feat: queue voice session cover generation jobs

backend/app/services/story_service.py

@@ -1241,13 +1241,16 @@ async def execute_generation_job_service(
     """Execute one previously accepted generation job inside the worker."""
 
     try:
-        request = GenerationRequest.model_validate(job.request_payload or {})
-        response = await _generate_generation_service_with_job(
-            request,
-            job.user_id,
-            db,
-            job=job,
-        )
+        if job.output_mode == "asset_generation":
+            response = await _generate_asset_generation_service_with_job(job, db)
+        else:
+            request = GenerationRequest.model_validate(job.request_payload or {})
+            response = await _generate_generation_service_with_job(
+                request,
+                job.user_id,
+                db,
+                job=job,
+            )
     except GenerationJobCanceledError:
         return _build_canceled_generation_response(job)
     except HTTPException as exc:
@@ -1294,6 +1297,39 @@ def _build_canceled_generation_response(job: GenerationJob) -> GenerationRespons
     )
 
 
+def _build_generation_response_from_story(
+    story: Story,
+    *,
+    job_id: str,
+) -> GenerationResponse:
+    """Build a unified generation response from one persisted story record."""
+
+    pages = None
+    if story.mode == "storybook":
+        pages = _storybook_pages_to_response(story.pages or [])
+
+    return GenerationResponse(
+        id=story.id,
+        generation_job_id=job_id,
+        title=story.title,
+        mode=story.mode,
+        story_text=story.story_text,
+        pages=pages,
+        cover_prompt=story.cover_prompt,
+        image_url=story.image_url,
+        cover_url=story.image_url,
+        audio_ready=story.audio_status == StoryAssetStatus.READY.value,
+        generation_status=story.generation_status,
+        text_status=story.text_status,
+        image_status=story.image_status,
+        audio_status=story.audio_status,
+        last_error=story.last_error,
+        child_profile_id=story.child_profile_id,
+        universe_id=story.universe_id,
+        retryable_assets=story.retryable_assets,
+    )
+
+
 async def run_generation_job_service(
     job_id: str,
     db: AsyncSession,
@@ -1309,6 +1345,54 @@ async def run_generation_job_service(
     return job
 
 
+async def _generate_asset_generation_service_with_job(
+    job: GenerationJob,
+    db: AsyncSession,
+) -> GenerationResponse:
+    """Run queued asset generation in the background worker."""
+
+    payload = job.request_payload or {}
+    story_id = payload.get("story_id") or job.story_id
+    requested_assets = list(dict.fromkeys(payload.get("assets") or []))
+    if story_id is None:
+        raise HTTPException(status_code=400, detail="资源任务缺少 story_id。")
+    if not requested_assets:
+        raise HTTPException(status_code=400, detail="资源任务缺少 assets。")
+
+    story = await get_story_detail(int(story_id), job.user_id, db)
+
+    if "image" in requested_assets:
+        if story.mode == "storybook":
+            await _complete_storybook_image_assets(story, db, job=job)
+        else:
+            await _complete_cover_image_asset(
+                story,
+                db,
+                raise_on_failure=True,
+                log_event="cover_generation_failed",
+                job=job,
+            )
+
+    if "audio" in requested_assets:
+        await _complete_audio_asset(
+            story,
+            db,
+            raise_on_failure=True,
+            job=job,
+        )
+
+    story = await get_story_detail(story.id, job.user_id, db)
+    await finish_generation_job(
+        db,
+        job=job,
+        story=story,
+        current_step="asset_generation_completed",
+        message="Asset generation completed in the background worker.",
+        metadata={"assets": requested_assets},
+    )
+    return _build_generation_response_from_story(story, job_id=job.id)
+
+
 async def retry_generation_job_service(
     job_id: str,
     user_id: str,
@@ -1534,6 +1618,46 @@ async def create_story_from_result(
     )
 
 
+async def queue_story_asset_generation(
+    story_id: int,
+    user_id: str,
+    assets: list[str],
+    db: AsyncSession,
+) -> dict:
+    """Queue one asset generation job for an already persisted story."""
+
+    await ensure_no_active_story_generation_job(db, story_id=story_id, user_id=user_id)
+    requested_assets = list(dict.fromkeys(assets))
+    if not requested_assets:
+        raise HTTPException(status_code=400, detail="至少需要一个待生成资源")
+
+    story = await get_story_detail(story_id, user_id, db)
+    if "image" in requested_assets:
+        has_image_prompt = bool(story.cover_prompt)
+        if story.mode == "storybook":
+            has_image_prompt = has_image_prompt or any(
+                isinstance(page, dict) and page.get("image_prompt")
+                for page in story.pages or []
+            )
+        if not has_image_prompt:
+            raise HTTPException(status_code=400, detail="当前故事没有可生成的图片提示词")
+
+    if "audio" in requested_assets and not story.story_text:
+        raise HTTPException(status_code=400, detail="当前故事没有可生成音频的正文")
+
+    job = await create_generation_job(
+        db,
+        user_id=user_id,
+        output_mode="asset_generation",
+        input_type=",".join(requested_assets),
+        request_payload={"story_id": story_id, "assets": requested_assets},
+        story_id=story_id,
+    )
+    await _dispatch_generation_job(db, job=job)
+    await db.refresh(job)
+    return generation_job_to_summary(job)
+
+
 async def _retry_cover_image_asset(story: Story, db: AsyncSession, *, job=None) -> None:
     """Retry cover generation for a text story."""
 

backend/app/services/voice_session_service.py

@@ -31,7 +31,7 @@ from app.services.memory_service import build_enhanced_memory_context
 from app.services.provider_router import generate_story_content, text_to_speech
 from app.services.story_service import (
     create_story_from_result,
-    generate_story_cover,
+    queue_story_asset_generation,
     validate_profile_and_universe,
 )
 from app.services.voice_session_safety import (
@@ -1710,15 +1710,21 @@ async def finalize_voice_session_service(
     generation_job_id: str | None = None
     if request.generate_cover and story.cover_prompt:
         try:
-            await generate_story_cover(story.id, user_id, db)
+            cover_job = await queue_story_asset_generation(
+                story.id,
+                user_id,
+                ["image"],
+                db,
+            )
+            generation_job_id = str(cover_job["id"])
             await _record_session_event(
                 db,
                 session_id=session.id,
                 turn_id=None,
-                event_type="session_cover_generation_succeeded",
+                event_type="session_cover_generation_queued",
                 status="succeeded",
-                message="Finalized story cover was generated after session save.",
-                metadata={"story_id": story.id},
+                message="Finalized story cover generation was queued after session save.",
+                metadata={"story_id": story.id, "generation_job_id": generation_job_id},
             )
         except HTTPException as exc:
             await _record_session_event(
@@ -1727,11 +1733,11 @@ async def finalize_voice_session_service(
                 turn_id=None,
                 event_type="session_cover_generation_failed",
                 status="failed",
-                message="Finalized story cover generation failed after session save.",
+                message="Finalized story cover generation failed before the worker could start.",
                 metadata={"story_id": story.id, "error": str(exc.detail)},
             )
             logger.warning(
-                "voice_session_finalize_cover_failed",
+                "voice_session_finalize_cover_queue_failed",
                 session_id=session.id,
                 story_id=story.id,
                 error=str(exc.detail),