storyforge/docs/superpowers/plans/2026-04-05-seedance2-ai-video-execution.md

Seedance 2.0 AI Video Execution Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: 在不打断现有 AI 视频链的前提下，新增 Seedance 2.0 兼容入口，并把前端、主 Agent、执行结果和回归链统一到可扩展的视频引擎模型。

Architecture: 保留现有 /v2/pipelines/ai-video 业务入口不变，把视频生成从“固定 huobao/doubao 默认值”升级为“provider-aware execution”。前端新增视频引擎选择与 Seedance 定向字段，后端统一把 provider/version/meta 写进 job artifacts 和结果，主 Agent 透传 provider 进入执行链并在结果卡中可追溯。

Tech Stack: FastAPI, SQLite-backed job records, existing HuobaoDramaClient integration layer, vanilla JS SPA, node test runner, unittest

---

Task 1: 后端视频引擎兼容层

Files:

• Modify: collector-service/app/core_main.py

• Modify: collector-service/app/integrations.py

• Test: tests/test_production_baseline.py

• Step 1: 写失败测试，锁定 video_provider 会进入 AI 视频 job artifacts 和执行结果

def test_ai_video_request_persists_video_provider(client, approved_headers):
    response = client.post(
        "/v2/pipelines/ai-video",
        json={
            "project_id": "project_demo",
            "assistant_id": "",
            "knowledge_base_id": "kb_demo",
            "title": "Seedance test",
            "brief": "创业者口播视频",
            "style": "realistic",
            "shots": 4,
            "duration": 5,
            "video_provider": "seedance2",
            "video_model": "seedance-2.0-pro",
        },
        headers=approved_headers,
    )
    assert response.status_code == 200
    payload = response.json()
    assert payload["artifacts"]["video_provider"] == "seedance2"
    assert payload["artifacts"]["video_model"] == "seedance-2.0-pro"

• Step 2: 运行测试确认先失败或至少缺少 seedance2 兼容断言

Run: python3 -m unittest tests.test_production_baseline -v Expected: 现有测试不覆盖 seedance2，新增断言失败或行为缺失。

• Step 3: 实现 provider 兼容层

def normalize_ai_video_provider(provider: str, model: str) -> tuple[str, str]:
    normalized = str(provider or "").strip().lower()
    if normalized in {"seedance", "seedance2", "seedance-2.0", "seedance_2_0"}:
        return "seedance2", model or "seedance-2.0-pro"
    if normalized in {"doubao", "jimeng", "huobao"}:
        return "doubao", model
    return normalized or "doubao", model

video_provider, video_model = normalize_ai_video_provider(request.video_provider, request.video_model)

"video_provider": video_provider,
"video_model": video_model,
"video_provider_label": "Seedance 2.0" if video_provider == "seedance2" else video_provider,

• Step 4: 在内部渲染步骤里把 provider 元信息回写进结果

"video_provider": video_provider,
"video_model": video_model,
"video_provider_label": "Seedance 2.0" if video_provider == "seedance2" else video_provider,

• Step 5: 再跑后端测试确认通过

Run: python3 -m unittest tests.test_production_baseline -v Expected: PASS

• Step 6: 提交这一批

git add collector-service/app/core_main.py collector-service/app/integrations.py tests/test_production_baseline.py
git commit -m "feat: add seedance2 ai video provider support"

Task 2: 前端 AI 视频任务表单与结果展示

Files:

• Modify: web/storyforge-web-v4/assets/app.js

• Test: web/storyforge-web-v4/tests/workbench-pages.test.mjs

• Step 1: 写失败测试，锁定 AI 视频表单出现视频引擎选择和 Seedance 定向字段

test("ai video action exposes seedance 2.0 provider controls", () => {
  const aiVideo = extractBetween(APP, "function openCreateAiVideoAction(defaults = {})", "function openCreateRealCutAction(defaults = {})");
  assert.match(aiVideo, /videoProvider/);
  assert.match(aiVideo, /Seedance 2.0/);
  assert.match(aiVideo, /cameraMotion/);
  assert.match(aiVideo, /visualStyle/);
});

• Step 2: 跑测试确认先失败

Run: node --test web/storyforge-web-v4/tests/workbench-pages.test.mjs Expected: FAIL because current form has no Seedance fields.

• Step 3: 实现前端表单扩展

{ name: "videoProvider", label: "视频引擎", type: "select", value: defaults.videoProvider || "doubao", options: [
  { value: "doubao", label: "当前默认" },
  { value: "seedance2", label: "Seedance 2.0" }
] }

{ name: "cameraMotion", label: "镜头运动", value: defaults.cameraMotion || "", placeholder: "例如：慢推近、环绕、手持跟拍" },
{ name: "visualStyle", label: "画面风格", value: defaults.visualStyle || "", placeholder: "例如：电影感、高反差、创业访谈质感" },

• Step 4: 提交时把 provider 和 Seedance 定向字段发给后端

video_provider: values.videoProvider || "doubao",
video_model: values.videoProvider === "seedance2" ? "seedance-2.0-pro" : "",
camera_motion: values.cameraMotion || "",
visual_style: values.visualStyle || "",

• Step 5: 在 job 详情/结果卡里显示视频引擎

<span class="tag blue">${escapeHtml(job.artifacts?.video_provider_label || job.artifacts?.video_provider || "默认引擎")}</span>

• Step 6: 跑前端测试确认通过

Run: node --test web/storyforge-web-v4/tests/workbench-pages.test.mjs Expected: PASS

• Step 7: 提交这一批

git add web/storyforge-web-v4/assets/app.js web/storyforge-web-v4/tests/workbench-pages.test.mjs
git commit -m "feat: add seedance2 controls to ai video flow"

Task 3: 主 Agent 透传与执行追溯

Files:

• Modify: collector-service/app/oneliner_features.py

• Test: tests/test_main_agent_governance.py

• Step 1: 写失败测试，锁定主 Agent 创建 AI 视频任务时会透传 provider

def test_oneliner_ai_video_run_preserves_video_provider(client, approved_headers):
    response = client.post(
        "/v2/oneliner/actions/execute",
        json={
            "action_key": "create-ai-video",
            "project_id": "project_demo",
            "platform": "douyin",
            "payload": {
                "video_provider": "seedance2",
                "video_model": "seedance-2.0-pro",
            },
        },
        headers=approved_headers,
    )
    assert response.status_code == 200
    payload = response.json()
    assert payload["payload"]["job"]["artifacts"]["video_provider"] == "seedance2"

• Step 2: 跑测试确认先失败

Run: python3 -m unittest tests.test_main_agent_governance -v Expected: FAIL because provider not yet propagated.

• Step 3: 实现主 Agent 透传

video_provider=str(requested_payload.get("video_provider") or requested_payload.get("videoProvider") or "doubao"),
video_model=str(requested_payload.get("video_model") or requested_payload.get("videoModel") or ""),

"video_provider": job.get("artifacts", {}).get("video_provider", ""),
"video_provider_label": job.get("artifacts", {}).get("video_provider_label", ""),

• Step 4: 结果卡补充视频引擎标签

result_sections.append({
    "title": "视频引擎",
    "items": [{"label": "本轮视频引擎", "value": provider_label}],
})

• Step 5: 跑后端治理测试确认通过

Run: python3 -m unittest tests.test_main_agent_governance -v Expected: PASS

• Step 6: 提交这一批

git add collector-service/app/oneliner_features.py tests/test_main_agent_governance.py
git commit -m "feat: propagate seedance2 through main agent"

Task 4: 整体回归、部署与版本记录

Files:

• Modify: CHANGELOG.md

• Step 1: 更新版本记录

### Seedance 2.0 生视频兼容接入

- AI 视频链新增 `Seedance 2.0` 视频引擎入口，并统一把 provider/version 回写到任务结果与主 Agent 执行卡。
- 前端 AI 视频创建表单新增视频引擎选择与 Seedance 定向参数。
- 主 Agent 创建 AI 视频任务时会透传视频引擎，结果页可追溯本轮使用的引擎。

• Step 2: 跑全量回归

Run: python3 -m unittest tests.test_main_agent_governance tests.test_platform_contracts tests.test_production_baseline -v Expected: PASS

Run: node --test web/storyforge-web-v4/tests/dashboard-home.test.mjs web/storyforge-web-v4/tests/workbench-pages.test.mjs Expected: PASS

Run: bash scripts/check_repo_baseline.sh Expected: baseline checks passed

• Step 3: 发布到 NAS 并跑 smoke

Run: FNOS_PASSWORD='Admin_yqs_asd20260101.' bash /Users/kris/code/StoryForge-gitea/scripts/deploy_fnos_storyforge_web.sh Expected: fnOS StoryForge Web V4 ready: http://192.168.31.188:19192/

Run: bash /Users/kris/code/StoryForge-gitea/scripts/smoke_fnos_storyforge_lan.sh Expected: fnOS lan smoke passed

• Step 4: 提交汇总版本

git add CHANGELOG.md
git commit -m "chore: record seedance2 ai video rollout"