feat(sense): T-G7 审计+前端（live 分流接线 + tier 指标卡 + Settings 灰度区块）

- proxy/routes：sense.mode=live 时转发前调 Grader 分流，档位 -> 池条目
  （tier_pool_hint 映射），响应头 X-Campus-Tier 透出（可解释性）；
  T1 决策按 review.sample_rate 抽样入 ReviewQueue（sense_t1 标签，
  人工核"判 T1 实为更高档"）；分级故障不影响代理可用性（D-G4）
- /api/metrics：sense 段（enabled/mode/观察数/已标签/档位分布/shadow 一致率）
- 前端：MetricsView「语义分级」卡片（观察/标签进度/一致率/晋升门提示）；
  SettingsView「Sense 灰度控制」区块（总开关 + collect/shadow/live 切换 +
  标签进度），npm run build 产物更新
- 全量 410 passed

gateway/proxy/routes.py

@@ -78,11 +78,59 @@ def build_proxy_router(cfg: ProxyConfig, pool) -> APIRouter:
                            _limits(), time.time())
         if not _limits().acquire_slot(ctx["key_id"]):
             raise QuotaError("并发请求已达该 key 上限")
+
+        # 语义分析器 live 分流（D-G7：mode=live 才启用；任何异常不影响代理可用性）
+        tier_used = None
         try:
-            return await _run_chat(body, dict(request.headers), ctx, cfg, ledger, pool)
+            from gateway.sense.config import build_sense_config
+            from gateway.sense.grader import Grader
+            from gateway.sense.observer import get_observer
+            from gateway.sense.store import SenseStore
+            scfg = build_sense_config(settings_store().to_dict())
+            if scfg.enabled and scfg.mode == "live":
+                sstore = SenseStore.init_db(scfg.db_path)
+                grader = Grader(scfg, sstore, get_observer(sstore))
+                qtext = "\n".join(str(m.get("content") or "")
+                                  for m in (body.get("messages") or []))
+                d = await grader.decide(qtext or str(body.get("model") or ""),
+                                        "proxy")
+                tier_used = d.tier
+                # 档位 -> 池条目（§2 消费方表）：t1/t2/t3 映射模型替换请求模型
+                hint = scfg.tier_pool_hint("proxy", tier_used)
+                e = _entry_by_tier_hint(pool, hint)
+                if e is not None:
+                    body = {**body, "model": e["model"]}
+        except Exception:
+            pass  # D-G4：分级故障不影响代理可用性
+        try:
+            resp = await _run_chat(body, dict(request.headers), ctx, cfg, ledger, pool)
+            if tier_used:
+                resp.headers["x-campus-tier"] = tier_used
+                # T1 审计抽样（§9.4）：按 review.sample_rate 入队人工核
+                try:
+                    import random as _random
+                    rate = float(load_config().get("review", {}).get("sample_rate", 0.1))
+                    if _random.random() < rate:
+                        get_review().enqueue(
+                            request_id + "-sense", str(body.get("model") or "proxy"),
+                            "(sense T1 审计抽样)", tags=["sense_t1"],
+                            reason="sense_audit")
+                except Exception:
+                    pass
+            return resp
         finally:
             _limits().release_slot(ctx["key_id"])
 
+    def _entry_by_tier_hint(pool, hint: str) -> Optional[Dict[str, Any]]:
+        """按档位名选池条目（tier_hint 优先，缺省按 tier 字段映射）。"""
+        entries = pool.list().get("entries", [])
+        for e in entries:
+            if not e.get("enabled") or e.get("backend") == "mock":
+                continue
+            if (e.get("tier_hint") or e.get("tier")) == hint:
+                return e
+        return None
+
     # ---------------- 管理面 ----------------
     def _guard_admin(request: Request) -> None:
         host = request.client.host if request.client else ""