.env.example

@@ -220,6 +220,20 @@ DOUBAO_MAX_RETRIES=2
 DOUBAO_VISION_MODEL=doubao-1-5-vision-pro-250328
 DOUBAO_VISION_LITE_MODEL=doubao-1-5-vision-lite-250315
 DOUBAO_VISION_USE_LITE=true
+# Embedding 向量化模型（原 large-text-240915 已下线，用多模态 embedding）
+DOUBAO_EMBEDDING_MODEL=doubao-embedding-vision-251215
+
+# ==================== 即梦（Jimeng）视觉 API —— 真人参考图兜底通道 (#2169) ====
+# 方舟 Seedance 走 B 端审核，真人参考图会被 50411 拦截；即梦走 C 端审核，普通真人照片可过审。
+# 需要在火山控制台开通即梦 cvtob 服务，使用 AK/SK（Region=cn-north-1, Service=cv）
+# 留空则真人拦截后直接返回错误提示，不会走即梦兜底。
+JIMENG_AK=
+JIMENG_SK=
+JIMENG_BASE_URL=https://visual.volcengineapi.com
+# 视频3.0 720P 首帧图生视频（1张图，稳定版；1080P标注下线中）
+JIMENG_REQ_KEY=jimeng_i2v_first_v30
+JIMENG_VIDEO_TIMEOUT=600
+JIMENG_VIDEO_POLL_INTERVAL=5
 
 # ==================== 积分/会员系统 (#1895) ====================
 # 积分系统总开关：默认 false（暂停积分系统）。

apps/worker/worker_app/tasks/viral_video.py

@@ -647,7 +647,7 @@ def _build_products_summary(image_analysis: dict) -> str:
         # 优先 VLM 生成的 summary 段（自然语言，给编导模型看效果最好）
         summary = (p.get("summary") or "").strip()
         if summary and len(summary) >= 30:
-            lines.append(f"- 图{i+1} {name}：{summary}")
+            lines.append(f"- 图{i + 1} {name}：{summary}")
             continue
         # 结构化字段兜底
         brand = p.get("brand") or ""
@@ -669,7 +669,7 @@ def _build_products_summary(image_analysis: dict) -> str:
         feats = p.get("key_features") or p.get("features") or []
         sellings = p.get("selling_points") or []
         scenes = p.get("suitable_scenes") or []
-        parts = [f"图{i+1} {name}"]
+        parts = [f"图{i + 1} {name}"]
         if brand and brand not in ("未知", "无法判断"):
             parts.append(f"品牌={brand}")
         if cat and cat not in ("无法判断", "非产品图"):
@@ -802,7 +802,7 @@ def _validate_and_normalize_script(raw, job: ViralVideoJob) -> dict:
                 continue
             shots.append(
                 {
-                    "time_range": str(s.get("time_range") or f"{i*3}-{(i+1)*3}秒"),
+                    "time_range": str(s.get("time_range") or f"{i * 3}-{(i + 1) * 3}秒"),
                     "shot_type_angle_movement": str(s.get("shot_type_angle_movement") or "中景平视，固定镜头"),
                     "scene_and_dialogue": str(s.get("scene_and_dialogue") or ""),
                     "action_details": str(s.get("action_details") or ""),
@@ -878,8 +878,8 @@ def _step_script_generation(job: ViralVideoJob, intent: dict, image_analysis: di
     style_hint = "无"
     if isinstance(job.style_guide, dict):
         style_hint = (
-            f"节奏{job.style_guide.get('cut_speed','')}、转场{job.style_guide.get('transition','')}、"
-            f"色调{job.style_guide.get('color_grade','')}、能量{job.style_guide.get('energy','')}"
+            f"节奏{job.style_guide.get('cut_speed', '')}、转场{job.style_guide.get('transition', '')}、"
+            f"色调{job.style_guide.get('color_grade', '')}、能量{job.style_guide.get('energy', '')}"
         )
 
     dur = max(5, min(30, int(getattr(job, "duration", 15) or 15)))
@@ -1103,14 +1103,14 @@ def _assemble_seedance_prompt(copy_result: dict, job: ViralVideoJob) -> str:
         ab = s.get("audio_bgm", "")
         t = s.get("transition", "")
         ref = s.get("reference_image_index")
-        lines.append(f"- 镜头{i+1}（{tr}）：")
+        lines.append(f"- 镜头{i + 1}（{tr}）：")
         lines.append(f"  景别/运镜：{cam}")
         lines.append(f"  画面与对白：{sd}")
         lines.append(f"  动作细节：{act}")
         lines.append(f"  音效/BGM：{ab}")
         lines.append(f"  转场：{t}")
         if ref is not None and isinstance(ref, int):
-            lines.append(f"  参考图片：第{ref+1}张产品图")
+            lines.append(f"  参考图片：第{ref + 1}张产品图")
     lines.append("")
     lines.append("【硬性约束】")
     for c in hc:
@@ -1177,8 +1177,9 @@ def _step_render(job: ViralVideoJob, copy_result: dict, tts_audio_url: str | Non
         reference_videos=ref_videos,
     )
 
-    # 肖像/内容审核拦截降级重试：去掉参考图/参考视频走纯文字 t2v 再试一次
-    # （不影响音频——去掉首帧和多参考图通常就能绕过真人照片审核，纯文字描述真人是允许的）
+    # #2169: 真人/肖像拦截已由 ai_client 内部自动切即梦（jimeng-3.0）通道重试——
+    # 保留首帧图、不走"去掉参考图纯 t2v 降级"（用户明确要求按参考照片生成）。
+    # 即梦也失败或非拦截类错误时，直接抛错给上层展示用户友好提示。
     def _check_and_reraise(result):
         if result and isinstance(result, dict):
             return result
@@ -1189,11 +1190,11 @@ def _step_render(job: ViralVideoJob, copy_result: dict, tts_audio_url: str | Non
         detail = err.get("detail") or ""
         err_code = err.get("error_code") or "unknown"
         status_code = err.get("status_code", 0)
-        provider = _mcfg.get("provider", "doubao")
-        err_msg = user_msg or f"视频生成失败（{provider} status={status_code} code={err_code}）"
+        err_provider = err.get("provider") or _mcfg.get("provider", "doubao")
+        err_msg = user_msg or f"视频生成失败（{err_provider} status={status_code} code={err_code}）"
         logger.error(
             "[爆款视频] 视频生成失败: provider=%s model=%s code=%s status=%s user_msg=%s detail=%s",
-            provider,
+            err_provider,
             model or "default",
             err_code,
             status_code,
@@ -1203,38 +1204,24 @@ def _step_render(job: ViralVideoJob, copy_result: dict, tts_audio_url: str | Non
         raise RuntimeError(err_msg)
 
     if not result or not isinstance(result, dict):
-        from packages.shared.ai_service import get_last_video_error
-
-        err = get_last_video_error() or {}
-        err_code = err.get("error_code") or "unknown"
-        is_portrait = err_code == "portrait_intercept" or any(
-            k in (err.get("detail", "") or "").lower()
-            for k in ("real person", "privacy", "portrait", "sensitivecontent", "face", "人脸", "肖像", "真人")
-        )
-        if is_portrait and (first_image or rest_images or ref_videos):
-            logger.warning("[爆款视频] 参考素材触发真人/肖像审核拦截(error_code=%s)，降级为纯文字t2v重试一次", err_code)
-            # 去掉所有参考图/参考视频，纯 t2v 泛化生成；音频仍保留（不影响肖像审核）
-            result = call_video_generation(
-                prompt=prompt,
-                image_url=None,
-                duration=dur,
-                ratio=ratio,
-                resolution=resolution,
-                output_dir=str(tmpdir),
-                model=model,
-                generate_audio=gen_audio,
-                reference_images=[],
-                reference_audios=ref_audios,
-                reference_videos=[],
-            )
         _check_and_reraise(result)
     video_path = result.get("video_path") or ""
     usage = result.get("usage")
     if not video_path or not Path(video_path).exists() or Path(video_path).stat().st_size == 0:
-        raise RuntimeError("Seedance 视频生成失败：返回空文件或路径不存在")
-    logger.info(
-        "[爆款视频] Seedance 单次生成完成: %s size=%d usage=%s", video_path, Path(video_path).stat().st_size, usage
-    )
+        raise RuntimeError("视频生成失败：返回空文件或路径不存在")
+    # #2169: 如果实际走了即梦兜底（真人拦截→jimeng），更新 job.video_model 让积分结算用 jimeng-3.0 价格
+    if isinstance(usage, dict):
+        actual_provider = usage.get("provider")
+        actual_model_key = usage.get("model_key")
+        if actual_provider == "jimeng" and actual_model_key:
+            logger.info(
+                "[爆款视频] 实际通过即梦通道生成(原model=%s)，更新video_model=%s 用于积分结算",
+                job.video_model,
+                actual_model_key,
+            )
+            job.video_model = actual_model_key
+    size = Path(video_path).stat().st_size
+    logger.info("[爆款视频] 单次生成完成: path=%s size=%d usage=%s", video_path, size, usage)
     return str(video_path), (usage if isinstance(usage, dict) else None)
 
 

packages/config/base.py

@@ -90,15 +90,21 @@ class SharedSettings(BaseSettings):
 
     # ── 豆包大模型（火山引擎方舟） ────────────────────────────────────────
     doubao_api_key: str = ""
-    doubao_model: str = "doubao-seed-1-6-250615"  # 推理模型（通用兜底）
-    doubao_fast_model: str = "doubao-1-5-pro-32k-250115"  # 快速结构化输出模型（编导脚本/意图解析/审核）
+    doubao_model: str = "doubao-seed-2-1-pro-260915"  # 推理模型（Seed 2.1 Pro，深度思考+多模态；原 seed-1-6 已下线）
+    doubao_fast_model: str = (
+        "doubao-seed-2-1-lite-260915"  # 快速模型（Seed 2.1 Lite，高 RPM，编导/审核/VLM lite；原 1-5-pro-32k 已 Retiring）
+    )
     doubao_base_url: str = "https://ark.cn-beijing.volces.com/api/v3"
     doubao_timeout: int = 30
     doubao_max_retries: int = 2
-    doubao_vision_model: str = "doubao-1-5-vision-pro-250328"  # 高精度视觉（备用）
-    doubao_vision_lite_model: str = "doubao-1-5-vision-lite-250315"  # 快速视觉（商品识别默认，速度优先）
+    doubao_vision_model: str = (
+        "doubao-seed-2-1-pro-260915"  # 高精度视觉（Seed 2.1 Pro 原生多模态；原 vision-pro-250328 已下线）
+    )
+    doubao_vision_lite_model: str = (
+        "doubao-seed-2-1-lite-260915"  # 快速视觉（Seed 2.1 Lite 原生多模态；原 vision-lite-250315 不可用）
+    )
     doubao_vision_use_lite: bool = True  # viral-video 图片分析默认用 lite 提速
-    doubao_embedding_model: str = "doubao-embedding-large-text-240915"
+    doubao_embedding_model: str = "doubao-embedding-vision-251215"  # 多模态向量化（原 large-text-240915 已 Retiring）
     doubao_video_model: str = "doubao-seedance-2-5-260628"
     doubao_video_timeout: int = 600  # 视频生成轮询总超时（秒）
     doubao_video_poll_interval: int = 10  # 轮询间隔（秒）
@@ -109,6 +115,17 @@ class SharedSettings(BaseSettings):
     dashscope_video_timeout: int = 900  # Wan 视频任务轮询总超时（秒）
     dashscope_video_poll_interval: int = 10
 
+    # ── 即梦（Jimeng）视觉 API —— 火山引擎 cvtob ──────────────────────────
+    # #2169: 真人参考图在方舟 Seedance 走 B 端审核会被 50411 拦截，
+    # 即梦走 C 端审核链路，普通真人照片可过审，作为参考图场景兜底通道。
+    # 鉴权：AK/SK V4 签名（Region=cn-north-1, Service=cv）
+    jimeng_ak: str = ""
+    jimeng_sk: str = ""
+    jimeng_base_url: str = "https://visual.volcengineapi.com"
+    jimeng_req_key: str = "jimeng_i2v_first_v30"  # 视频3.0 720P 首帧图生视频（1张图，稳定版；1080P 标注下线中）
+    jimeng_video_timeout: int = 600  # 即梦轮询总超时（秒）
+    jimeng_video_poll_interval: int = 5  # 轮询间隔（秒）
+
     # ── MediaKit (火山引擎 AI 媒体工具) ──────────────────────────────────
     mediakit_api_key: str = ""
     mediakit_base_url: str = "https://mediakit.cn-beijing.volces.com/api/v1"

packages/domain/points_rules.py

@@ -31,6 +31,8 @@ VIRAL_VIDEO_MODEL_PRICES: dict[tuple[str, str, bool], float] = {
     ("wan-3.0", "480p", False): 0.3,
     ("wan-3.0", "720p", False): 0.6,
     ("wan-3.0", "1080p", False): 1.2,
+    # #2169: 即梦（Jimeng）视频3.0 720P 首帧图生视频，0.28 元/秒（C 端审核，真人可过）
+    ("jimeng-3.0", "720p", False): 0.28,
 }
 
 # 固定成本（元）：VLM 分析 + LLM 文案 + TTS + OSS + 服务器
@@ -151,6 +153,20 @@ VIRAL_VIDEO_MODEL_CONFIG: dict[str, dict] = {
         "billing_mode": "per_second",
         "is_default": False,
     },
+    # #2169: 即梦视频3.0（内部兜底通道，方舟 Seedance 返回真人拦截 50411 时自动切到即梦重试，
+    # 不暴露给前端让用户直接选择，但需支持计费结算）
+    "jimeng-3.0": {
+        "key": "jimeng-3.0",
+        "display_name": "即梦3.0 — 真人图生视频（兜底）",
+        "model_id": "jimeng_i2v_first_v30",
+        "provider": "jimeng",
+        "supports_audio": False,  # 即梦返回无声视频，音频由后续 ffmpeg 合成 TTS
+        "supported_resolutions": ["720p"],
+        "max_duration": 10,  # 即梦 i2v 首帧最长 10s（frames=241）
+        "billing_mode": "per_second",
+        "is_default": False,
+        "_internal_fallback_only": True,  # 标记：不对外暴露到模型选择列表
+    },
 }
 
 
@@ -173,6 +189,9 @@ def list_viral_video_models(
             continue
         if cfg.get("provider") == "dashscope" and not dashscope_available:
             continue
+        # #2169: 即梦是内部兜底通道，不在前端模型列表展示
+        if cfg.get("_internal_fallback_only"):
+            continue
         out.append(
             {
                 "key": cfg["key"],

packages/shared/ai_client.py

@@ -34,20 +34,22 @@ _HTTP_STATUS_ERROR = httpx.HTTPStatusError if hasattr(httpx, "HTTPStatusError")
 logger = logging.getLogger(__name__)
 
 
-# 视频模型 ID 解析逻辑（#2159 多模型支持）。
-# 内部使用简短别名（seedance-2.5 / seedance-2.0 / wan-3.0 等）做 PRICING key 和前端选择值；
+# 视频模型 ID 解析逻辑（#2159 多模型支持，#2169 接入即梦）。
+# 内部使用简短别名（seedance-2.5 / seedance-2.0 / wan-3.0 / jimeng-3.0 等）做 PRICING key；
 # 实际调用时按 VIRAL_VIDEO_MODEL_CONFIG（domain/points_rules.py）的 provider/model_id 分发。
-# - provider=doubao → 火山方舟
+# - provider=doubao → 火山方舟 Seedance
 # - provider=dashscope → 阿里云 DashScope（Wan 系列）
+# - provider=jimeng → 火山引擎即梦 cvtob（jimeng_i2v_first_v30，真人参考图走 C 端审核）
 
 
 def _resolve_video_provider_and_id(model: str | None) -> tuple[str, str, dict]:
     """把内部 model key 解析成 (provider, model_id, cfg)。
 
-    - provider: "doubao" | "dashscope"
+    - provider: "doubao" | "dashscope" | "jimeng"
     - model_id: 对应 API 的真实模型 ID
     - cfg: VIRAL_VIDEO_MODEL_CONFIG 条目
-    未识别或空值回落到默认 seedance-2.5。已经是 doubao-/ep- 开头的完整 ID 视为 doubao provider。
+    未识别或空值回落到默认 seedance-2.5。已经是 doubao-/ep- 开头的完整 ID 视为 doubao provider；
+    "jimeng" 开头视为 jimeng provider（内部兜底，不暴露给前端）。
     """
     from packages.domain.points_rules import get_viral_video_model_config
 
@@ -60,6 +62,10 @@ def _resolve_video_provider_and_id(model: str | None) -> tuple[str, str, dict]:
     # 已经是 doubao-/ep- 开头：直接透传，默认视为 doubao provider
     if m.startswith("doubao-") or m.startswith("ep-"):
         return "doubao", m, {"provider": "doubao", "model_id": m, "supports_audio": True}
+    # 显式 jimeng 关键字：路由到即梦（内部兜底通道使用）
+    if m.startswith("jimeng"):
+        cfg = get_viral_video_model_config("jimeng-3.0")
+        return "jimeng", cfg.get("model_id", "jimeng_i2v_first_v30"), cfg
     # 别名 → 从 domain config 查
     cfg = get_viral_video_model_config(m)
     provider = cfg.get("provider", "doubao")
@@ -180,6 +186,7 @@ class DoubaoClient:
         self.vision_model: str = settings.doubao_vision_model
         self.vision_lite_model: str = settings.doubao_vision_lite_model
         self.fast_model: str = settings.doubao_fast_model
+        self.embedding_model: str = settings.doubao_embedding_model
         # 最近一次视频生成的详细错误（error_code + user_message + raw detail），供上层读取后展示给用户
         self.last_video_error: dict = {}
 
@@ -194,7 +201,7 @@ class DoubaoClient:
             "Content-Type": "application/json",
         }
         payload: dict[str, Any] = {
-            "model": getattr(self, "embedding_model", None) or "doubao-embedding-large-text-240915",
+            "model": self.embedding_model,
             "input": text.strip(),
             "encoding_format": "float",
         }
@@ -480,6 +487,18 @@ class DoubaoClient:
                 }
                 return None
 
+        if provider == "jimeng":
+            # #2169: 即梦 cvtob（jimeng_i2v_first_v30）— 真人参考图兜底通道
+            return self._call_jimeng_video_generation(
+                prompt=prompt,
+                image_url=image_url,
+                duration=duration,
+                ratio=ratio,
+                resolution=resolution,
+                output_dir=output_dir,
+                generate_audio=generate_audio,
+            )
+
         ref_audios = [u for u in (reference_audios or [])[:10] if u and isinstance(u, str)]
         ref_videos = [u for u in (reference_videos or [])[:3] if u and isinstance(u, str)]
         ref_imgs = [u for u in (reference_images or [])[:9] if u and isinstance(u, str)]
@@ -650,6 +669,28 @@ class DoubaoClient:
                 last_err,
                 (body or "")[:500],
             )
+            # #2169: 方舟返回 portrait_intercept 且有参考图 → 自动切即梦重试一次（保留首帧图）
+            if err_code == "portrait_intercept" and image_url:
+                logger.warning(
+                    "[viral-video] Seedance 真人拦截(code=%s)，自动切即梦通道重试(首帧图) img=%s",
+                    err_code,
+                    bool(image_url),
+                )
+                jm_result = self._call_jimeng_video_generation(
+                    prompt=prompt,
+                    image_url=image_url,
+                    duration=duration,
+                    ratio=ratio,
+                    resolution=resolution,
+                    output_dir=output_dir,
+                    generate_audio=False,  # 即梦 i2v 不带音频，音频由后续 ffmpeg 合成
+                    _portrait_fallback=True,
+                )
+                if jm_result is not None:
+                    return jm_result
+                # 即梦也失败了，保留即梦的 last_video_error（已经由 _call_jimeng 设置）
+                logger.error("[viral-video] 即梦通道重试也失败: %s", self.last_video_error)
+                return None
             return None
 
         logger.info("Seedance 任务已创建: task_id=%s ratio=%s", task_id, create_payload["ratio"])
@@ -793,6 +834,100 @@ class DoubaoClient:
             }
             return None
 
+    def _call_jimeng_video_generation(
+        self,
+        *,
+        prompt: str,
+        image_url: str | None,
+        duration: int,
+        ratio: str | None,
+        resolution: str,
+        output_dir: str | None,
+        generate_audio: bool = False,
+        _portrait_fallback: bool = False,
+    ) -> dict | None:
+        """#2169: 调用即梦 cvtob 客户端做图生视频（真人参考图兜底通道）。
+
+        - 即梦 i2v 首帧接口只接受 1 张图、无原生音频（返回无声视频，音频由 ffmpeg 后合）。
+        - 成功返回 {"video_path": str, "usage": {...}}；失败写 self.last_video_error 并返回 None。
+        - _portrait_fallback=True 时在日志里标注是从方舟拦截切过来的。
+        """
+        from packages.shared.jimeng_client import get_jimeng_client
+
+        jm = get_jimeng_client()
+        if jm is None:
+            detail = "即梦 client 不可用（JIMENG_AK/SK 未配置）"
+            if _portrait_fallback:
+                # 从真人拦截切过来但即梦没配，仍把错误归到 portrait_intercept，让上层提示用户
+                self.last_video_error = {
+                    "error_code": "portrait_intercept",
+                    "user_message": "参考素材包含真人照片被安全策略拦截，即梦兜底通道未启用，请联系管理员配置 JIMENG_AK/SK。",
+                    "status_code": 0,
+                    "detail": detail,
+                    "provider": "jimeng",
+                }
+            else:
+                self.last_video_error = {
+                    "error_code": "auth_error",
+                    "user_message": "即梦视频通道未配置（JIMENG_AK/SK 缺失），请联系管理员。",
+                    "status_code": 0,
+                    "detail": detail,
+                    "provider": "jimeng",
+                }
+            logger.error("[jimeng] %s, portrait_fallback=%s", detail, _portrait_fallback)
+            return None
+        if not image_url:
+            self.last_video_error = {
+                "error_code": "invalid_param",
+                "user_message": "即梦图生视频必须提供参考图片。",
+                "status_code": 0,
+                "detail": "empty image_url for jimeng i2v",
+                "provider": "jimeng",
+            }
+            return None
+        # 即梦 i2v 无声视频，generate_audio 强制 False
+        jm.last_video_error = {}
+        tag = "[portrait-fallback→jimeng]" if _portrait_fallback else "[jimeng-direct]"
+        logger.info("%s 调用即梦: dur=%s ratio=%s res=%s img=%s", tag, duration, ratio, resolution, bool(image_url))
+        try:
+            result = jm.video_generation(
+                prompt=prompt,
+                image_url=image_url,
+                duration=duration,
+                ratio=ratio,
+                resolution=resolution,
+                output_dir=output_dir,
+                generate_audio=False,
+            )
+        except Exception as je:
+            logger.error("%s 即梦 video_generation 异常: %s", tag, je, exc_info=True)
+            self.last_video_error = {
+                "error_code": "unknown",
+                "user_message": f"即梦视频生成异常：{je!s}"[:200],
+                "status_code": 0,
+                "detail": str(je),
+                "provider": "jimeng",
+            }
+            return None
+        if not result and jm.last_video_error:
+            # 透传即梦错误；如果即梦也返回 portrait_intercept，说明图片真的有问题，直接给用户
+            jm_err = dict(jm.last_video_error)
+            jm_err["provider"] = "jimeng"
+            if _portrait_fallback and jm_err.get("error_code") == "portrait_intercept":
+                jm_err["user_message"] = (
+                    "参考素材真人肖像审核未通过（方舟+即梦双通道均被拦截），请更换非真人或授权清晰的照片后重试。"
+                )
+            self.last_video_error = jm_err
+            return None
+        if result:
+            # 补充 usage 里的 provider 标记
+            u = result.get("usage") or {}
+            u.setdefault("provider", "jimeng")
+            u.setdefault("model_key", "jimeng-3.0")
+            result["usage"] = u
+            logger.info("%s 即梦生成成功: %s", tag, result.get("video_path"))
+        return result
+
     def get_last_video_error(self) -> dict:
         """返回最近一次 video_generation 失败的详细错误。空 dict 表示上次成功或未调用。"""
         return dict(self.last_video_error or {})

packages/shared/jimeng_client.py

@@ -0,0 +1,526 @@
+"""即梦（Jimeng）视觉 API 客户端 —— 火山引擎 cvtob。
+
+#2169: 真人参考图在方舟 Seedance 走 B 端审核会被 50411 拦截，
+即梦走 C 端审核链路，普通真人照片可过审。接入即梦 i2v 作为参考图场景兜底通道。
+
+接口协议（jimeng_i2v_first_v30 —— 视频3.0 720P 首帧图生视频）：
+- 接口地址：https://visual.volcengineapi.com
+- 鉴权：火山 V4 签名（Region=cn-north-1, Service=cv），使用 AK/SK
+- 提交任务：POST ?Action=CVSync2AsyncSubmitTask&Version=2022-08-31
+  body: {"req_key": "jimeng_i2v_first_v30", "image_urls": ["<url>"], "prompt": "...", "seed": -1, "frames": 121}
+  -> {"code": 10000, "data": {"task_id": "..."}}
+- 查询任务：POST ?Action=CVSync2AsyncGetResult&Version=2022-08-31
+  body: {"req_key": "jimeng_i2v_first_v30", "task_id": "..."}
+  -> {"code": 10000, "data": {"status": "in_queue|generating|done", "video_url": "..."}}
+- 视频 URL 有效期 1 小时，必须立即下载到本地。
+"""
+
+from __future__ import annotations
+
+import hashlib
+import hmac
+import json
+import logging
+import os
+import time
+import uuid
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+from urllib.parse import quote, urlparse
+
+import httpx
+
+from packages.shared.config import get_shared_settings
+
+# 网络/超时类异常父类集合
+_HTTP_NETWORK_ERRORS = ()
+try:
+    _HTTP_NETWORK_ERRORS = (httpx.TimeoutException, httpx.NetworkError)
+except Exception:
+    _HTTP_NETWORK_ERRORS = (Exception,)
+
+logger = logging.getLogger(__name__)
+
+_JIMENG_CLIENT_SINGLETON: "JimengClient | None" = None
+
+# ── V4 签名常量 ────────────────────────────────────────────────────────
+_JIMENG_REGION = "cn-north-1"
+_JIMENG_SERVICE = "cv"
+_JIMENG_VERSION = "2022-08-31"
+_ACTION_SUBMIT = "CVSync2AsyncSubmitTask"
+_ACTION_POLL = "CVSync2AsyncGetResult"
+_CONTENT_TYPE = "application/json"
+_SIGNED_HEADERS_LIST = ["content-type", "host", "x-content-sha256", "x-date"]
+_SIGNED_HEADERS_STR = ";".join(_SIGNED_HEADERS_LIST)
+
+
+def _norm_query(params: dict[str, str]) -> str:
+    """构造规范查询串：按 key 排序，URL 编码（safe=-_.~），空格->%20。"""
+    parts = []
+    for k in sorted(params.keys()):
+        v = params[k]
+        ek = quote(str(k), safe="-_.~")
+        ev = quote(str(v), safe="-_.~").replace("+", "%20")
+        parts.append(f"{ek}={ev}")
+    return "&".join(parts)
+
+
+def _hmac_sha256(key: bytes, msg: str) -> bytes:
+    return hmac.new(key, msg.encode("utf-8"), hashlib.sha256).digest()
+
+
+def _sha256_hex(data: bytes) -> str:
+    return hashlib.sha256(data).hexdigest()
+
+
+def _sign_v4(
+    ak: str,
+    sk: str,
+    method: str,
+    host: str,
+    query: dict[str, str],
+    body_bytes: bytes,
+    x_date: str,
+) -> dict[str, str]:
+    """火山 V4 签名，返回需要附加到请求的 headers 字典。
+
+    x_date 形如 "20260101T120000Z"（UTC）。
+    short_date = x_date[:8]（YYYYMMDD）。
+    """
+    short_date = x_date[:8]
+    payload_hash = _sha256_hex(body_bytes)
+    canon_uri = "/"
+    canon_query = _norm_query(query)
+    canon_headers = f"content-type:{_CONTENT_TYPE}\nhost:{host}\nx-content-sha256:{payload_hash}\nx-date:{x_date}\n"
+    canon_request = f"{method}\n{canon_uri}\n{canon_query}\n{canon_headers}\n{_SIGNED_HEADERS_STR}\n{payload_hash}"
+    credential_scope = f"{short_date}/{_JIMENG_REGION}/{_JIMENG_SERVICE}/request"
+    string_to_sign = f"HMAC-SHA256\n{x_date}\n{credential_scope}\n{_sha256_hex(canon_request.encode('utf-8'))}"
+    k_date = _hmac_sha256(sk.encode("utf-8"), short_date)
+    k_region = _hmac_sha256(k_date, _JIMENG_REGION)
+    k_service = _hmac_sha256(k_region, _JIMENG_SERVICE)
+    k_signing = _hmac_sha256(k_service, "request")
+    signature = hmac.new(k_signing, string_to_sign.encode("utf-8"), hashlib.sha256).hexdigest()
+    authorization = (
+        f"HMAC-SHA256 Credential={ak}/{credential_scope}, SignedHeaders={_SIGNED_HEADERS_STR}, Signature={signature}"
+    )
+    return {
+        "Content-Type": _CONTENT_TYPE,
+        "Host": host,
+        "X-Content-Sha256": payload_hash,
+        "X-Date": x_date,
+        "Authorization": authorization,
+    }
+
+
+# ── 错误分类 ──────────────────────────────────────────────────────────
+
+# 即梦业务码 -> 是否可重试映射
+_JIMENG_RETRYABLE_CODES = {50511, 50516, 50429, 50430, 50500, 50501}
+_JIMENG_NON_RETRYABLE_CODES = {50411, 50412, 50413, 50512, 50513, 50514}
+
+
+def _classify_jimeng_error(status_code: int, body: str, biz_code: int | None = None) -> tuple[str, str, bool]:
+    """即梦错误分类，返回 (error_code, user_message, is_retryable)。"""
+    code = biz_code if biz_code is not None else 0
+    body_lower = (body or "").lower()
+
+    # 业务码优先
+    if code == 50411:
+        return (
+            "portrait_intercept",
+            "即梦通道：参考图片前审核未通过（Pre Img Risk Not Pass），请更换参考图后重试。",
+            False,
+        )
+    if code == 50511:
+        return "task_failed", "即梦通道：输出图片后审核未通过，可稍后重试。", True
+    if code in (50412, 50413, 50512):
+        return "invalid_param", "即梦通道：提示词或文本审核不通过，请调整文案后重试。", False
+    if code == 50516:
+        return "task_failed", "即梦通道：输出视频后审核未通过，可稍后重试。", True
+    if code in (50429, 50430):
+        return "rate_limit", "即梦通道：QPS/并发超限，请稍等 1-2 分钟后重试。", True
+    if code in (50500, 50501):
+        return "network_error", "即梦通道：服务内部错误，可稍后重试。", True
+
+    # HTTP 层兜底
+    if status_code in (401, 403):
+        return "auth_error", "即梦通道：AK/SK 鉴权失败，请联系管理员检查 JIMENG_AK/SK 配置。", False
+    if status_code == 429:
+        return "rate_limit", "即梦通道：服务限流，请稍后重试。", True
+    if status_code == 404:
+        return "model_not_found", "即梦通道：接口不存在（req_key 或 Action 错误），请联系管理员。", False
+    if status_code in (402, 400) and any(kw in body_lower for kw in ("quota", "billing", "insufficient", "余额")):
+        return "quota_exceeded", "即梦通道：账户余额/配额不足，请联系管理员充值。", False
+    if status_code == 400:
+        msg = ""
+        try:
+            msg = str(json.loads(body or "{}").get("message", "") or "")
+        except Exception:
+            pass
+        return "invalid_param", f"即梦通道：参数错误：{msg or body[:200]}", False
+    if status_code == 0:
+        return "network_error", "即梦通道：网络连接失败，请稍后重试。", True
+    # 任务内失败
+    if code and code != 10000:
+        return "unknown", f"即梦通道：视频生成失败（错误码 {code}），请稍后重试。", code in _JIMENG_RETRYABLE_CODES
+    detail = body[:200]
+    return "unknown", f"即梦通道：视频生成失败（HTTP {status_code}）：{detail}", False
+
+
+# ── 即梦客户端 ────────────────────────────────────────────────────────
+
+
+class JimengClient:
+    """火山引擎即梦视觉 API（cvtob）异步客户端，支持图生视频首帧（jimeng_i2v_first_v30）。"""
+
+    def __init__(self) -> None:
+        settings = get_shared_settings()
+        self.ak: str = getattr(settings, "jimeng_ak", "") or os.getenv("JIMENG_AK", "")
+        self.sk: str = getattr(settings, "jimeng_sk", "") or os.getenv("JIMENG_SK", "")
+        self.base_url: str = (getattr(settings, "jimeng_base_url", "") or "https://visual.volcengineapi.com").rstrip(
+            "/"
+        )
+        self.req_key: str = getattr(settings, "jimeng_req_key", "") or "jimeng_i2v_first_v30"
+        self.poll_interval: int = int(getattr(settings, "jimeng_video_poll_interval", 5) or 5)
+        self.total_timeout: int = int(getattr(settings, "jimeng_video_timeout", 600) or 600)
+        self.max_retries: int = 2
+        self.last_video_error: dict = {}
+        # 解析 base_url 里的 host（用于签名 Host 头）
+        parsed = urlparse(self.base_url)
+        self.host: str = parsed.netloc or "visual.volcengineapi.com"
+
+    @property
+    def is_available(self) -> bool:
+        return bool(self.ak and self.sk)
+
+    def get_last_video_error(self) -> dict:
+        return dict(self.last_video_error or {})
+
+    def _set_error(self, error_code: str, user_message: str, status_code: int = 0, detail: str = "", **extra) -> None:
+        self.last_video_error = {
+            "error_code": error_code,
+            "user_message": user_message,
+            "status_code": status_code,
+            "detail": detail[:500] if detail else "",
+            "provider": "jimeng",
+            **extra,
+        }
+
+    # ── 内部 HTTP：签名 + 请求 ──────────────────────────────────────
+
+    def _signed_request(
+        self,
+        method: str,
+        action: str,
+        body_obj: dict[str, Any],
+        timeout: float = 60.0,
+    ) -> tuple[int, str, dict]:
+        """发送一次带 V4 签名的请求，返回 (status_code, body_text, parsed_json)。"""
+        body_bytes = json.dumps(body_obj, ensure_ascii=False).encode("utf-8")
+        query = {"Action": action, "Version": _JIMENG_VERSION}
+        x_date = datetime.now(timezone.utc).strftime("%Y%m%dT%H%M%SZ")
+        headers = _sign_v4(self.ak, self.sk, method, self.host, query, body_bytes, x_date)
+        url = f"{self.base_url}/?{_norm_query(query)}"
+        resp = httpx.request(
+            method,
+            url,
+            headers=headers,
+            content=body_bytes,
+            timeout=timeout,
+        )
+        sc = int(getattr(resp, "status_code", 0) or 0)
+        text = getattr(resp, "text", "") or ""
+        try:
+            data = resp.json()
+        except Exception:
+            data = {}
+        return sc, text, data
+
+    # ── 提交任务 ────────────────────────────────────────────────────
+
+    def _submit_task(
+        self,
+        prompt: str,
+        image_url: str,
+        frames: int = 121,
+        seed: int = -1,
+    ) -> str | None:
+        """提交图生视频任务，成功返回 task_id；失败写 last_video_error 并返回 None。"""
+        body: dict[str, Any] = {
+            "req_key": self.req_key,
+            "prompt": prompt.strip()[:800],
+            "image_urls": [image_url],
+            "seed": int(seed) if seed and seed > 0 else -1,
+            "frames": int(frames),
+        }
+        last_sc = 0
+        last_body = ""
+        for attempt in range(self.max_retries + 1):
+            try:
+                sc, text, data = self._signed_request("POST", _ACTION_SUBMIT, body, timeout=60.0)
+                last_sc, last_body = sc, text
+                if sc >= 400:
+                    logger.error("[jimeng] 提交 HTTP %d: %s", sc, text[:500])
+                    if sc >= 500 and attempt < self.max_retries:
+                        time.sleep(0.8 * (2**attempt))
+                        continue
+                    biz_code = data.get("code") if isinstance(data, dict) else None
+                    err_code, user_msg, _ = _classify_jimeng_error(sc, text, biz_code)
+                    self._set_error(err_code, user_msg, sc, text, req_key=self.req_key)
+                    return None
+                code = data.get("code") if isinstance(data, dict) else None
+                if code == 10000:
+                    d = data.get("data") or {}
+                    tid = d.get("task_id")
+                    if tid:
+                        return str(tid)
+                err_code, user_msg, retry = _classify_jimeng_error(sc, text, code)
+                logger.error(
+                    "[jimeng] 提交业务错误 code=%s msg=%s",
+                    code,
+                    (data.get("message") if isinstance(data, dict) else ""),
+                )
+                if retry and attempt < self.max_retries:
+                    time.sleep(0.8 * (2**attempt))
+                    continue
+                self._set_error(err_code, user_msg, sc, text, req_key=self.req_key, biz_code=code)
+                return None
+            except _HTTP_NETWORK_ERRORS as ne:
+                last_sc, last_body = 0, f"network error: {ne}"
+                logger.warning(
+                    "[jimeng] 提交网络异常 %s，重试 %d/%d", type(ne).__name__, attempt + 1, self.max_retries + 1
+                )
+                if attempt < self.max_retries:
+                    time.sleep(0.8 * (2**attempt))
+                    continue
+                self._set_error("network_error", "即梦通道：提交任务网络异常，请稍后重试。", 0, str(ne))
+                return None
+            except Exception as e:
+                last_sc, last_body = 0, f"exception: {e}"
+                logger.error("[jimeng] 提交异常: %s", e, exc_info=True)
+                if attempt < self.max_retries:
+                    time.sleep(0.8 * (2**attempt))
+                    continue
+                self._set_error("unknown", f"即梦通道：提交任务异常：{e!s}"[:200], 0, str(e))
+                return None
+        if not self.last_video_error:
+            err_code, user_msg, _ = _classify_jimeng_error(last_sc, last_body)
+            self._set_error(err_code, user_msg, last_sc, last_body)
+        return None
+
+    # ── 轮询结果 ────────────────────────────────────────────────────
+
+    def _poll_result(self, task_id: str) -> str | None:
+        """轮询任务直到 done/failed/expired/timeout，成功返回 video_url。"""
+        deadline = time.time() + self.total_timeout
+        poll_count = 0
+        last_status = ""
+        poll_body = {"req_key": self.req_key, "task_id": task_id}
+        while time.time() < deadline:
+            poll_count += 1
+            try:
+                sc, text, data = self._signed_request("POST", _ACTION_POLL, poll_body, timeout=30.0)
+                if sc >= 400:
+                    logger.warning("[jimeng] 轮询 HTTP %d: %s", sc, text[:300])
+                    if poll_count < 3:
+                        time.sleep(self.poll_interval)
+                        continue
+                    err_code, user_msg, _ = _classify_jimeng_error(sc, text)
+                    self._set_error(err_code, user_msg, sc, text, task_id=task_id)
+                    return None
+                code = data.get("code") if isinstance(data, dict) else None
+                d = data.get("data") if isinstance(data, dict) else None
+                if code != 10000 or not isinstance(d, dict):
+                    err_code, user_msg, retry = _classify_jimeng_error(sc, text, code)
+                    logger.error(
+                        "[jimeng] 轮询业务错误 task=%s code=%s msg=%s",
+                        task_id,
+                        code,
+                        (data.get("message") if isinstance(data, dict) else ""),
+                    )
+                    if retry and poll_count < 3:
+                        time.sleep(self.poll_interval)
+                        continue
+                    self._set_error(err_code, user_msg, sc, text, task_id=task_id, biz_code=code)
+                    return None
+                status = d.get("status", "") or ""
+                last_status = status
+                if status == "done":
+                    video_url = d.get("video_url") or ""
+                    if video_url:
+                        logger.info("[jimeng] 任务 %s 完成 polls=%d", task_id, poll_count)
+                        return str(video_url)
+                    logger.error("[jimeng] 任务 %s done 但无 video_url: %s", task_id, str(d)[:500])
+                    self._set_error(
+                        "unknown",
+                        "即梦通道：任务成功但未返回视频URL，请联系管理员。",
+                        200,
+                        str(d)[:500],
+                        task_id=task_id,
+                    )
+                    return None
+                if status in ("not_found", "expired"):
+                    logger.error("[jimeng] 任务 %s 状态 %s", task_id, status)
+                    self._set_error(
+                        "network_error" if status == "expired" else "unknown",
+                        f"即梦通道：任务{'已过期' if status == 'expired' else '未找到'}，请重新提交。",
+                        200,
+                        f"task {status}",
+                        task_id=task_id,
+                    )
+                    return None
+                if poll_count % 6 == 0:
+                    logger.info("[jimeng] 轮询中 task=%s status=%s polls=%d", task_id, status, poll_count)
+            except _HTTP_NETWORK_ERRORS as ne:
+                logger.warning("[jimeng] 轮询网络异常 %s", ne)
+            except Exception as e:
+                logger.debug("[jimeng] 轮询异常: %s", e)
+            time.sleep(self.poll_interval)
+        logger.error(
+            "[jimeng] 任务 %s 轮询超时（%ds）polls=%d last_status=%s",
+            task_id,
+            self.total_timeout,
+            poll_count,
+            last_status,
+        )
+        self._set_error(
+            "network_error",
+            f"即梦通道：视频生成超时（>{self.total_timeout}s），任务仍在排队，请稍后重试。",
+            0,
+            f"timeout after {self.total_timeout}s, polls={poll_count}, last_status={last_status}",
+            task_id=task_id,
+            last_status=last_status,
+        )
+        return None
+
+    # ── 下载视频 ────────────────────────────────────────────────────
+
+    def _download_video(self, video_url: str, output_dir: str, task_id: str) -> str | None:
+        os.makedirs(output_dir, exist_ok=True)
+        suffix = Path(urlparse(video_url).path).suffix or ".mp4"
+        if suffix.lower() not in (".mp4", ".mov", ".webm"):
+            suffix = ".mp4"
+        safe_tid = "".join(c if c.isalnum() or c in "-_" else "_" for c in task_id)[:40]
+        out_path = os.path.join(output_dir, f"jimeng_{safe_tid}_{uuid.uuid4().hex[:8]}{suffix}")
+        try:
+            with httpx.stream("GET", video_url, timeout=180, follow_redirects=True) as resp:
+                dsc = int(getattr(resp, "status_code", 0) or 0)
+                if dsc >= 400:
+                    logger.error("[jimeng] 下载 HTTP %d", dsc)
+                    self._set_error("network_error", "即梦通道：视频下载失败（HTTP错误），请稍后重试。", dsc)
+                    return None
+                with open(out_path, "wb") as f:
+                    for chunk in resp.iter_bytes(chunk_size=1024 * 256):
+                        if chunk:
+                            f.write(chunk)
+        except Exception as e:
+            logger.error("[jimeng] 下载视频失败: %s", e, exc_info=True)
+            self._set_error("network_error", f"即梦通道：视频下载失败：{e!s}"[:200], 0, str(e))
+            return None
+        size = os.path.getsize(out_path) if os.path.exists(out_path) else 0
+        if size < 1024:
+            logger.error("[jimeng] 下载文件过小: %d bytes", size)
+            self._set_error("unknown", "即梦通道：视频下载文件过小，请稍后重试。", 0, f"downloaded only {size} bytes")
+            try:
+                os.remove(out_path)
+            except Exception:
+                pass
+            return None
+        logger.info("[jimeng] 视频已下载: %s (%d bytes)", out_path, size)
+        return out_path
+
+    # ── 对外主入口 ──────────────────────────────────────────────────
+
+    def video_generation(
+        self,
+        prompt: str,
+        *,
+        image_url: str,
+        duration: int = 5,
+        ratio: str | None = "9:16",
+        resolution: str = "720p",
+        output_dir: str | None = None,
+        generate_audio: bool = False,
+    ) -> dict | None:
+        """即梦图生视频主入口。
+
+        成功返回 {"video_path": str, "usage": {"provider","duration_seconds","frames","req_key","billing_mode"}}；
+        失败返回 None，详情在 self.last_video_error。
+
+        注意：jimeng_i2v_first_v30 不支持原生音频（generate_audio 被忽略，返回无声视频），
+        音频由后续 ffmpeg 合成阶段叠加 TTS。
+        支持时长：5s(frames=121)/10s(frames=241)，>10s 截断并打 warning。
+        分辨率固定 720P；ratio 对首帧 i2v 无效（自动按图片比例）。
+        """
+        self.last_video_error = {}
+        if not self.is_available:
+            self._set_error(
+                "auth_error",
+                "即梦通道未配置（JIMENG_AK/SK 缺失），请联系管理员。",
+                detail="jimeng ak/sk empty",
+            )
+            logger.error("[jimeng] AK/SK 未配置，无法调用")
+            return None
+        if not prompt or not prompt.strip():
+            self._set_error("invalid_param", "视频生成提示词不能为空。", detail="empty prompt")
+            return None
+        if not image_url or not image_url.strip():
+            self._set_error("invalid_param", "即梦图生视频必须提供参考图片。", detail="empty image_url")
+            return None
+
+        dur = int(duration or 5)
+        if dur <= 5:
+            frames = 121
+            real_dur = 5
+        elif dur <= 10:
+            frames = 241
+            real_dur = 10
+        else:
+            logger.warning("[jimeng] 请求时长 %ds 超出即梦 i2v 上限 10s，截断到 10s（frames=241）", dur)
+            frames = 241
+            real_dur = 10
+
+        out_dir = output_dir or "/tmp"
+        logger.info(
+            "[jimeng] 提交任务: req_key=%s dur=%ds(frames=%d) ratio=%s res=%s gen_audio=%s img=%s",
+            self.req_key,
+            real_dur,
+            frames,
+            ratio,
+            resolution,
+            generate_audio,
+            bool(image_url),
+        )
+
+        task_id = self._submit_task(prompt=prompt, image_url=image_url, frames=frames, seed=-1)
+        if not task_id:
+            return None
+        logger.info("[jimeng] 任务已提交: task_id=%s", task_id)
+
+        video_url = self._poll_result(task_id)
+        if not video_url:
+            return None
+
+        local_path = self._download_video(video_url, out_dir, task_id)
+        if not local_path:
+            return None
+
+        usage = {
+            "provider": "jimeng",
+            "duration_seconds": real_dur,
+            "frames": frames,
+            "req_key": self.req_key,
+            "billing_mode": "per_second",
+        }
+        return {"video_path": local_path, "usage": usage}
+
+
+def get_jimeng_client() -> "JimengClient | None":
+    """返回即梦客户端单例；未配置 AK/SK 时返回 None。"""
+    global _JIMENG_CLIENT_SINGLETON
+    if _JIMENG_CLIENT_SINGLETON is None:
+        _JIMENG_CLIENT_SINGLETON = JimengClient()
+    if not _JIMENG_CLIENT_SINGLETON.is_available:
+        return None
+    return _JIMENG_CLIENT_SINGLETON

tests/unit/test_ai_client_video.py

@@ -563,7 +563,7 @@ class TestResolveVideoModelId:
 
 class TestVideoGenerationLastError:
     def test_create_400_portrait_returns_user_message(self, tmp_path):
-        """HTTP 400 + 真人拦截关键词 → portrait_intercept 错误码，用户提示友好。"""
+        """#2169: HTTP 400 + 真人拦截关键词 → 自动尝试即梦兜底；即梦未配时返回 portrait_intercept。"""
         client = _make_client(max_retries=0)
         create_resp = MagicMock()
         create_resp.status_code = 400
@@ -576,6 +576,8 @@ class TestVideoGenerationLastError:
             patch("packages.shared.ai_client.httpx.post", return_value=create_resp),
             patch("packages.shared.ai_client.time.sleep", return_value=None),
             patch("packages.shared.ai_client.get_shared_settings") as mock_s,
+            # jimeng 未配置，fallback 后仍返回 portrait_intercept（提示用户需要配置即梦）
+            patch("packages.shared.jimeng_client.get_jimeng_client", return_value=None),
         ):
             mock_s.return_value = MagicMock(
                 doubao_video_poll_interval=0, doubao_video_timeout=1, doubao_video_model="seedance"
@@ -584,8 +586,10 @@ class TestVideoGenerationLastError:
             assert result is None
             err = client.get_last_video_error()
             assert err["error_code"] == "portrait_intercept"
-            assert "真人" in err["user_message"]
-            assert err["status_code"] == 400
+            # 即梦兜底未启用时提示包含"真人照片"/"即梦"等关键字
+            assert "真人" in err["user_message"] or "肖像" in err["user_message"] or "即梦" in err["user_message"]
+            # 方舟本身 status_code=400（最后一个错误可能被即梦兜底覆盖，但 error_code 不变）
+            assert err["status_code"] in (0, 400)
 
     def test_create_401_returns_auth_error(self, tmp_path):
         client = _make_client(max_retries=0)

tests/unit/test_jimeng_client.py

@@ -0,0 +1,376 @@
+"""tests for packages/shared/jimeng_client.py (#2169 即梦 i2v 客户端)."""
+
+from __future__ import annotations
+
+import json
+import os
+from unittest.mock import MagicMock, mock_open, patch
+
+import pytest
+
+
+@pytest.fixture(autouse=True)
+def reset_singleton():
+    import packages.shared.jimeng_client as j
+
+    j._JIMENG_CLIENT_SINGLETON = None
+    yield
+    j._JIMENG_CLIENT_SINGLETON = None
+
+
+def _make_settings(ak="test-ak", sk="test-sk", req_key="jimeng_i2v_first_v30", timeout=60, poll_interval=2):
+    return MagicMock(
+        jimeng_ak=ak,
+        jimeng_sk=sk,
+        jimeng_base_url="https://visual.volcengineapi.com",
+        jimeng_req_key=req_key,
+        jimeng_video_timeout=timeout,
+        jimeng_video_poll_interval=poll_interval,
+    )
+
+
+# ── V4 签名单元测试 ──────────────────────────────────────────────────
+
+
+class TestV4Signature:
+    def test_sign_returns_required_headers(self):
+        from packages.shared.jimeng_client import _sign_v4
+
+        headers = _sign_v4(
+            ak="AK_TEST",
+            sk="SK_TEST",
+            method="POST",
+            host="visual.volcengineapi.com",
+            query={"Action": "CVSync2AsyncSubmitTask", "Version": "2022-08-31"},
+            body_bytes=b'{"req_key":"jimeng_i2v_first_v30"}',
+            x_date="20260101T120000Z",
+        )
+        assert headers["Content-Type"] == "application/json"
+        assert headers["Host"] == "visual.volcengineapi.com"
+        assert headers["X-Date"] == "20260101T120000Z"
+        assert "X-Content-Sha256" in headers
+        assert headers["Authorization"].startswith("HMAC-SHA256 Credential=AK_TEST/20260101/cn-north-1/cv/request")
+        assert "SignedHeaders=content-type;host;x-content-sha256;x-date" in headers["Authorization"]
+        assert "Signature=" in headers["Authorization"]
+        # 签名是 64 字符 hex
+        sig = headers["Authorization"].split("Signature=")[-1]
+        assert len(sig) == 64
+        assert all(c in "0123456789abcdef" for c in sig)
+
+    def test_sign_deterministic(self):
+        """相同输入必须产生相同签名（幂等）。"""
+        from packages.shared.jimeng_client import _sign_v4
+
+        kwargs = dict(
+            ak="AK",
+            sk="SK",
+            method="POST",
+            host="h",
+            query={"A": "1", "B": "2"},
+            body_bytes=b"{}",
+            x_date="20260101T000000Z",
+        )
+        h1 = _sign_v4(**kwargs)
+        h2 = _sign_v4(**kwargs)
+        assert h1["Authorization"] == h2["Authorization"]
+        assert h1["X-Content-Sha256"] == h2["X-Content-Sha256"]
+
+    def test_sign_different_body_different_sig(self):
+        from packages.shared.jimeng_client import _sign_v4
+
+        base = dict(ak="AK", sk="SK", method="POST", host="h", query={}, x_date="20260101T000000Z")
+        h1 = _sign_v4(body_bytes=b"a", **base)
+        h2 = _sign_v4(body_bytes=b"b", **base)
+        assert h1["Authorization"] != h2["Authorization"]
+
+    def test_payload_sha256_matches(self):
+        import hashlib
+
+        from packages.shared.jimeng_client import _sign_v4
+
+        body = b'{"prompt":"hello"}'
+        h = _sign_v4("ak", "sk", "POST", "h", {}, body, "20260101T000000Z")
+        expected = hashlib.sha256(body).hexdigest()
+        assert h["X-Content-Sha256"] == expected
+
+    def test_norm_query_sorted_and_encoded(self):
+        from packages.shared.jimeng_client import _norm_query
+
+        q = _norm_query({"B": "2", "A": "1", "C": "a b"})
+        # key 排序 + 空格→%20
+        assert q.startswith("A=1")
+        assert "B=2" in q
+        assert "C=a%20b" in q
+
+
+# ── 可用性 / 单例 ─────────────────────────────────────────────────────
+
+
+class TestAvailability:
+    def test_unavailable_without_ak_sk(self):
+        from packages.shared.jimeng_client import JimengClient, get_jimeng_client
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings(ak="", sk="")
+            # 重置单例
+            import packages.shared.jimeng_client as j
+
+            j._JIMENG_CLIENT_SINGLETON = None
+            assert get_jimeng_client() is None
+
+    def test_available_with_ak_sk(self):
+        from packages.shared.jimeng_client import get_jimeng_client
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings()
+            import packages.shared.jimeng_client as j
+
+            j._JIMENG_CLIENT_SINGLETON = None
+            c = get_jimeng_client()
+            assert c is not None
+            assert c.is_available is True
+            assert c.req_key == "jimeng_i2v_first_v30"
+
+
+# ── 错误分类 ────────────────────────────────────────────────────────
+
+
+class TestClassifyError:
+    def test_50411_is_portrait_intercept_non_retryable(self):
+        from packages.shared.jimeng_client import _classify_jimeng_error
+
+        code, msg, retry = _classify_jimeng_error(200, '{"code":50411,"message":"Pre Img Risk"}', 50411)
+        assert code == "portrait_intercept"
+        assert retry is False
+
+    def test_50429_is_rate_limit_retryable(self):
+        from packages.shared.jimeng_client import _classify_jimeng_error
+
+        code, msg, retry = _classify_jimeng_error(200, "", 50429)
+        assert code == "rate_limit"
+        assert retry is True
+
+    def test_50430_is_rate_limit(self):
+        from packages.shared.jimeng_client import _classify_jimeng_error
+
+        code, _, _ = _classify_jimeng_error(200, "", 50430)
+        assert code == "rate_limit"
+
+    def test_50500_is_network_error_retryable(self):
+        from packages.shared.jimeng_client import _classify_jimeng_error
+
+        code, _, retry = _classify_jimeng_error(200, "", 50500)
+        assert code == "network_error"
+        assert retry is True
+
+    def test_50412_is_invalid_param_non_retryable(self):
+        from packages.shared.jimeng_client import _classify_jimeng_error
+
+        code, _, retry = _classify_jimeng_error(200, "", 50412)
+        assert code == "invalid_param"
+        assert retry is False
+
+    def test_401_auth(self):
+        from packages.shared.jimeng_client import _classify_jimeng_error
+
+        code, msg, retry = _classify_jimeng_error(401, "auth fail", None)
+        assert code == "auth_error"
+        assert retry is False
+
+    def test_400_text_audit(self):
+        from packages.shared.jimeng_client import _classify_jimeng_error
+
+        code, _, _ = _classify_jimeng_error(400, "text error", None)
+        assert code == "invalid_param"
+
+
+# ── video_generation 主流程 ──────────────────────────────────────────
+
+
+class TestVideoGenerationHappyPath:
+    def test_missing_ak_returns_none(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings(ak="", sk="")
+            c = JimengClient()
+        assert c.video_generation("hi", image_url="http://x/y.jpg") is None
+        err = c.last_video_error
+        assert err["error_code"] == "auth_error"
+
+    def test_empty_prompt_returns_none(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings()
+            c = JimengClient()
+        assert c.video_generation("   ", image_url="http://x/y.jpg") is None
+        assert c.last_video_error["error_code"] == "invalid_param"
+
+    def test_empty_image_url_returns_none(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings()
+            c = JimengClient()
+        assert c.video_generation("prompt", image_url="") is None
+        assert c.last_video_error["error_code"] == "invalid_param"
+
+    def test_duration_5s_frames_121(self):
+        """5s → frames=121，10s→frames=241，>10s 截断到10s。"""
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings(timeout=1, poll_interval=0)
+            c = JimengClient()
+
+        captured = {}
+
+        def fake_submit(prompt, image_url, frames, seed=-1):
+            captured["frames"] = frames
+            return "task-xyz"
+
+        def fake_poll(tid):
+            captured["tid"] = tid
+            return "http://example.com/v.mp4"
+
+        def fake_download(url, out_dir, tid):
+            captured["url"] = url
+            return "/tmp/fake.mp4"
+
+        # 构造一个假文件
+        os.makedirs("/tmp", exist_ok=True)
+        with open("/tmp/fake.mp4", "wb") as f:
+            f.write(b"x" * 2048)
+
+        with (
+            patch.object(c, "_submit_task", side_effect=fake_submit),
+            patch.object(c, "_poll_result", side_effect=fake_poll),
+            patch.object(c, "_download_video", side_effect=fake_download),
+        ):
+            r = c.video_generation("test", image_url="http://x/y.jpg", duration=5, output_dir="/tmp")
+        assert r is not None
+        assert captured["frames"] == 121
+        assert r["usage"]["duration_seconds"] == 5
+        assert r["usage"]["billing_mode"] == "per_second"
+        assert r["usage"]["req_key"] == "jimeng_i2v_first_v30"
+
+    def test_duration_10s_frames_241(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings(timeout=1, poll_interval=0)
+            c = JimengClient()
+        captured = {}
+
+        def fake_submit(prompt, image_url, frames, seed=-1):
+            captured["frames"] = frames
+            return "tid"
+
+        def fake_poll(tid):
+            return "http://x/v.mp4"
+
+        def fake_download(url, out_dir, tid):
+            with open("/tmp/fake2.mp4", "wb") as f:
+                f.write(b"x" * 2048)
+            return "/tmp/fake2.mp4"
+
+        with (
+            patch.object(c, "_submit_task", side_effect=fake_submit),
+            patch.object(c, "_poll_result", side_effect=fake_poll),
+            patch.object(c, "_download_video", side_effect=fake_download),
+        ):
+            r = c.video_generation("hi", image_url="http://x/y.jpg", duration=10, output_dir="/tmp")
+        assert captured["frames"] == 241
+        assert r["usage"]["duration_seconds"] == 10
+
+    def test_duration_over_10s_truncates_to_10s(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings(timeout=1, poll_interval=0)
+            c = JimengClient()
+        captured = {}
+
+        def fake_submit(prompt, image_url, frames, seed=-1):
+            captured["frames"] = frames
+            return "tid"
+
+        def fake_poll(tid):
+            return "http://x/v.mp4"
+
+        def fake_download(url, out_dir, tid):
+            with open("/tmp/fake3.mp4", "wb") as f:
+                f.write(b"x" * 2048)
+            return "/tmp/fake3.mp4"
+
+        with (
+            patch.object(c, "_submit_task", side_effect=fake_submit),
+            patch.object(c, "_poll_result", side_effect=fake_poll),
+            patch.object(c, "_download_video", side_effect=fake_download),
+        ):
+            r = c.video_generation("hi", image_url="http://x/y.jpg", duration=30, output_dir="/tmp")
+        assert captured["frames"] == 241
+        assert r["usage"]["duration_seconds"] == 10
+
+
+class TestSubmitTaskErrors:
+    def test_submit_50411_writes_portrait_intercept(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings()
+            c = JimengClient()
+        fake_resp = MagicMock(status_code=200, text='{"code":50411,"message":"Pre Img Risk Not Pass"}')
+        fake_resp.json.return_value = {"code": 50411, "message": "Pre Img Risk Not Pass"}
+        with patch("packages.shared.jimeng_client.httpx.request", return_value=fake_resp):
+            tid = c._submit_task("p", "http://x/y.jpg", frames=121)
+        assert tid is None
+        assert c.last_video_error["error_code"] == "portrait_intercept"
+
+    def test_submit_returns_task_id(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings()
+            c = JimengClient()
+        fake_resp = MagicMock(status_code=200, text='{"code":10000,"data":{"task_id":"abc"}}')
+        fake_resp.json.return_value = {"code": 10000, "data": {"task_id": "abc"}}
+        with patch("packages.shared.jimeng_client.httpx.request", return_value=fake_resp):
+            tid = c._submit_task("p", "http://x/y.jpg", frames=121)
+        assert tid == "abc"
+
+
+class TestPollResult:
+    def test_poll_done_returns_video_url(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings(timeout=10, poll_interval=0)
+            c = JimengClient()
+        done_resp = MagicMock(status_code=200)
+        done_resp.json.return_value = {"code": 10000, "data": {"status": "done", "video_url": "http://x/v.mp4"}}
+        with (
+            patch("packages.shared.jimeng_client.httpx.request", return_value=done_resp),
+            patch("packages.shared.jimeng_client.time.sleep"),
+        ):
+            url = c._poll_result("abc")
+        assert url == "http://x/v.mp4"
+
+    def test_poll_timeout_returns_none(self):
+        from packages.shared.jimeng_client import JimengClient
+
+        with patch("packages.shared.jimeng_client.get_shared_settings") as ms:
+            ms.return_value = _make_settings(timeout=1, poll_interval=0)
+            c = JimengClient()
+        queue_resp = MagicMock(status_code=200)
+        queue_resp.json.return_value = {"code": 10000, "data": {"status": "in_queue"}}
+        # time.time 会被调用，模拟超时
+        with (
+            patch("packages.shared.jimeng_client.httpx.request", return_value=queue_resp),
+            patch("packages.shared.jimeng_client.time.sleep"),
+        ):
+            url = c._poll_result("abc")
+        assert url is None
+        assert c.last_video_error["error_code"] == "network_error"
+        assert "超时" in c.last_video_error["user_message"]