fix: bring a cached tier back as a tier, not as a bare string

Adding applied_effort to LLMResponse put it through the cache round
trip, where JSON stores a StrEnum as its plain value. Rehydrated raw, a
hit would hand downstream a str while the annotation says Effort, and
every `is Effort.LOW` in the library would quietly answer False on the
hit path only -- the same trap thinking_observation already has a
coercion for.

A value outside this version's vocabulary degrades to None rather than
failing the entry: projects sharing one Redis would otherwise keep
invalidating each other's writes over an attribution field, and None is
the honest reading of a tier this version cannot name.

src/polygateway/middleware/cache.py

@@ -53,6 +53,29 @@ def _coerce_observation(raw: Any) -> ThinkingObservation:
         return ThinkingObservation.UNKNOWN
 
 
+def _coerce_applied_effort(raw: Any) -> Effort | None:
+    """缓存里的档位字符串 → 枚举;域外取值降级为 `None`,**不作废整条缓存**。
+
+    与 `_coerce_observation` 同源同向,理由逐条相同: 多项目共用一个 Redis 时,
+    先升级的进程可能写入本版没有的档位名,未升级的进程若把这些条目判成未命中,
+    两个版本就会互相打对方的缓存。归因字段不该有能力废掉内容完好的响应。
+
+    降级到 `None` 而不是别的档: 它的语义是"库不知道这次跑在哪档",对一个读不懂
+    的取值这是唯一诚实的说法——随便挑一档等于替上游声称了一件它没说过的事。
+    """
+    if raw is None:
+        return None
+    try:
+        return Effort(raw)
+    except ValueError:
+        logger.warning(
+            "缓存条目的 applied_effort 取值 {!r} 不在本版档位词汇内(多半由更新版本的"
+            "进程写入),已降级为 None;响应内容照常复活——归因字段不作废缓存",
+            raw,
+        )
+        return None
+
+
 def digest_messages(messages: list[dict[str, Any]]) -> list[dict[str, Any]]:
     """多模态 content part 先各自 sha256 摘要再参与序列化;文本原文参与。
 
@@ -176,6 +199,8 @@ class CacheMW:
             # 键缺失即升级前写入的旧条目,交给 dataclass 默认值
             if "thinking_observation" in fields:
                 fields["thinking_observation"] = _coerce_observation(fields["thinking_observation"])
+            if "applied_effort" in fields:
+                fields["applied_effort"] = _coerce_applied_effort(fields["applied_effort"])
             fields.update(
                 cache_hit=True,
                 latency_ms=0,