feat: queue voice session cover generation jobs

2026-04-20 17:05:01 +08:00
parent 4d7072fb66
commit df4f5f4cdd
6 changed files with 254 additions and 21 deletions
--- a/backend/app/services/story_service.py
+++ b/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."""

--- a/backend/app/services/voice_session_service.py
+++ b/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),
--- a/backend/tests/test_generation_jobs.py
+++ b/backend/tests/test_generation_jobs.py
@@ -8,7 +8,7 @@ from httpx import ASGITransport, AsyncClient
 from sqlalchemy import select

 from app.db.database import get_db
-from app.db.models import GenerationJob, GenerationJobEvent
+from app.db.models import GenerationJob, GenerationJobEvent, Story
 from app.main import app
 from app.services.adapters import AdapterConfig
 from app.services.adapters.storybook.primary import Storybook, StorybookPage
@@ -20,7 +20,7 @@ from app.services.generation_jobs import (
    mark_stale_generation_jobs,
    record_generation_event,
 )
-from app.services.story_service import run_generation_job_service
+from app.services.story_service import queue_story_asset_generation, run_generation_job_service

 pytestmark = pytest.mark.asyncio

@@ -225,6 +225,86 @@ async def test_asset_retry_records_job_events_and_updates_retryable_assets(
        app.dependency_overrides.clear()


+async def test_queue_story_asset_generation_dispatches_background_job(
+    db_session,
+    test_story,
+):
+    task_delay_path = "app.tasks.generation_workflow.run_generation_workflow_task.delay"
+
+    with patch(task_delay_path) as mock_delay:
+        summary = await queue_story_asset_generation(
+            test_story.id,
+            test_story.user_id,
+            ["image"],
+            db_session,
+        )
+
+    assert summary["output_mode"] == "asset_generation"
+    assert summary["input_type"] == "image"
+    assert summary["status"] == "running"
+    assert summary["current_step"] == "request_accepted"
+    mock_delay.assert_called_once_with(summary["id"])
+
+    job = (
+        await db_session.execute(
+            select(GenerationJob).where(GenerationJob.id == summary["id"])
+        )
+    ).scalar_one()
+    assert job.story_id == test_story.id
+    assert job.output_mode == "asset_generation"
+
+
+async def test_asset_generation_job_worker_completes_cover_image(
+    db_session,
+    test_story,
+):
+    job = await create_generation_job(
+        db_session,
+        user_id=test_story.user_id,
+        output_mode="asset_generation",
+        input_type="image",
+        request_payload={"story_id": test_story.id, "assets": ["image"]},
+        story_id=test_story.id,
+    )
+
+    with patch(
+        "app.services.story_service.generate_image",
+        new_callable=AsyncMock,
+    ) as mock_generate_image:
+        mock_generate_image.return_value = "https://example.com/async-cover.png"
+
+        await run_generation_job_service(job.id, db_session)
+
+    refreshed_job = (
+        await db_session.execute(select(GenerationJob).where(GenerationJob.id == job.id))
+    ).scalar_one()
+    assert refreshed_job.status == "completed"
+    assert refreshed_job.current_step == "asset_generation_completed"
+    assert refreshed_job.result_snapshot["image_status"] == "ready"
+
+    story = (
+        await db_session.execute(
+            select(Story).where(Story.id == test_story.id)
+        )
+    ).scalar_one()
+    assert story.image_url == "https://example.com/async-cover.png"
+
+    events = (
+        await db_session.execute(
+            select(GenerationJobEvent)
+            .where(GenerationJobEvent.job_id == job.id)
+            .order_by(GenerationJobEvent.id)
+        )
+    ).scalars().all()
+    assert [event.event_type for event in events] == [
+        "request_accepted",
+        "worker_started",
+        "cover_image_started",
+        "cover_image_succeeded",
+        "asset_generation_completed",
+    ]
+
+
 async def test_storybook_generation_is_queued_then_worker_records_page_image_events(
    db_session,
    auth_token,
--- a/backend/tests/test_voice_sessions.py
+++ b/backend/tests/test_voice_sessions.py
@@ -99,9 +99,9 @@ async def test_voice_session_correct_turn_and_finalize_to_story(
            new_callable=AsyncMock,
        ) as mock_tts,
        patch(
-            "app.services.voice_session_service.generate_story_cover",
+            "app.services.voice_session_service.queue_story_asset_generation",
            new_callable=AsyncMock,
-        ) as mock_generate_cover,
+        ) as mock_queue_asset_generation,
    ):
        mock_generate.side_effect = [
            StoryOutput(
@@ -118,7 +118,23 @@ async def test_voice_session_correct_turn_and_finalize_to_story(
            ),
        ]
        mock_tts.side_effect = [b"turn-1-audio", b"turn-2-audio"]
-        mock_generate_cover.return_value = "https://example.com/voice-cover.png"
+        mock_queue_asset_generation.return_value = {
+            "id": "cover-job-123",
+            "story_id": 1,
+            "output_mode": "asset_generation",
+            "input_type": "image",
+            "status": "running",
+            "current_step": "request_accepted",
+            "progress_percent": 5,
+            "progress_label": "已接收请求",
+            "is_terminal": False,
+            "can_cancel": True,
+            "can_retry": False,
+            "result_snapshot": {},
+            "error_message": None,
+            "created_at": "2026-04-20T00:00:00Z",
+            "updated_at": "2026-04-20T00:00:00Z",
+        }

        transport = ASGITransport(app=app)
        try:
@@ -156,6 +172,7 @@ async def test_voice_session_correct_turn_and_finalize_to_story(
                finalize_data = response.json()
                story_id = finalize_data["story_id"]
                assert finalize_data["status"] == "completed"
+                assert finalize_data["generation_job_id"] == "cover-job-123"

                response = await client.get(f"/api/stories/{story_id}")
                assert response.status_code == 200
@@ -172,7 +189,7 @@ async def test_voice_session_correct_turn_and_finalize_to_story(
                assert session_data["final_story_id"] == story_id
                assert session_data["can_continue"] is False
                assert session_data["story_state"]["final_summary"]
-                mock_generate_cover.assert_awaited_once()
+                mock_queue_asset_generation.assert_awaited_once()
        finally:
            app.dependency_overrides.clear()

--- a/docs/technical/voice-co-creation-phase-a-migration-api-draft.md
+++ b/docs/technical/voice-co-creation-phase-a-migration-api-draft.md
@@ -594,6 +594,11 @@ async def resolve_voice_turn_confirmation(...)
 async def finalize_voice_session(...)
 ```

+说明：
+
+- 当 `generate_cover=true` 且已产出 `cover_prompt` 时，finalize 后会额外排队一个 `asset_generation` job
+- 前端可通过返回的 `generation_job_id` 继续复用现有 generation trace / job detail 组件
+
 ### 获取语音共创 analytics

 ```python
--- a/docs/technical/voice-co-creation-phase-a-tech-spec.md
+++ b/docs/technical/voice-co-creation-phase-a-tech-spec.md
@@ -27,8 +27,9 @@
 - 前端明确展示“本轮系统理解为”与“建议家长确认后再继续”提示
 - 低置信度确认链路已有后端测试覆盖，可作为下一阶段继续接 ASR 与更细确认交互的基础
 - 已新增用户转写安全检查、assistant 输出柔性改写与 `safety_flags` 事件记录
- finalize 会生成更稳定的标题/摘要，并在条件允许时自动衔接封面补全
+- finalize 会生成更稳定的标题/摘要，并在条件允许时自动排队封面补全 job
 - 已新增 `voice session analytics` 聚合指标，可跟踪 turn 成功率、ASR/TTS 失败、低置信度触发和 finalize 转化率
+- `voice session finalize` 现在会返回可追踪的 `generation_job_id`，让正式 Story 资产补全重新接回现有 generation trace 主干

 Phase A 的核心目标不是做“完全实时的语音陪伴”，而是验证以下最小闭环：