存量积累：生图素人感约束+评图分+幂等防重跑+审核回路

- 015-017迁移：image_candidate 文案复审/AI视觉分/重生标记
- constants C7素人感约束(反电商摆拍对齐真实笔记)+C3叠字口子
- celery visibility_timeout=2h 防长任务被误判重投重复烧钱(task75教训)
- image_scorer 评图分(只筛选+展示,真实信号才进飞轮权重)
- storyboard/image_gen/pipeline_io 生图存量
- task_actions/tasks/task_service/flywheel 审核回路+飞轮存量

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

backend/app/services/ai_engine/gemini_factory.py

@@ -41,7 +41,7 @@ class AIClients:
     _pool_loop_id: int | None = field(default=None, repr=False)
     _gemini_key: str | None = field(default=None, repr=False)   # 局部变量不打印
     _model_image: str = "gpt-image-2"
-    _model_text: str = "claude-sonnet-4-5"  # apiports无gpt-4o-mini,文案用claude中文质量好
+    _model_text: str = "claude-opus-4-8"  # 最强档(倩倩姐红线):Claude系一律4.8,绝不降级
 
     def _client(self) -> httpx.AsyncClient:
         """主通道(apiports) client，按当前事件循环缓存"""
@@ -111,16 +111,44 @@ class AIClients:
         resp.raise_for_status()
         return _extract_gemini_image(resp.json())
 
+    async def _chat_post_failover(self, payload: dict, timeout: float) -> dict:
+        """
+        chat/completions 发送器，带 codeproxy 回落。
+        主通道(apiports, claude-opus-4-8)若 5xx/超时/连接错，自动切 codeproxy 重试一次。
+        ⚠ codeproxy 账号池只支持 gpt 系(gpt-5.5)，无 claude，故回落时模型换成 gpt-5.5。
+        这是强档↔强档切换：红线"Claude系4.8/GPT系5.5"本就是两个平级最强档、互为兜底，
+        非降级(降级=落 sonnet/弱档)。codeproxy 回落档由 CODEPROXY_CHAT_MODEL 配置(默认 gpt-5.5)。
+        codeproxy 未配置(_alt_token 为空)时不回落，原样抛错。
+        """
+        main_base = (os.environ.get("IMAGE_API_BASE") or os.environ.get("APIPORTS_BASE_URL") or "").rstrip("/")
+        try:
+            resp = await self._client().post(f"{main_base}/chat/completions", json=payload, timeout=timeout)
+            resp.raise_for_status()
+            return resp.json()
+        except (httpx.HTTPStatusError, httpx.TransportError, httpx.TimeoutException) as exc:
+            # 仅 5xx(服务端过载)或网络层错误才回落；4xx(参数/鉴权)回落也没用，直接抛。
+            status = getattr(getattr(exc, "response", None), "status_code", None)
+            retryable = status is None or status >= 500
+            if not (retryable and self._alt_token):
+                raise
+            alt_base = (self._alt_base or os.environ.get("CODEPROXY_BASE_URL") or "").rstrip("/")
+            alt_payload = dict(payload)
+            alt_payload["model"] = os.environ.get("CODEPROXY_CHAT_MODEL", "gpt-5.5")
+            logger.warning("apiports chat 失败(%s)，回落 codeproxy 用 %s 重试（强档兜底，非降级）",
+                           status or type(exc).__name__, alt_payload["model"])
+            client = self._client_for(alt_base, self._alt_token)
+            resp = await client.post(f"{alt_base}/chat/completions", json=alt_payload, timeout=timeout)
+            resp.raise_for_status()
+            return resp.json()
+
     async def chat_complete(self, messages: list[dict], model: str | None = None, max_tokens: int = 4096, temperature: float = 0.75) -> str:
-        """文字生成（文案生成用）"""
-        base = (os.environ.get("IMAGE_API_BASE") or os.environ.get("APIPORTS_BASE_URL") or "").rstrip("/")
+        """文字生成（文案生成用）。apiports 503 时自动回落 codeproxy。"""
         payload = {"model": model or self._model_text, "messages": messages, "max_tokens": max_tokens, "temperature": temperature}
         # 单批≤4条文案正常 40-55s 返回；apiports 网关 ~60s 上限。客户端超时设 75s：
         # 略高于网关上限即可，过长（如180s）会在 apiports 卡顿时干等，拖慢整体。
         timeout = float(os.environ.get("TEXT_LLM_TIMEOUT", "75"))
-        resp = await self._client().post(f"{base}/chat/completions", json=payload, timeout=timeout)
-        resp.raise_for_status()
-        return resp.json()["choices"][0]["message"]["content"] or ""
+        data = await self._chat_post_failover(payload, timeout)
+        return data["choices"][0]["message"]["content"] or ""
 
     async def gpt_vision_analyze(self, prompt: str, images: list[bytes], model: str | None = None) -> str:
         """
@@ -139,16 +167,15 @@ class AIClients:
             })
         used_model = model or os.environ.get("MODEL_TEXT", "claude-opus-4-8")
         messages = [{"role": "user", "content": content}]
-        base = (os.environ.get("IMAGE_API_BASE") or os.environ.get("APIPORTS_BASE_URL") or "").rstrip("/")
         payload = {
             "model": used_model,
             "messages": messages,
             "max_tokens": 2048,
             "temperature": 0.2,
         }
-        resp = await self._client().post(f"{base}/chat/completions", json=payload, timeout=90.0)
-        resp.raise_for_status()
-        return resp.json()["choices"][0]["message"]["content"] or ""
+        # 评图也走 codeproxy 回落：apiports 503 时切备用站，模型档不变（守红线）。
+        data = await self._chat_post_failover(payload, 90.0)
+        return data["choices"][0]["message"]["content"] or ""
 
     # duck-type: text_variants._call_llm 用的属性
     @property