'status': 'pending'} for d in {d['url']: d for d in data.get('meetingDocuments', []) + data.get('witnessDocuments', []) if d.get('format') == 'PDF'}.values()]}}
             for result in self.attach_or_finish(parsed, lane):
                 yield result
@@ -514,7 +514,7 @@ class USOfficialSpider(NewsSpider):
             retries[parsed['url']] = {'document': retry_doc, 'attempts': previous.get('attempts', 0) + 1, 'next_at': (self.run.started_at + timedelta(hours=1)).timestamp()}
         elif not incomplete:
             retries.pop(parsed['url'], None)
-        if reasons or incomplete:
+        if reasons:
             parsed['extra']['selection_reasons'] = reasons
             parsed['extra']['content_scope'] = 'partial' if incomplete else 'article'
             self.run.stats['partial'] = self.run.stats.get('partial', 0) + int(incomplete)
@@ -523,6 +523,8 @@ class USOfficialSpider(NewsSpider):
             yield item
         else:
             self.run.stats['excluded'] = self.run.stats.get('excluded', 0) + 1
+            if incomplete:
+                self.run.stats['unclassified_attachment_gaps'] = self.run.stats.get('unclassified_attachment_gaps', 0) + 1
         yield from self.complete(lane)
 
     def parse_attachment(self, response):
@@ -532,6 +534,8 @@ class USOfficialSpider(NewsSpider):
         try:
             if response.status != 200:
                 raise ValueError(f'HTTP {response.status}')
+            if response.headers.get('Content-Type', b'').lower().startswith(b'image/') and response.body.startswith((b'\x89PNG\r\n\x1a\n', b'\xff\xd8\xff', b'GIF87a', b'GIF89a')):
+                raise ValueError('官方图片附件尚未提取文字，请核对原文')
             text = pdf_text(response.body)
             attachment['status'] = 'complete'
             parsed['content_html'] += f"<section><h2>{escape(attachment['title'])}</h2><pre>{escape(text)}</pre></section>"