feat(health): refine health monitor metrics

- add TPS to model and provider health payloads

- exclude user-cancelled 499 requests from health statistics

- update model/provider health cards with average latency, average TTFB, TPS, and availability

apps/aether-gateway/src/handlers/admin/observability/usage/summary_routes.rs

@@ -486,6 +486,7 @@ fn build_admin_usage_keyword_search_query(
         model: base_query.model.clone(),
         api_format: base_query.api_format.clone(),
         statuses: base_query.statuses.clone(),
+        exclude_status_codes: base_query.exclude_status_codes.clone(),
         is_stream: base_query.is_stream,
         error_only: base_query.error_only,
         keywords,

apps/aether-gateway/src/handlers/public/catalog_helpers.rs

@@ -23,6 +23,8 @@ use serde_json::json;
 use std::collections::{BTreeMap, BTreeSet};
 use std::time::{SystemTime, UNIX_EPOCH};
 
+const USER_CANCELLED_STATUS_CODE: u16 = 499;
+
 pub(crate) fn request_candidate_status_label(status: RequestCandidateStatus) -> &'static str {
     match status {
         RequestCandidateStatus::Available => "available",
@@ -576,6 +578,7 @@ pub(crate) async fn build_model_health_monitor_payload(
             created_until_unix_secs: now_unix_secs,
             user_id: None,
             provider_name: None,
+            exclude_status_codes: vec![USER_CANCELLED_STATUS_CODE],
             group_by: UsageBreakdownGroupBy::Model,
         })
         .await
@@ -595,6 +598,7 @@ pub(crate) async fn build_model_health_monitor_payload(
                 created_from_unix_secs: Some(since_unix_secs),
                 created_until_unix_secs: Some(now_unix_secs),
                 model: Some(row.group_key.clone()),
+                exclude_status_codes: vec![USER_CANCELLED_STATUS_CODE],
                 limit: Some(per_model_limit),
                 newest_first: true,
                 ..UsageAuditListQuery::default()
@@ -640,6 +644,7 @@ pub(crate) async fn build_model_health_monitor_payload(
             "success_rate": success_rate,
             "avg_latency_ms": avg_latency_ms,
             "avg_first_byte_ms": first_byte_average,
+            "avg_tps": model_health_average_tps(&row),
             "last_event_at": last_event_at,
             "events": event_payload,
             "timeline": timeline,
@@ -691,6 +696,7 @@ pub(crate) async fn build_provider_health_monitor_payload(
             created_until_unix_secs: now_unix_secs,
             user_id: None,
             provider_name: None,
+            exclude_status_codes: vec![USER_CANCELLED_STATUS_CODE],
             group_by: UsageBreakdownGroupBy::Provider,
         })
         .await
@@ -754,6 +760,7 @@ async fn build_provider_health_payload(
             created_until_unix_secs: now_unix_secs,
             user_id: None,
             provider_name: Some(provider.name.clone()),
+            exclude_status_codes: vec![USER_CANCELLED_STATUS_CODE],
             group_by: UsageBreakdownGroupBy::Model,
         })
         .await
@@ -768,6 +775,7 @@ async fn build_provider_health_payload(
             created_from_unix_secs: Some(since_unix_secs),
             created_until_unix_secs: Some(now_unix_secs),
             provider_name: Some(provider.name.clone()),
+            exclude_status_codes: vec![USER_CANCELLED_STATUS_CODE],
             limit: Some(provider_event_limit),
             newest_first: true,
             ..UsageAuditListQuery::default()
@@ -804,7 +812,7 @@ async fn build_provider_health_payload(
         ));
     }
 
-    let (total_attempts, success_count, failed_count, success_rate, avg_latency_ms) =
+    let (total_attempts, success_count, failed_count, success_rate, avg_latency_ms, avg_tps) =
         if let Some(row) = provider_stats {
             let total_attempts = row.request_count;
             let success_count = row.success_count.min(total_attempts);
@@ -820,9 +828,10 @@ async fn build_provider_health_payload(
                 failed_count,
                 success_rate,
                 model_health_average_latency_ms(row),
+                model_health_average_tps(row),
             )
         } else {
-            (0, 0, 0, 1.0, None)
+            (0, 0, 0, 1.0, None, None)
         };
 
     let (timeline, time_range_start, time_range_end) = build_model_health_timeline(
@@ -847,6 +856,7 @@ async fn build_provider_health_payload(
         "success_rate": success_rate,
         "avg_latency_ms": avg_latency_ms,
         "avg_first_byte_ms": model_health_average_first_byte_ms(&provider_events),
+        "avg_tps": avg_tps,
         "model_count": model_breakdown.len(),
         "last_event_at": last_event_at,
         "timeline": timeline,
@@ -917,6 +927,7 @@ fn model_health_payload_from_row(
         "success_rate": success_rate,
         "avg_latency_ms": model_health_average_latency_ms(row),
         "avg_first_byte_ms": model_health_average_first_byte_ms(events),
+        "avg_tps": model_health_average_tps(row),
         "last_event_at": last_event_at,
         "events": event_payload,
         "timeline": timeline,
@@ -939,6 +950,13 @@ fn model_health_average_latency_ms(row: &StoredUsageBreakdownSummaryRow) -> Opti
     None
 }
 
+fn model_health_average_tps(row: &StoredUsageBreakdownSummaryRow) -> Option<f64> {
+    if row.output_tokens == 0 || row.response_time_sum_ms <= 0.0 {
+        return None;
+    }
+    Some(row.output_tokens as f64 / (row.response_time_sum_ms / 1000.0))
+}
+
 fn model_health_average_first_byte_ms(events: &[StoredRequestUsageAudit]) -> Option<f64> {
     let mut sum = 0u64;
     let mut count = 0u64;

apps/aether-gateway/src/handlers/public/support/user_me_usage.rs

@@ -848,6 +848,7 @@ pub(super) async fn handle_users_me_usage_get(
                 created_until_unix_secs,
                 user_id: Some(auth.user.id.clone()),
                 provider_name: None,
+                exclude_status_codes: Vec::new(),
                 group_by: UsageBreakdownGroupBy::Model,
             })
             .await
@@ -868,6 +869,7 @@ pub(super) async fn handle_users_me_usage_get(
                     created_until_unix_secs,
                     user_id: Some(auth.user.id.clone()),
                     provider_name: None,
+                    exclude_status_codes: Vec::new(),
                     group_by: UsageBreakdownGroupBy::Provider,
                 })
                 .await
@@ -888,6 +890,7 @@ pub(super) async fn handle_users_me_usage_get(
                 created_until_unix_secs,
                 user_id: Some(auth.user.id.clone()),
                 provider_name: None,
+                exclude_status_codes: Vec::new(),
                 group_by: UsageBreakdownGroupBy::ApiFormat,
             })
             .await
@@ -926,6 +929,7 @@ pub(super) async fn handle_users_me_usage_get(
                 model: None,
                 api_format: None,
                 statuses: None,
+                exclude_status_codes: Vec::new(),
                 is_stream: None,
                 error_only: false,
                 keywords,
@@ -979,6 +983,7 @@ pub(super) async fn handle_users_me_usage_get(
                     model: None,
                     api_format: None,
                     statuses: None,
+                    exclude_status_codes: Vec::new(),
                     is_stream: None,
                     error_only: false,
                     limit: None,
@@ -1005,6 +1010,7 @@ pub(super) async fn handle_users_me_usage_get(
                     model: None,
                     api_format: None,
                     statuses: None,
+                    exclude_status_codes: Vec::new(),
                     is_stream: None,
                     error_only: false,
                     limit: Some(limit),
@@ -1141,6 +1147,7 @@ pub(super) async fn handle_users_me_usage_active_get(
                 model: None,
                 api_format: None,
                 statuses: Some(vec!["pending".to_string(), "streaming".to_string()]),
+                exclude_status_codes: Vec::new(),
                 is_stream: None,
                 error_only: false,
                 limit: Some(50),

crates/aether-data-contracts/src/repository/usage/types.rs

@@ -731,6 +731,7 @@ pub struct UsageAuditListQuery {
     pub model: Option<String>,
     pub api_format: Option<String>,
     pub statuses: Option<Vec<String>>,
+    pub exclude_status_codes: Vec<u16>,
     pub is_stream: Option<bool>,
     pub error_only: bool,
     pub limit: Option<usize>,
@@ -747,6 +748,7 @@ pub struct UsageAuditKeywordSearchQuery {
     pub model: Option<String>,
     pub api_format: Option<String>,
     pub statuses: Option<Vec<String>>,
+    pub exclude_status_codes: Vec<u16>,
     pub is_stream: Option<bool>,
     pub error_only: bool,
     pub keywords: Vec<String>,
@@ -986,6 +988,7 @@ pub struct UsageBreakdownSummaryQuery {
     pub created_until_unix_secs: u64,
     pub user_id: Option<String>,
     pub provider_name: Option<String>,
+    pub exclude_status_codes: Vec<u16>,
     pub group_by: UsageBreakdownGroupBy,
 }
 

crates/aether-data/src/repository/usage/memory.rs

@@ -263,6 +263,12 @@ fn usage_matches_list_query(item: &StoredRequestUsageAudit, query: &UsageAuditLi
             return false;
         }
     }
+    if item
+        .status_code
+        .is_some_and(|status_code| query.exclude_status_codes.contains(&status_code))
+    {
+        return false;
+    }
     if let Some(is_stream) = query.is_stream {
         if item.is_stream != is_stream {
             return false;
@@ -323,6 +329,12 @@ fn usage_matches_keyword_search_query(
             return false;
         }
     }
+    if item
+        .status_code
+        .is_some_and(|status_code| query.exclude_status_codes.contains(&status_code))
+    {
+        return false;
+    }
     if let Some(is_stream) = query.is_stream {
         if item.is_stream != is_stream {
             return false;
@@ -540,6 +552,12 @@ fn usage_matches_breakdown_summary_query(
             return false;
         }
     }
+    if item
+        .status_code
+        .is_some_and(|status_code| query.exclude_status_codes.contains(&status_code))
+    {
+        return false;
+    }
     match query.group_by {
         UsageBreakdownGroupBy::Model | UsageBreakdownGroupBy::Provider => true,
         UsageBreakdownGroupBy::ApiFormat => item.api_format.is_some(),

crates/aether-data/src/repository/usage/postgres/mod.rs

@@ -86,6 +86,24 @@ fn dashboard_utc_to_unix_secs(value: DateTime<Utc>) -> u64 {
     value.timestamp().max(0) as u64
 }
 
+fn push_postgres_usage_excluded_status_codes(
+    builder: &mut QueryBuilder<'_, Postgres>,
+    has_where: &mut bool,
+    status_codes: &[u16],
+) {
+    if status_codes.is_empty() {
+        return;
+    }
+    builder.push(if *has_where { " AND " } else { " WHERE " });
+    *has_where = true;
+    builder.push("(\"usage\".status_code IS NULL OR \"usage\".status_code NOT IN (");
+    let mut separated = builder.separated(", ");
+    for status_code in status_codes {
+        separated.push_bind(i32::from(*status_code));
+    }
+    separated.push_unseparated("))");
+}
+
 fn dashboard_utc_midnight(value: DateTime<Utc>) -> DateTime<Utc> {
     DateTime::<Utc>::from_naive_utc_and_offset(
         value
@@ -2529,6 +2547,7 @@ ORDER BY request_count DESC, "usage".provider_name ASC
         }
         if let Some(user_id) = query.user_id.as_deref() {
             builder.push(if has_where { " AND " } else { " WHERE " });
+            has_where = true;
             builder
                 .push("\"usage\".user_id = ")
                 .push_bind(user_id.to_string());
@@ -2566,6 +2585,11 @@ ORDER BY request_count DESC, "usage".provider_name ASC
                 separated.push_unseparated(")");
             }
         }
+        push_postgres_usage_excluded_status_codes(
+            &mut builder,
+            &mut has_where,
+            &query.exclude_status_codes,
+        );
         if let Some(is_stream) = query.is_stream {
             builder.push(if has_where { " AND " } else { " WHERE " });
             has_where = true;
@@ -2625,6 +2649,7 @@ OR (\"usage\".error_message IS NOT NULL AND BTRIM(\"usage\".error_message) <> ''
         }
         if let Some(user_id) = query.user_id.as_deref() {
             builder.push(if has_where { " AND " } else { " WHERE " });
+            has_where = true;
             builder
                 .push("\"usage\".user_id = ")
                 .push_bind(user_id.to_string());
@@ -2662,6 +2687,11 @@ OR (\"usage\".error_message IS NOT NULL AND BTRIM(\"usage\".error_message) <> ''
                 separated.push_unseparated(")");
             }
         }
+        push_postgres_usage_excluded_status_codes(
+            &mut builder,
+            &mut has_where,
+            &query.exclude_status_codes,
+        );
         if let Some(is_stream) = query.is_stream {
             builder.push(if has_where { " AND " } else { " WHERE " });
             has_where = true;
@@ -2840,6 +2870,11 @@ OR (\"usage\".error_message IS NOT NULL AND BTRIM(\"usage\".error_message) <> ''
                 separated.push_unseparated(")");
             }
         }
+        push_postgres_usage_excluded_status_codes(
+            &mut builder,
+            &mut has_where,
+            &query.exclude_status_codes,
+        );
         if let Some(is_stream) = query.is_stream {
             builder.push(if has_where { " AND " } else { " WHERE " });
             has_where = true;
@@ -2926,6 +2961,11 @@ OR (\"usage\".error_message IS NOT NULL AND BTRIM(\"usage\".error_message) <> ''
                 separated.push_unseparated(")");
             }
         }
+        push_postgres_usage_excluded_status_codes(
+            &mut builder,
+            &mut has_where,
+            &query.exclude_status_codes,
+        );
         if let Some(is_stream) = query.is_stream {
             builder.push(if has_where { " AND " } else { " WHERE " });
             has_where = true;
@@ -4801,6 +4841,11 @@ WITH filtered_usage AS (
                 .push("\"usage\".provider_name = ")
                 .push_bind(provider_name.to_string());
         }
+        push_postgres_usage_excluded_status_codes(
+            &mut builder,
+            &mut has_where,
+            &query.exclude_status_codes,
+        );
         builder.push(filtered_extra_where);
         builder.push(
             r#"
@@ -4893,6 +4938,9 @@ ORDER BY request_count DESC, group_key ASC
         &self,
         query: &UsageBreakdownSummaryQuery,
     ) -> Result<Vec<StoredUsageBreakdownSummaryRow>, DataLayerError> {
+        if !query.exclude_status_codes.is_empty() {
+            return self.summarize_usage_breakdown_raw(query).await;
+        }
         if query.provider_name.is_some() {
             return self.summarize_usage_breakdown_raw(query).await;
         }
@@ -4918,6 +4966,7 @@ ORDER BY request_count DESC, group_key ASC
                     created_until_unix_secs: dashboard_utc_to_unix_secs(raw_end),
                     user_id: Some(user_id.to_string()),
                     provider_name: None,
+                    exclude_status_codes: query.exclude_status_codes.clone(),
                     group_by: query.group_by,
                 })
                 .await?;
@@ -4941,6 +4990,7 @@ ORDER BY request_count DESC, group_key ASC
                     created_until_unix_secs: dashboard_utc_to_unix_secs(raw_end),
                     user_id: Some(user_id.to_string()),
                     provider_name: None,
+                    exclude_status_codes: query.exclude_status_codes.clone(),
                     group_by: query.group_by,
                 })
                 .await?;

crates/aether-data/src/repository/usage/sqlite.rs

@@ -466,6 +466,7 @@ fn push_sqlite_usage_list_filters(
             separated.push_unseparated(")");
         }
     }
+    push_sqlite_usage_excluded_status_codes(builder, has_where, &query.exclude_status_codes);
     if let Some(is_stream) = query.is_stream {
         push_sqlite_usage_where(builder, has_where);
         builder
@@ -482,6 +483,23 @@ OR (error_message IS NOT NULL AND TRIM(error_message) <> ''))",
     }
 }
 
+fn push_sqlite_usage_excluded_status_codes(
+    builder: &mut QueryBuilder<'_, Sqlite>,
+    has_where: &mut bool,
+    status_codes: &[u16],
+) {
+    if status_codes.is_empty() {
+        return;
+    }
+    push_sqlite_usage_where(builder, has_where);
+    builder.push("(status_code IS NULL OR status_code NOT IN (");
+    let mut separated = builder.separated(", ");
+    for status_code in status_codes {
+        separated.push_bind(i64::from(*status_code));
+    }
+    separated.push_unseparated("))");
+}
+
 fn push_sqlite_usage_keyword_filters(
     builder: &mut QueryBuilder<'_, Sqlite>,
     query: &UsageAuditKeywordSearchQuery,
@@ -497,6 +515,7 @@ fn push_sqlite_usage_keyword_filters(
             model: query.model.clone(),
             api_format: query.api_format.clone(),
             statuses: query.statuses.clone(),
+            exclude_status_codes: query.exclude_status_codes.clone(),
             is_stream: query.is_stream,
             error_only: query.error_only,
             limit: None,
@@ -2471,6 +2490,11 @@ FROM "usage"
             "provider_name",
             query.provider_name.as_deref(),
         );
+        push_sqlite_usage_excluded_status_codes(
+            &mut builder,
+            &mut has_where,
+            &query.exclude_status_codes,
+        );
         if matches!(query.group_by, UsageBreakdownGroupBy::ApiFormat) {
             push_sqlite_usage_where(&mut builder, &mut has_where);
             builder.push("api_format IS NOT NULL");

frontend/src/api/endpoints/types/provider.ts

@@ -631,6 +631,7 @@ export interface ModelStatusMonitor {
   success_rate: number
   avg_latency_ms?: number | null
   avg_first_byte_ms?: number | null
+  avg_tps?: number | null
   provider_count?: number
   last_event_at?: string | null
   events: ModelHealthEvent[]
@@ -655,6 +656,7 @@ export interface ProviderStatusMonitor {
   success_rate: number
   avg_latency_ms?: number | null
   avg_first_byte_ms?: number | null
+  avg_tps?: number | null
   model_count: number
   last_event_at?: string | null
   timeline?: string[]

frontend/src/features/providers/components/ModelHealthMonitorCard.vue

@@ -1,16 +1,15 @@
 <template>
-  <Card
-    variant="default"
-    class="overflow-hidden"
-  >
+  <Card variant="default" class="overflow-hidden">
     <div class="px-6 py-3.5 border-b border-border/60">
-      <div class="flex flex-col gap-3 sm:flex-row sm:items-center sm:justify-between">
+      <div
+        class="flex flex-col gap-3 sm:flex-row sm:items-center sm:justify-between"
+      >
         <div>
           <h3 class="text-base font-semibold">
             {{ title }}
           </h3>
           <p class="mt-1 text-xs text-muted-foreground">
-            基于真实请求统计模型可用率、响应延迟与首包延迟
+            基于真实请求统计模型可用率、平均耗时、平均TTFB 与平均速度
           </p>
         </div>
         <div class="flex items-center gap-3">
@@ -20,27 +19,14 @@
               <SelectValue />
             </SelectTrigger>
             <SelectContent>
-              <SelectItem value="1">
-                1 小时
-              </SelectItem>
-              <SelectItem value="6">
-                6 小时
-              </SelectItem>
-              <SelectItem value="12">
-                12 小时
-              </SelectItem>
-              <SelectItem value="24">
-                24 小时
-              </SelectItem>
-              <SelectItem value="48">
-                48 小时
-              </SelectItem>
+              <SelectItem value="1"> 1 小时 </SelectItem>
+              <SelectItem value="6"> 6 小时 </SelectItem>
+              <SelectItem value="12"> 12 小时 </SelectItem>
+              <SelectItem value="24"> 24 小时 </SelectItem>
+              <SelectItem value="48"> 48 小时 </SelectItem>
             </SelectContent>
           </Select>
-          <RefreshButton
-            :loading="loading"
-            @click="refreshData"
-          />
+          <RefreshButton :loading="loading" @click="refreshData" />
         </div>
       </div>
     </div>
@@ -60,59 +46,80 @@
       >
         <Bot class="w-12 h-12 mb-3 opacity-30" />
         <p>暂无模型健康监控数据</p>
-        <p class="text-xs mt-1">
-          模型尚未产生请求记录
-        </p>
+        <p class="text-xs mt-1">模型尚未产生请求记录</p>
       </div>
 
-      <div
-        v-else
-        class="grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4"
-      >
+      <div v-else class="grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4">
         <div
           v-for="monitor in monitors"
           :key="monitor.model"
           class="relative overflow-hidden rounded-xl border border-border/60 bg-card/60 p-4 transition-colors hover:border-primary/50"
         >
-          <div class="absolute inset-x-0 top-0 h-px bg-gradient-to-r from-transparent via-primary/40 to-transparent" />
+          <div
+            class="absolute inset-x-0 top-0 h-px bg-gradient-to-r from-transparent via-primary/40 to-transparent"
+          />
           <div class="flex items-start justify-between gap-3">
             <div class="flex min-w-0 items-center gap-3">
-              <div class="flex h-11 w-11 flex-shrink-0 items-center justify-center rounded-xl border border-border/60 bg-muted/50">
+              <div
+                class="flex h-11 w-11 flex-shrink-0 items-center justify-center rounded-xl border border-border/60 bg-muted/50"
+              >
                 <Bot class="h-5 w-5 text-muted-foreground" />
               </div>
               <h4 class="min-w-0 truncate text-sm font-semibold">
                 {{ monitor.display_name || monitor.model }}
               </h4>
             </div>
-            <Badge
-              :variant="getHealthBadgeVariant(monitor)"
-              class="shrink-0"
-            >
+            <Badge :variant="getHealthBadgeVariant(monitor)" class="shrink-0">
               {{ getHealthLabel(monitor) }}
             </Badge>
           </div>
 
-          <div class="mt-4 grid grid-cols-3 gap-2">
-            <div class="rounded-lg border border-border/40 bg-muted/20 px-3 py-2">
-              <div class="flex items-center gap-1.5 text-[11px] text-muted-foreground">
+          <div class="mt-4 grid grid-cols-4 gap-2">
+            <div
+              class="rounded-lg border border-border/40 bg-muted/20 px-3 py-2"
+            >
+              <div
+                class="flex items-center gap-1.5 text-[11px] leading-tight text-muted-foreground"
+              >
                 <Gauge class="h-3.5 w-3.5" />
-                延迟
+                平均耗时
               </div>
               <div class="mt-1 text-sm font-semibold tabular-nums">
                 {{ formatMs(monitor.avg_latency_ms) }}
               </div>
             </div>
-            <div class="rounded-lg border border-border/40 bg-muted/20 px-3 py-2">
-              <div class="flex items-center gap-1.5 text-[11px] text-muted-foreground">
+            <div
+              class="rounded-lg border border-border/40 bg-muted/20 px-3 py-2"
+            >
+              <div
+                class="flex items-center gap-1.5 text-[11px] leading-tight text-muted-foreground"
+              >
                 <Radio class="h-3.5 w-3.5" />
-                Ping
+                平均TTFB
               </div>
               <div class="mt-1 text-sm font-semibold tabular-nums">
                 {{ formatMs(monitor.avg_first_byte_ms) }}
               </div>
             </div>
-            <div class="rounded-lg border border-border/40 bg-muted/20 px-3 py-2">
-              <div class="flex items-center gap-1.5 text-[11px] text-muted-foreground">
+            <div
+              class="rounded-lg border border-border/40 bg-muted/20 px-3 py-2"
+            >
+              <div
+                class="flex items-center gap-1.5 text-[11px] leading-tight text-muted-foreground"
+              >
+                <Zap class="h-3.5 w-3.5" />
+                平均速度
+              </div>
+              <div class="mt-1 text-sm font-semibold tabular-nums">
+                {{ formatTps(monitor.avg_tps) }}
+              </div>
+            </div>
+            <div
+              class="rounded-lg border border-border/40 bg-muted/20 px-3 py-2"
+            >
+              <div
+                class="flex items-center gap-1.5 text-[11px] leading-tight text-muted-foreground"
+              >
                 <Activity class="h-3.5 w-3.5" />
                 可用率
               </div>
@@ -125,7 +132,9 @@
             </div>
           </div>
 
-          <div class="mt-4 flex items-center justify-between gap-3 text-[11px] uppercase tracking-wide text-muted-foreground">
+          <div
+            class="mt-4 flex items-center justify-between gap-3 text-[11px] uppercase tracking-wide text-muted-foreground"
+          >
             <span>History (60pts)</span>
             <span class="truncate normal-case tracking-normal">
               {{ getModelMetaText(monitor) }}
@@ -144,11 +153,7 @@
                     :class="getTimelineColor(segment)"
                   />
                 </TooltipTrigger>
-                <TooltipContent
-                  side="top"
-                  :side-offset="8"
-                  class="max-w-xs"
-                >
+                <TooltipContent side="top" :side-offset="8" class="max-w-xs">
                   <div class="text-xs whitespace-pre-line">
                     {{ buildTimelineTooltip(monitor, segment, index) }}
                   </div>
@@ -157,9 +162,13 @@
             </div>
           </TooltipProvider>
 
-          <div class="mt-2 flex items-center justify-between text-[10px] text-muted-foreground">
+          <div
+            class="mt-2 flex items-center justify-between text-[10px] text-muted-foreground"
+          >
             <span>{{ formatTimestamp(monitor.time_range_start) }}</span>
-            <span>{{ formatTimestamp(monitor.time_range_end || generatedAt) }}</span>
+            <span>{{
+              formatTimestamp(monitor.time_range_end || generatedAt)
+            }}</span>
           </div>
         </div>
       </div>
@@ -168,182 +177,220 @@
 </template>
 
 <script setup lang="ts">
-import { ref, onMounted, watch } from 'vue'
-import { Activity, Bot, Gauge, Loader2, Radio } from 'lucide-vue-next'
-import Card from '@/components/ui/card.vue'
-import Badge from '@/components/ui/badge.vue'
-import Label from '@/components/ui/label.vue'
-import Select from '@/components/ui/select.vue'
-import SelectTrigger from '@/components/ui/select-trigger.vue'
-import SelectValue from '@/components/ui/select-value.vue'
-import SelectContent from '@/components/ui/select-content.vue'
-import SelectItem from '@/components/ui/select-item.vue'
-import RefreshButton from '@/components/ui/refresh-button.vue'
-import { Tooltip, TooltipContent, TooltipProvider, TooltipTrigger } from '@/components/ui/tooltip'
-import { getModelStatusMonitor, getPublicModelStatusMonitor } from '@/api/endpoints/health'
-import type { ModelStatusMonitor } from '@/api/endpoints/types'
-import { useToast } from '@/composables/useToast'
-import { parseApiError } from '@/utils/errorParser'
+import { ref, onMounted, watch } from "vue";
+import { Activity, Bot, Gauge, Loader2, Radio, Zap } from "lucide-vue-next";
+import Card from "@/components/ui/card.vue";
+import Badge from "@/components/ui/badge.vue";
+import Label from "@/components/ui/label.vue";
+import Select from "@/components/ui/select.vue";
+import SelectTrigger from "@/components/ui/select-trigger.vue";
+import SelectValue from "@/components/ui/select-value.vue";
+import SelectContent from "@/components/ui/select-content.vue";
+import SelectItem from "@/components/ui/select-item.vue";
+import RefreshButton from "@/components/ui/refresh-button.vue";
+import {
+  Tooltip,
+  TooltipContent,
+  TooltipProvider,
+  TooltipTrigger,
+} from "@/components/ui/tooltip";
+import {
+  getModelStatusMonitor,
+  getPublicModelStatusMonitor,
+} from "@/api/endpoints/health";
+import type { ModelStatusMonitor } from "@/api/endpoints/types";
+import { useToast } from "@/composables/useToast";
+import { parseApiError } from "@/utils/errorParser";
 
-const props = withDefaults(defineProps<{
-  title?: string
-  isAdmin?: boolean
-  showProviderInfo?: boolean
-}>(), {
-  title: '模型健康监控',
-  isAdmin: false,
-  showProviderInfo: false
-})
+const props = withDefaults(
+  defineProps<{
+    title?: string;
+    isAdmin?: boolean;
+    showProviderInfo?: boolean;
+  }>(),
+  {
+    title: "模型健康监控",
+    isAdmin: false,
+    showProviderInfo: false,
+  },
+);
 
-const { error: showError } = useToast()
+const { error: showError } = useToast();
 
-const loading = ref(false)
-const loadingMonitors = ref(false)
-const monitors = ref<ModelStatusMonitor[]>([])
-const generatedAt = ref<string | null>(null)
-const lookbackHours = ref('6')
+const loading = ref(false);
+const loadingMonitors = ref(false);
+const monitors = ref<ModelStatusMonitor[]>([]);
+const generatedAt = ref<string | null>(null);
+const lookbackHours = ref("6");
 
 async function loadMonitors() {
-  loadingMonitors.value = true
+  loadingMonitors.value = true;
   try {
     const params = {
       lookback_hours: parseInt(lookbackHours.value),
       model_limit: 12,
-      per_model_limit: 100
-    }
+      per_model_limit: 100,
+    };
 
     const data = props.isAdmin
       ? await getModelStatusMonitor(params)
-      : await getPublicModelStatusMonitor(params)
-    monitors.value = data.models || []
-    generatedAt.value = data.generated_at || null
+      : await getPublicModelStatusMonitor(params);
+    monitors.value = data.models || [];
+    generatedAt.value = data.generated_at || null;
   } catch (err: unknown) {
-    showError(parseApiError(err, '加载模型健康监控数据失败'), '错误')
+    showError(parseApiError(err, "加载模型健康监控数据失败"), "错误");
   } finally {
-    loadingMonitors.value = false
+    loadingMonitors.value = false;
   }
 }
 
 async function refreshData() {
-  loading.value = true
+  loading.value = true;
   try {
-    await loadMonitors()
+    await loadMonitors();
   } finally {
-    loading.value = false
+    loading.value = false;
   }
 }
 
 function getHealthLabel(monitor: ModelStatusMonitor) {
-  if (monitor.total_attempts <= 0) return '未知'
-  if (monitor.success_rate >= 0.95) return '正常'
-  if (monitor.success_rate >= 0.8) return '波动'
-  return '异常'
+  if (monitor.total_attempts <= 0) return "未知";
+  if (monitor.success_rate >= 0.95) return "正常";
+  if (monitor.success_rate >= 0.8) return "波动";
+  return "异常";
 }
 
-function getHealthBadgeVariant(monitor: ModelStatusMonitor): 'default' | 'secondary' | 'destructive' | 'outline' | 'success' | 'warning' | 'dark' {
-  if (monitor.total_attempts <= 0) return 'outline'
-  if (monitor.success_rate >= 0.95) return 'success'
-  if (monitor.success_rate >= 0.8) return 'warning'
-  return 'destructive'
+function getHealthBadgeVariant(
+  monitor: ModelStatusMonitor,
+):
+  | "default"
+  | "secondary"
+  | "destructive"
+  | "outline"
+  | "success"
+  | "warning"
+  | "dark" {
+  if (monitor.total_attempts <= 0) return "outline";
+  if (monitor.success_rate >= 0.95) return "success";
+  if (monitor.success_rate >= 0.8) return "warning";
+  return "destructive";
 }
 
 function getSuccessRateClass(rate: number) {
-  if (rate >= 0.95) return 'text-green-600 dark:text-green-400'
-  if (rate >= 0.8) return 'text-amber-600 dark:text-amber-400'
-  return 'text-red-600 dark:text-red-400'
+  if (rate >= 0.95) return "text-green-600 dark:text-green-400";
+  if (rate >= 0.8) return "text-amber-600 dark:text-amber-400";
+  return "text-red-600 dark:text-red-400";
 }
 
 function getModelMetaText(monitor: ModelStatusMonitor) {
-  const attempts = `${formatCompactNumber(monitor.total_attempts)} 次请求`
-  if (props.showProviderInfo && typeof monitor.provider_count === 'number') {
-    return `${monitor.provider_count} 个提供商 / ${attempts}`
+  const attempts = `${formatCompactNumber(monitor.total_attempts)} 次请求`;
+  if (props.showProviderInfo && typeof monitor.provider_count === "number") {
+    return `${monitor.provider_count} 个提供商 / ${attempts}`;
   }
-  return attempts
+  return attempts;
 }
 
 function formatMs(value?: number | null) {
-  if (typeof value !== 'number' || Number.isNaN(value)) return '-'
-  const absValue = Math.abs(value)
-  if (absValue < 1000) return `${Math.round(value)} ms`
-  if (absValue < 60_000) return `${formatDurationNumber(value / 1000)} s`
-  return `${formatDurationNumber(value / 60_000)} min`
+  if (typeof value !== "number" || Number.isNaN(value)) return "-";
+  const absValue = Math.abs(value);
+  if (absValue < 1000) return `${Math.round(value)} ms`;
+  if (absValue < 60_000) return `${formatDurationNumber(value / 1000)} s`;
+  return `${formatDurationNumber(value / 60_000)} min`;
 }
 
 function formatDurationNumber(value: number) {
-  return new Intl.NumberFormat('zh-CN', {
-    maximumFractionDigits: Math.abs(value) < 10 ? 2 : 1
-  }).format(value)
+  return new Intl.NumberFormat("zh-CN", {
+    maximumFractionDigits: Math.abs(value) < 10 ? 2 : 1,
+  }).format(value);
 }
 
 function formatPercent(value: number) {
-  if (typeof value !== 'number' || Number.isNaN(value)) return '-'
-  return `${(value * 100).toFixed(2)}%`
+  if (typeof value !== "number" || Number.isNaN(value)) return "-";
+  return `${(value * 100).toFixed(2)}%`;
+}
+
+function formatTps(value?: number | null) {
+  if (typeof value !== "number" || Number.isNaN(value)) return "-";
+  return `${new Intl.NumberFormat("zh-CN", {
+    maximumFractionDigits: value < 10 ? 2 : value < 100 ? 1 : 0,
+  }).format(value)} tps`;
 }
 
 function formatCompactNumber(value: number) {
-  return new Intl.NumberFormat('zh-CN', { notation: 'compact', maximumFractionDigits: 1 }).format(value)
+  return new Intl.NumberFormat("zh-CN", {
+    notation: "compact",
+    maximumFractionDigits: 1,
+  }).format(value);
 }
 
 function timelineSegments(monitor: ModelStatusMonitor) {
-  const timeline = Array.isArray(monitor.timeline) ? monitor.timeline : []
-  if (timeline.length > 0) return timeline
-  return Array.from({ length: 60 }, () => 'unknown')
+  const timeline = Array.isArray(monitor.timeline) ? monitor.timeline : [];
+  if (timeline.length > 0) return timeline;
+  return Array.from({ length: 60 }, () => "unknown");
 }
 
 function getTimelineColor(status: string) {
   switch (status) {
-    case 'healthy':
-      return 'bg-green-500/85 dark:bg-green-400/90'
-    case 'warning':
-      return 'bg-amber-400/85 dark:bg-amber-300/85'
-    case 'unhealthy':
-      return 'bg-red-500/85 dark:bg-red-400/90'
+    case "healthy":
+      return "bg-green-500/85 dark:bg-green-400/90";
+    case "warning":
+      return "bg-amber-400/85 dark:bg-amber-300/85";
+    case "unhealthy":
+      return "bg-red-500/85 dark:bg-red-400/90";
     default:
-      return 'bg-gray-300 dark:bg-gray-600'
+      return "bg-gray-300 dark:bg-gray-600";
   }
 }
 
 function getTimelineLabel(status: string) {
   switch (status) {
-    case 'healthy':
-      return '健康'
-    case 'warning':
-      return '波动'
-    case 'unhealthy':
-      return '异常'
+    case "healthy":
+      return "健康";
+    case "warning":
+      return "波动";
+    case "unhealthy":
+      return "异常";
     default:
-      return '无请求'
+      return "无请求";
   }
 }
 
-function buildTimelineTooltip(monitor: ModelStatusMonitor, status: string, index: number) {
-  const segmentCount = timelineSegments(monitor).length
-  const startMs = monitor.time_range_start ? new Date(monitor.time_range_start).getTime() : Date.now() - parseInt(lookbackHours.value) * 60 * 60 * 1000
-  const endMs = monitor.time_range_end ? new Date(monitor.time_range_end).getTime() : Date.now()
-  const interval = Math.max(endMs - startMs, 1) / segmentCount
-  const cellStart = new Date(startMs + index * interval).toISOString()
-  const cellEnd = new Date(startMs + (index + 1) * interval).toISOString()
-  return `${formatTimestamp(cellStart)} - ${formatTimestamp(cellEnd)}\n模型：${monitor.model}\n状态：${getTimelineLabel(status)}`
+function buildTimelineTooltip(
+  monitor: ModelStatusMonitor,
+  status: string,
+  index: number,
+) {
+  const segmentCount = timelineSegments(monitor).length;
+  const startMs = monitor.time_range_start
+    ? new Date(monitor.time_range_start).getTime()
+    : Date.now() - parseInt(lookbackHours.value) * 60 * 60 * 1000;
+  const endMs = monitor.time_range_end
+    ? new Date(monitor.time_range_end).getTime()
+    : Date.now();
+  const interval = Math.max(endMs - startMs, 1) / segmentCount;
+  const cellStart = new Date(startMs + index * interval).toISOString();
+  const cellEnd = new Date(startMs + (index + 1) * interval).toISOString();
+  return `${formatTimestamp(cellStart)} - ${formatTimestamp(cellEnd)}\n模型：${monitor.model}\n状态：${getTimelineLabel(status)}`;
 }
 
 function formatTimestamp(timestamp?: string | null) {
-  if (!timestamp) return '未知时间'
-  const date = new Date(timestamp)
-  if (Number.isNaN(date.getTime())) return '未知时间'
-  return date.toLocaleString('zh-CN', {
-    month: '2-digit',
-    day: '2-digit',
-    hour: '2-digit',
-    minute: '2-digit',
-    second: '2-digit'
-  })
+  if (!timestamp) return "未知时间";
+  const date = new Date(timestamp);
+  if (Number.isNaN(date.getTime())) return "未知时间";
+  return date.toLocaleString("zh-CN", {
+    month: "2-digit",
+    day: "2-digit",
+    hour: "2-digit",
+    minute: "2-digit",
+    second: "2-digit",
+  });
 }
 
 watch(lookbackHours, () => {
-  loadMonitors()
-})
+  loadMonitors();
+});
 
 onMounted(() => {
-  refreshData()
-})
+  refreshData();
+});
 </script>

frontend/src/features/providers/components/ProviderHealthMonitorCard.vue

@@ -109,15 +109,19 @@
                 </div>
               </div>
 
-              <div class="grid w-full grid-cols-3 gap-2 lg:max-w-xl">
+              <div class="grid w-full grid-cols-4 gap-2 lg:max-w-2xl">
                 <MetricBox
-                  label="延迟"
+                  label="平均耗时"
                   :value="formatMs(provider.avg_latency_ms)"
                 />
                 <MetricBox
-                  label="Ping"
+                  label="平均TTFB"
                   :value="formatMs(provider.avg_first_byte_ms)"
                 />
+                <MetricBox
+                  label="平均速度"
+                  :value="formatTps(provider.avg_tps)"
+                />
                 <MetricBox
                   label="可用率"
                   :value="formatPercent(provider.success_rate)"
@@ -161,15 +165,19 @@
                   </Badge>
                 </div>
 
-                <div class="mt-4 grid grid-cols-3 gap-2">
+                <div class="mt-4 grid grid-cols-4 gap-2">
                   <MetricBox
-                    label="延迟"
+                    label="平均耗时"
                     :value="formatMs(model.avg_latency_ms)"
                   />
                   <MetricBox
-                    label="Ping"
+                    label="平均TTFB"
                     :value="formatMs(model.avg_first_byte_ms)"
                   />
+                  <MetricBox
+                    label="平均速度"
+                    :value="formatTps(model.avg_tps)"
+                  />
                   <MetricBox
                     label="可用率"
                     :value="formatPercent(model.success_rate)"
@@ -253,7 +261,7 @@ const MetricBox = defineComponent({
   },
   setup(props) {
     return () => h('div', { class: 'rounded-lg border border-border/40 bg-muted/20 px-3 py-2' }, [
-      h('div', { class: 'text-[11px] text-muted-foreground' }, props.label),
+      h('div', { class: 'text-[11px] leading-tight text-muted-foreground' }, props.label),
       h('div', { class: ['mt-1 text-sm font-semibold tabular-nums', props.valueClass] }, props.value)
     ])
   }
@@ -356,6 +364,13 @@ function formatPercent(value: number) {
   return `${(value * 100).toFixed(2)}%`
 }
 
+function formatTps(value?: number | null) {
+  if (typeof value !== 'number' || Number.isNaN(value)) return '-'
+  return `${new Intl.NumberFormat('zh-CN', {
+    maximumFractionDigits: value < 10 ? 2 : value < 100 ? 1 : 0
+  }).format(value)} tps`
+}
+
 function formatCompactNumber(value: number) {
   return new Intl.NumberFormat('zh-CN', { notation: 'compact', maximumFractionDigits: 1 }).format(value)
 }

frontend/src/mocks/handler.ts

@@ -272,6 +272,7 @@ const MOCK_MODEL_STATUS = {
       success_rate: 0.9896,
       avg_latency_ms: 1736,
       avg_first_byte_ms: 176,
+      avg_tps: 84.6,
       provider_count: 3,
       last_event_at: new Date().toISOString(),
       events: generateHealthEvents(60, 0.989, 0.008, 0.003, 1600, 460),
@@ -288,6 +289,7 @@ const MOCK_MODEL_STATUS = {
       success_rate: 0.9751,
       avg_latency_ms: 1280,
       avg_first_byte_ms: 221,
+      avg_tps: 76.2,
       provider_count: 2,
       last_event_at: new Date().toISOString(),
       events: generateHealthEvents(60, 0.975, 0.02, 0.005, 1200, 520),
@@ -304,6 +306,7 @@ const MOCK_MODEL_STATUS = {
       success_rate: 0.9517,
       avg_latency_ms: 940,
       avg_first_byte_ms: 184,
+      avg_tps: 101.8,
       provider_count: 2,
       last_event_at: new Date().toISOString(),
       events: generateHealthEvents(55, 0.952, 0.04, 0.008, 860, 300),
@@ -320,6 +323,7 @@ const MOCK_MODEL_STATUS = {
       success_rate: 0.835,
       avg_latency_ms: 2310,
       avg_first_byte_ms: 420,
+      avg_tps: 38.4,
       provider_count: 1,
       last_event_at: new Date().toISOString(),
       events: generateHealthEvents(45, 0.835, 0.145, 0.02, 2200, 780),
@@ -344,6 +348,7 @@ const MOCK_PROVIDER_HEALTH_STATUS = {
       success_rate: 0.9896,
       avg_latency_ms: 1736,
       avg_first_byte_ms: 176,
+      avg_tps: 84.6,
       model_count: 2,
       last_event_at: new Date().toISOString(),
       timeline: generateHealthTimeline(0.9, 0.05),
@@ -362,6 +367,7 @@ const MOCK_PROVIDER_HEALTH_STATUS = {
       success_rate: 0.9751,
       avg_latency_ms: 1280,
       avg_first_byte_ms: 221,
+      avg_tps: 76.2,
       model_count: 1,
       last_event_at: new Date().toISOString(),
       timeline: generateHealthTimeline(0.84, 0.09),
@@ -380,6 +386,7 @@ const MOCK_PROVIDER_HEALTH_STATUS = {
       success_rate: 0.9517,
       avg_latency_ms: 940,
       avg_first_byte_ms: 184,
+      avg_tps: 101.8,
       model_count: 1,
       last_event_at: new Date().toISOString(),
       timeline: generateHealthTimeline(0.78, 0.14),
@@ -398,6 +405,7 @@ const MOCK_PROVIDER_HEALTH_STATUS = {
       success_rate: 1,
       avg_latency_ms: null,
       avg_first_byte_ms: null,
+      avg_tps: null,
       model_count: 0,
       last_event_at: null,
       timeline: Array.from({ length: 60 }, () => 'unknown'),
@@ -1584,6 +1592,7 @@ const mockHandlers: Record<string, (config: AxiosRequestConfig) => Promise<Axios
         success_rate: model.success_rate,
         avg_latency_ms: model.avg_latency_ms,
         avg_first_byte_ms: model.avg_first_byte_ms,
+        avg_tps: model.avg_tps,
         last_event_at: model.last_event_at,
         events: model.events.slice(0, 10),
         timeline: model.timeline,