feat: add concurrency safety, caption detection, admin enhancements, and performance improvements

tests/test_summarizer.py

@@ -366,6 +366,38 @@ class TestSummarizeOneFlow:
         result = await summarize_one(db_session, sample_paper)
         assert result["status"] == "skipped"
 
+    @pytest.mark.asyncio
+    async def test_post_processing_runs_in_thread(
+        self, db_session, sample_paper, mock_pi_output, _summarize_tmp_paths
+    ):
+        """后处理（图片提取/ChromaDB）在工作线程而非事件循环线程执行。"""
+        import threading
+
+        seen_threads: list[int] = []
+        main_thread = threading.current_thread().ident
+
+        def spy_extract(arxiv_id, schema):
+            seen_threads.append(threading.current_thread().ident)
+
+        with (
+            patch("app.services.summarizer.download_pdf", new_callable=AsyncMock),
+            patch(
+                "app.services.summary_generator.call_pi",
+                new_callable=AsyncMock,
+                return_value=(mock_pi_output, "test-session-id"),
+            ),
+            patch(
+                "app.services.summary_persister._maybe_extract_images",
+                side_effect=spy_extract,
+            ),
+            patch("app.services.summary_persister._maybe_index_chroma"),
+        ):
+            result = await summarize_one(db_session, sample_paper)
+
+        assert result["status"] == "done"
+        assert seen_threads, "后处理未被调用"
+        assert seen_threads[0] != main_thread, "后处理应在工作线程执行，不阻塞事件循环"
+
 
 # ═══════════════════════════════════════════════════════════════════════
 # 批量操作测试