feat(api): 叙事 prompt、职业上下文、读路径章节、WS 解耦与错误脱敏

- 回忆录：事实边界补充允许清单；传记文体示例与 JSON 叙事要求对齐
- default 职业提示 occupation_context；cadre/military 退休语境
- GET 章节读路径零写入，prepare_chapter_read_view + markdown_for_response
- 文本归一抽到 core/text_normalize；移除弃用 reply 策略与 recompose_chapters_for_story
- ConversationService：WS 连接/用户段落/结束对话；对外错误固定文案
- 测试：HTTP 脱敏契约、章节读视图、occupation 与 background_voice

api/app/features/memoir/service.py

@@ -21,11 +21,7 @@ from app.features.memoir.chapter_markdown_compose import (
     materialize_chapter_markdown_from_loaded_chapter,
 )
 from app.features.memoir.cover_eligibility import primary_chapter_memoir_image
-from app.features.memoir.helpers import (
-    chapter_to_dict,
-    chapter_to_list_dict,
-    is_image_permanently_unavailable,
-)
+from app.features.memoir.helpers import chapter_to_dict, chapter_to_list_dict
 from app.features.memoir.memoir_images.settings import MemoirImageSettings
 from app.features.memoir.models import Book, Chapter, ChapterStoryLink
 from app.features.memoir.reading_segment_materialize import (
@@ -37,6 +33,29 @@ from app.ports.storage import ObjectStorage
 logger = get_logger(__name__)
 
 
+def prepare_chapter_read_view(chapter: Chapter) -> tuple[Chapter, str | None]:
+    """
+    读路径：不写入 DB。返回 (chapter, markdown_for_response)。
+    None 表示序列化使用 ORM 上的 canonical_markdown。
+    """
+    has_story_links = bool(getattr(chapter, "story_links", None))
+    has_snapshot = chapter.reading_segments_json is not None
+    dirty = chapter.markdown_compose_dirty is True
+
+    if not dirty:
+        return chapter, None
+
+    if has_snapshot:
+        return chapter, None
+
+    if has_story_links:
+        md = materialize_chapter_markdown_from_loaded_chapter(chapter)
+        m = (md or "").strip()
+        return chapter, m if m else None
+
+    return chapter, None
+
+
 async def get_or_create_book(user_id: str, db: AsyncSession):
     """Get the user's current book or return None."""
     return await repo.get_current_book(user_id, db)
@@ -67,29 +86,6 @@ class MemoirService:
         bundle = await self._memory.retrieve(user_id, query, top_k=top_k)
         return bundle.model_dump()
 
-    async def _cleanup_unavailable_images(self, ch: Chapter) -> None:
-        cleaned = False
-        for rec in getattr(ch, "images", None) or []:
-            if rec and is_image_permanently_unavailable(rec):
-                logger.info("清理不可用配图: chapter={}, image={}", ch.id, rec.id)
-                await self._db.delete(rec)
-                cleaned = True
-        if cleaned:
-            await self._db.commit()
-            await self._db.refresh(ch)
-
-    async def _ensure_chapter_materialized(self, chapter: Chapter) -> Chapter:
-        has_story_links = bool(getattr(chapter, "story_links", None))
-        has_snapshot = chapter.reading_segments_json is not None
-        if not has_story_links or (has_snapshot and not chapter.markdown_compose_dirty):
-            return chapter
-
-        markdown = materialize_chapter_markdown_from_loaded_chapter(chapter)
-        await repo.append_chapter_compose_version_async(self._db, chapter, markdown)
-        await self._db.commit()
-        refreshed = await repo.get_chapter_by_id(chapter.id, self._db)
-        return refreshed or chapter
-
     async def get_current_book(self, user_id: str) -> dict:
         book = await repo.get_current_book(user_id, self._db)
         if not book:
@@ -152,8 +148,10 @@ class MemoirService:
         chapters_raw = list(result.unique().scalars().all())
         chapters: List[Chapter] = []
         for ch in chapters_raw:
-            ch2 = await self._ensure_chapter_materialized(ch)
-            if chapter_meets_minimum_display(ch2):
+            ch2, md_override = prepare_chapter_read_view(ch)
+            if chapter_meets_minimum_display(
+                ch2, canonical_markdown_override=md_override
+            ):
                 chapters.append(ch2)
         asset_ids = collect_asset_ids_for_chapters(chapters)
         asset_map = await signed_urls_for_asset_ids(self._db, asset_ids)
@@ -180,11 +178,18 @@ class MemoirService:
         asset_map = await signed_urls_for_asset_ids(self._db, asset_ids)
         all_chapters: List[dict] = []
         for ch in chapters:
-            ch = await self._ensure_chapter_materialized(ch)
-            if not chapter_meets_minimum_display(ch):
+            ch, md_override = prepare_chapter_read_view(ch)
+            if not chapter_meets_minimum_display(
+                ch, canonical_markdown_override=md_override
+            ):
                 continue
-            await self._cleanup_unavailable_images(ch)
-            all_chapters.append(chapter_to_list_dict(ch, asset_url_map=asset_map))
+            all_chapters.append(
+                chapter_to_list_dict(
+                    ch,
+                    asset_url_map=asset_map,
+                    markdown_for_response=md_override,
+                )
+            )
         return all_chapters
 
     async def get_chapter(self, chapter_id: str, user_id: str) -> dict:
@@ -195,14 +200,19 @@ class MemoirService:
             raise HTTPException(status_code=403, detail="无权访问此章节")
         if not chapter.is_active:
             raise HTTPException(status_code=404, detail="Chapter not found")
-        chapter = await self._ensure_chapter_materialized(chapter)
-        await self._cleanup_unavailable_images(chapter)
-        if not chapter_meets_minimum_display(chapter):
+        chapter, md_override = prepare_chapter_read_view(chapter)
+        if not chapter_meets_minimum_display(
+            chapter, canonical_markdown_override=md_override
+        ):
             raise HTTPException(status_code=404, detail="Chapter not found")
         asset_map = await signed_urls_for_asset_ids(
             self._db, collect_asset_ids_for_chapter(chapter)
         )
-        return chapter_to_dict(chapter, asset_url_map=asset_map)
+        return chapter_to_dict(
+            chapter,
+            asset_url_map=asset_map,
+            markdown_for_response=md_override,
+        )
 
     async def disable_chapter(self, chapter_id: str, user_id: str) -> dict:
         chapter = await repo.get_chapter_by_id(chapter_id, self._db)